• Everything AMD Revealed at Its Computex 2025 Press Conference in 19 Minutes

    AMD was at Computex in Taipei this week to introduce some new hardware that PC builders might want to consider for their desktop builds.The first major announcement is new graphics cards, AMD’s Radeon RX 9060 XT, which will launch on June 5. The company is trying to make waves by pricing the 9060XT at a surprisingly low for the 8GB model and for 16GB version. "The fastest graphics card available under ” AMD SVP Jack Huynh said at the company's press conference.The RX 9060 XT cards have been designed for 1440p gaming. The products also outperform Nvidia’s slightly pricier RTX 5060 Ti by about 6% across 40 games when measured without AI-powered upscaling or frame generation, Huynh said. Stay tuned for our review. Still, we’re skeptical AMD can maintain the low price, considering the company’s other graphics card, the RX 9070XT, has been selling closer to to rather than the original starting price. Trump’s tariffs, low supplies, and greed from AMD’s GPU partners have all been blamed as factors for the price increases. Recommended by Our EditorsThe second major announcement involved new heavy-duty Threadripper 9000 CPUs meant to to power high-end workstation PCs and home desktop builds that want to go beyond a mere 16 cores. The new Threadrippers have been upgraded over the older 7000 series by using a 4-nanometer Zen 5 architecture. The Threadripper Pro 9000 series is meant for corporate workstation, with the most powerful chip, the 9995WX, containing a whopping 96 CPU cores and 192 threads. “There’s no contest here. This is absolute workload domination,” Huynh said while comparing the 9995WX against Intel’s Xeon W9-3595X chip. The company will also release Threadripper 9000 processors for home PCs with three CPU models that’ll offer 24, 32 or 64 CPU cores. Although no pricing was announced, expect the home desktop versions to cost from to based on the pricing for the 7000 series. The new Threadripper chips are slated to arrive in July.
    #everything #amd #revealed #its #computex
    Everything AMD Revealed at Its Computex 2025 Press Conference in 19 Minutes
    AMD was at Computex in Taipei this week to introduce some new hardware that PC builders might want to consider for their desktop builds.The first major announcement is new graphics cards, AMD’s Radeon RX 9060 XT, which will launch on June 5. The company is trying to make waves by pricing the 9060XT at a surprisingly low for the 8GB model and for 16GB version. "The fastest graphics card available under ” AMD SVP Jack Huynh said at the company's press conference.The RX 9060 XT cards have been designed for 1440p gaming. The products also outperform Nvidia’s slightly pricier RTX 5060 Ti by about 6% across 40 games when measured without AI-powered upscaling or frame generation, Huynh said. Stay tuned for our review. Still, we’re skeptical AMD can maintain the low price, considering the company’s other graphics card, the RX 9070XT, has been selling closer to to rather than the original starting price. Trump’s tariffs, low supplies, and greed from AMD’s GPU partners have all been blamed as factors for the price increases. Recommended by Our EditorsThe second major announcement involved new heavy-duty Threadripper 9000 CPUs meant to to power high-end workstation PCs and home desktop builds that want to go beyond a mere 16 cores. The new Threadrippers have been upgraded over the older 7000 series by using a 4-nanometer Zen 5 architecture. The Threadripper Pro 9000 series is meant for corporate workstation, with the most powerful chip, the 9995WX, containing a whopping 96 CPU cores and 192 threads. “There’s no contest here. This is absolute workload domination,” Huynh said while comparing the 9995WX against Intel’s Xeon W9-3595X chip. The company will also release Threadripper 9000 processors for home PCs with three CPU models that’ll offer 24, 32 or 64 CPU cores. Although no pricing was announced, expect the home desktop versions to cost from to based on the pricing for the 7000 series. The new Threadripper chips are slated to arrive in July. #everything #amd #revealed #its #computex
    ME.PCMAG.COM
    Everything AMD Revealed at Its Computex 2025 Press Conference in 19 Minutes
    AMD was at Computex in Taipei this week to introduce some new hardware that PC builders might want to consider for their desktop builds.The first major announcement is new graphics cards, AMD’s Radeon RX 9060 XT, which will launch on June 5. The company is trying to make waves by pricing the 9060XT at a surprisingly low $299 for the 8GB model and $349 for 16GB version. "The fastest graphics card available under $350,” AMD SVP Jack Huynh said at the company's press conference.(Credit: PCMag/Matthew Buzzi)The RX 9060 XT cards have been designed for 1440p gaming. The products also outperform Nvidia’s slightly pricier RTX 5060 Ti by about 6% across 40 games when measured without AI-powered upscaling or frame generation, Huynh said. Stay tuned for our review. Still, we’re skeptical AMD can maintain the low price, considering the company’s other graphics card, the RX 9070XT, has been selling closer to $729 to $899, rather than the original $599 starting price. Trump’s tariffs, low supplies, and greed from AMD’s GPU partners have all been blamed as factors for the price increases. Recommended by Our Editors(Credit: PCMag/Matthew Buzzi)The second major announcement involved new heavy-duty Threadripper 9000 CPUs meant to to power high-end workstation PCs and home desktop builds that want to go beyond a mere 16 cores. The new Threadrippers have been upgraded over the older 7000 series by using a 4-nanometer Zen 5 architecture. The Threadripper Pro 9000 series is meant for corporate workstation, with the most powerful chip, the 9995WX, containing a whopping 96 CPU cores and 192 threads. “There’s no contest here. This is absolute workload domination,” Huynh said while comparing the 9995WX against Intel’s Xeon W9-3595X chip. The company will also release Threadripper 9000 processors for home PCs with three CPU models that’ll offer 24, 32 or 64 CPU cores. Although no pricing was announced, expect the home desktop versions to cost from $1,500 to $5,000, based on the pricing for the 7000 series. The new Threadripper chips are slated to arrive in July.
    0 Комментарии 0 Поделились 0 предпросмотр
  • AMD launches new Ryzen Threadripper CPUs to challenge Intel’s workstation dominance

    Marking an aggressive push into the professional workstation and high-end desktop segments, AMD has launched its latest high-performance computing processors at Computex 2025.

    The processors are purpose-built to handle tough workloads such as VFX rendering, scientific simulation, CAD, and AI development for enterprise-grade workstations in industries such as engineering, healthcare, defense, and AI development. With the Ryzen Threadripper Pro 9000 WX-Series and Ryzen Threadripper 9000 Series, AMD is eyeing to compete against Intel’s Xeon W-3500 and Xeon W-2500, the long-dominant in enterprise workstations.
    #amd #launches #new #ryzen #threadripper
    AMD launches new Ryzen Threadripper CPUs to challenge Intel’s workstation dominance
    Marking an aggressive push into the professional workstation and high-end desktop segments, AMD has launched its latest high-performance computing processors at Computex 2025. The processors are purpose-built to handle tough workloads such as VFX rendering, scientific simulation, CAD, and AI development for enterprise-grade workstations in industries such as engineering, healthcare, defense, and AI development. With the Ryzen Threadripper Pro 9000 WX-Series and Ryzen Threadripper 9000 Series, AMD is eyeing to compete against Intel’s Xeon W-3500 and Xeon W-2500, the long-dominant in enterprise workstations. #amd #launches #new #ryzen #threadripper
    WWW.NETWORKWORLD.COM
    AMD launches new Ryzen Threadripper CPUs to challenge Intel’s workstation dominance
    Marking an aggressive push into the professional workstation and high-end desktop (HEDT) segments, AMD has launched its latest high-performance computing processors at Computex 2025. The processors are purpose-built to handle tough workloads such as VFX rendering, scientific simulation, CAD, and AI development for enterprise-grade workstations in industries such as engineering, healthcare, defense, and AI development. With the Ryzen Threadripper Pro 9000 WX-Series and Ryzen Threadripper 9000 Series, AMD is eyeing to compete against Intel’s Xeon W-3500 and Xeon W-2500, the long-dominant in enterprise workstations.
    0 Комментарии 0 Поделились 0 предпросмотр
  • RT Xeo ¯³: Illustrations by Japanese artist Katsuhiro Otomo for the anime titled "Akira"

    RT Xeo ¯³Illustrations by Japanese artist Katsuhiro Otomo for the anime titled "Akira"
    #xeo #illustrations #japanese #artist #katsuhiro
    RT Xeo ¯³: Illustrations by Japanese artist Katsuhiro Otomo for the anime titled "Akira"
    RT Xeo ¯³Illustrations by Japanese artist Katsuhiro Otomo for the anime titled "Akira" #xeo #illustrations #japanese #artist #katsuhiro
    X.COM
    RT Xeo ¯³: Illustrations by Japanese artist Katsuhiro Otomo for the anime titled "Akira"
    RT Xeo ¯³Illustrations by Japanese artist Katsuhiro Otomo for the anime titled "Akira"
    0 Комментарии 0 Поделились 0 предпросмотр
  • How Dell’s AI Infrastructure Updates Deliver Choice, Control And Scale

    Dell Technologies World focuses its keynote on Inventing the Future with AIDell Technologies
    Dell Technologies unveiled a significant expansion of its Dell AI Factory platform at its annual Dell Technologies World conference today, announcing over 40 product enhancements designed to help enterprises deploy artificial intelligence workloads more efficiently across both on-premises environments and cloud systems.

    The Dell AI Factory is not a physical manufacturing facility but a comprehensive framework combining advanced infrastructure, validated solutions, services, and an open ecosystem to help businesses harness the full potential of artificial intelligence across diverse environments—from data centers and cloud to edge locations and AI PCs.

    The company has attracted over 3,000 AI Factory customers since launching the platform last year. In an earlier call with industry analysts, Dell shared research stating that 79% of production AI workloads are running outside of public cloud environments—a trend driven by cost, security, and data governance concerns. During the keynote, Michael Dell provided more color on the value of Dell’s AI factory concept. He said, "The Dell AI factory is up to 60% more cost effective than the public cloud, and recent studies indicate that about three-fourths of AI initiatives are meeting or exceeding expectations. That means organizations are driving ROI and productivity gains from 20% to 40% in some cases.

    Making AI Easier to Deploy
    Organizations need the freedom to run AI workloads wherever makes the most sense for their business, without sacrificing performance or control. While IT leaders embraced the public cloud for their initial AI services, many organizations are now looking for a more nuanced approach where the company can control over their most critical AI assets while maintaining the flexibility to use cloud resources when appropriate. Over 80 percent of the companies Lopez Research interviewed said they struggled to find the budget and technical talent to deploy AI. These AI deployment challenges have only increased as more AI models and AI infrastructure services were launched.
    Silicon Diversity and Customer Choice
    A central theme of Dell's AI Factory message is how Dell makes AI easier to deploy while delivering choice. Dell is offering customers choice through silicon diversity in its designs, but also with ISV models. The company announced it has added Intel to its AI Factory portfolio with Intel Gaudi 3 AI accelerators and Intel Xeon processors, with a strong focus on inferencing workloads.
    Dell also announced its fourth update to the Dell AI Platform with AMD, rolling out two new PowerEdge servers—the XE9785 and the XE9785L—equipped with the latest AMD Instinct MI350 series GPUs. The Dell AI Factory with NVIDIA combines Dell's infrastructure with NVIDIA's AI software and GPU technologies to deliver end-to-end solutions that can reduce setup time by up to 86% compared to traditional approaches. The company also continues to strengthen its partnership with NVIDIA, announcing products that leverage NVIDIA's Blackwell family and other updates launched at NVIDIA GTC. As of today, Dell supports choice by delivering AI solutions with all of the primary GPU and AI accelerator infrastructure providers.
    Client-Side AI Advancements
    At the edge of the AI Factory ecosystem, Dell announced enhancements to the Dell Pro Max in a mobile form factor, leveraging Qualcomm’s AI 100 discrete NPUs designed for AI engineers and data scientists who need fast inferencing capabilities. With up to 288 TOPs at 16-bit floating point precision, these devices can power up to a 70-billion parameter model, delivering 7x the inferencing speed and 4x the accuracy over a 40 TOPs NPU. According to Dell, the Pro Max Plus line can run a 109-billion parameter AI model.
    The Dell Pro Max Plus targets AI engineers and data scientist with with the ability to run a ... More 109-billion parameter modelDell Technologies
    The Pro Max and Plus launches follow Dell's previous announcement of AI PCs featuring Dell Pro Max with GB 10 and GB 300 processors powered by NVIDIA's Grace Blackwell architecture. Overall, Dell has simplified its PC portfolio but made it easier for customers to choose the right system for their workloads by providing the latest chips from AMD, Intel, Nvidia, and Qualcomm.
    On-Premise AI Deployment Gains Ecosystem Momentum
    Following the theme of choice, organizations need the flexibility to run AI workloads on-premises and in the cloud. Dell is making significant strides in enabling on-premise AI deployments with major software partners. The company announced it is the first provider to bring Cohere capabilities on-premises, combining Cohere's generative AI models with Dell's secure, scalable infrastructure for turnkey enterprise solutions.
    Similar partnerships with Mistral and Glean were also announced, with Dell facilitating their first on-premise deployments. Additionally, Dell is supporting Google's Gemini on-premises with Google Distributed Cloud.
    To simplify model deployment, Dell now offers customers the ability to choose models on Hugging Face and deploy them in an automated fashion using containers and scripts. Enterprises increasingly recognize that while public cloud AI has its place, a hybrid AI infrastructure approach could deliver better economics and security for production workloads.
    The imperative for scalable yet efficient AI infrastructure at the edge is a growing need. As Michael Dell said during his Dell Technologies World keynote, “Over 75% of enterprise data will soon be created and processed at the edge, and AI will follow that data; it's not the other way around. The future of AI will be decentralized, low latency, and hyper-efficient.”
    Dell's ability to offer robust hybrid and fully on-premises solutions for AI is proving to be a significant advantage as companies increasingly seek on-premises support and even potentially air-gapped solutions for their most sensitive AI workloads. Key industries adopting the Dell AI Factory include finance, retail, energy, and healthcare providers.
    Scaling AI Requires a Focus on Energy Efficiency
    Simplifying AI also requires product innovations that deliver cost-effective, energy-efficient technology. As AI workloads drive unprecedented power consumption, Dell has prioritized energy efficiency in its latest offerings. The company introduced the Dell PowerCool Enclosed Rear Door Heat Exchangerwith Dell Integrated Rack Controller. This new cooling solution captures nearly 100% of the heat coming from GPU-intensive workloads. This innovation lowers cooling energy requirements for a rack by 60%, allowing customers to deploy 16% more racks with the same power infrastructure.
    Dell's new systems are rated to operate at 32 to 37 degrees Celsius, supporting significantly warmer temperatures than traditional air-cooled or water-chilled systems, further reducing power consumption for cooling. The PowerEdge XE9785L now offers Dell liquid cooling for flexible power management. Even if a company isn't aiming for a specific sustainability goal, every organization wants to improve energy utilization.
    Early Adopter Use Cases Highlight AI's Opportunity
    With over 200 product enhancements to its AI Factory in just one year, Dell Technologies is positioning itself as a central player in the rapidly evolving enterprise AI infrastructure market. It offers the breadth of solutions and expertise organizations require to successfully implement production-grade AI systems in a secure and scalable fashion. However, none of this technology matters if enterprises can't find a way to create business value by adopting it. Fortunately, examples from the first wave of enterprise early adopters highlight ways AI can deliver meaningful returns in productivity and customer experience. Let's look at two use cases presented at Dell Tech World.Michael Dell, CEO of Dell Technologies, interviews Larry Feinsmith, the Managing Director and Head ... More of Global Tech Strategy, Innovation & Partnerships at JPMorgan Chase at Dell Technologies WorldDell Technologies
    The Power of LLMs in Finance at JPMorgan Chase
    JPMorgan Chase took the stage to make AI real from a customer’s perspective. The financial firm uses Dell's compute hardware, software-defined storage, client, and peripheral solutions. Larry Feinsmith, the Managing Director and Head of Global Tech Strategy, Innovation & Partnerships at JPMorgan Chase, said, "We have a hybrid, multi-cloud, multi-provider strategy. Our private cloud is an incredibly strategic asset for us. We still have many applications and data on-premises for resiliency, latency, and a variety of other benefits."
    Feinsmith also spoke of the company's Large Language Modelstrategy. He said, “Our strategy is to use a constellation of models, both foundational and open, which requires a tremendous amount of compute in our data centers, in the public cloud, and, of course, at the edge. The one constant thing, whether you're training models, fine-tuning models, or finding a great use case that has large-scale inferencing, is that they all will drive compute. We think Dell is incredibly well positioned to help JPMorgan Chase and other companies in their AI journey.”
    Feinsmith noted that using AI isn't new for JPMorgan Chase. For over a decade, JPMorgan Chase has leveraged various types of AI, such as machine learning models for fraud detection, personalization, and marketing operations. The company uses what Feinsmith called its LLM suite, which over 200,000 people at JPMorgan Chase use today. The generative AI application is used for QA summarization and content generation using JPMorgan Chase's data in a highly secure way. Next, it has used the LLM suite architecture to build applications for its financial advisors, contact center agents, and any employee interacting with its clients. Its third use case highlighted changes in the software development area. JPMorgan Chase rolled out code generation AI capabilities to over 40,000 engineers. It has achieved as much as 20% productivity in the code creation and expects to leverage AI throughout the software development life cycle. Going forward, the financial firm expects to use AI agents and reasoning models to execute complex business processes end-to-end.Seemantini Godbole, EVP and Chief Digital and Information Officer at Lowe's shares the retailers AI ... More strategy at Dell Technologies WorldDell Technologies.
    How AI Makes It Easier For Employees to Serve Customers at Lowe's
    Lowe's Home Improvement Stores provided another example of how companies are leveraging Dell Technology and AI to transform the customer and employee experience. Seemantini Godbole, EVP and Chief Digital and Information Officer at Lowe's, shared insights on designing the strategy for AI when she said, "How should we deploy AI? We wanted to do impactful and meaningful things. We did not want to die a death of 1000 pilots, and we organized our efforts across how we sell, how we shop, and how we work. How we sell was for our associates. How we shop is for our customers, and how we work is for our headquarters employees. For whatever reason, most companies have begun with their workforce in the headquarters. We said, No, we are going to put AI in the hands of 300,000 associates."
    For example, she described a generative AI companion app for store associates. "Every store associate now has on his or her zebra device a ChatGPT-like experience for home improvement.", said Godbole. Lowe's is also deploying computer vision algorithms at the edge to understand issues such as whether a customer in a particular aisle is waiting for help. The system will then send notifications to the associates in that department. Customers can also ask various home improvement questions, such as what paint finish to use in a bathroom, at Lowes.com/AI.
    Designing A World Where AI Infrastructure Delivers Human Opportunity
    Michael Dell said, "We are entering the age of ubiquitous intelligence, where AI becomes as essential as electricity, with AI, you can distill years of experience into instant insights, speeding up decisions and uncovering patterns in massive data. But it’s not here to replace humans. AI is a collaborator that frees your teams to do what they do best, to innovate, to imagine, and to solve the world's toughest problems."
    While there are many AI deployment challenges ahead, the customer examples shared at Dell Technologies World provide a glimpse into a world where AI infrastructure and services benefits both customers and employees. The challenge now is to do this sustainably and ethically at scale.
    #how #dells #infrastructure #updates #deliver
    How Dell’s AI Infrastructure Updates Deliver Choice, Control And Scale
    Dell Technologies World focuses its keynote on Inventing the Future with AIDell Technologies Dell Technologies unveiled a significant expansion of its Dell AI Factory platform at its annual Dell Technologies World conference today, announcing over 40 product enhancements designed to help enterprises deploy artificial intelligence workloads more efficiently across both on-premises environments and cloud systems. The Dell AI Factory is not a physical manufacturing facility but a comprehensive framework combining advanced infrastructure, validated solutions, services, and an open ecosystem to help businesses harness the full potential of artificial intelligence across diverse environments—from data centers and cloud to edge locations and AI PCs. The company has attracted over 3,000 AI Factory customers since launching the platform last year. In an earlier call with industry analysts, Dell shared research stating that 79% of production AI workloads are running outside of public cloud environments—a trend driven by cost, security, and data governance concerns. During the keynote, Michael Dell provided more color on the value of Dell’s AI factory concept. He said, "The Dell AI factory is up to 60% more cost effective than the public cloud, and recent studies indicate that about three-fourths of AI initiatives are meeting or exceeding expectations. That means organizations are driving ROI and productivity gains from 20% to 40% in some cases. Making AI Easier to Deploy Organizations need the freedom to run AI workloads wherever makes the most sense for their business, without sacrificing performance or control. While IT leaders embraced the public cloud for their initial AI services, many organizations are now looking for a more nuanced approach where the company can control over their most critical AI assets while maintaining the flexibility to use cloud resources when appropriate. Over 80 percent of the companies Lopez Research interviewed said they struggled to find the budget and technical talent to deploy AI. These AI deployment challenges have only increased as more AI models and AI infrastructure services were launched. Silicon Diversity and Customer Choice A central theme of Dell's AI Factory message is how Dell makes AI easier to deploy while delivering choice. Dell is offering customers choice through silicon diversity in its designs, but also with ISV models. The company announced it has added Intel to its AI Factory portfolio with Intel Gaudi 3 AI accelerators and Intel Xeon processors, with a strong focus on inferencing workloads. Dell also announced its fourth update to the Dell AI Platform with AMD, rolling out two new PowerEdge servers—the XE9785 and the XE9785L—equipped with the latest AMD Instinct MI350 series GPUs. The Dell AI Factory with NVIDIA combines Dell's infrastructure with NVIDIA's AI software and GPU technologies to deliver end-to-end solutions that can reduce setup time by up to 86% compared to traditional approaches. The company also continues to strengthen its partnership with NVIDIA, announcing products that leverage NVIDIA's Blackwell family and other updates launched at NVIDIA GTC. As of today, Dell supports choice by delivering AI solutions with all of the primary GPU and AI accelerator infrastructure providers. Client-Side AI Advancements At the edge of the AI Factory ecosystem, Dell announced enhancements to the Dell Pro Max in a mobile form factor, leveraging Qualcomm’s AI 100 discrete NPUs designed for AI engineers and data scientists who need fast inferencing capabilities. With up to 288 TOPs at 16-bit floating point precision, these devices can power up to a 70-billion parameter model, delivering 7x the inferencing speed and 4x the accuracy over a 40 TOPs NPU. According to Dell, the Pro Max Plus line can run a 109-billion parameter AI model. The Dell Pro Max Plus targets AI engineers and data scientist with with the ability to run a ... More 109-billion parameter modelDell Technologies The Pro Max and Plus launches follow Dell's previous announcement of AI PCs featuring Dell Pro Max with GB 10 and GB 300 processors powered by NVIDIA's Grace Blackwell architecture. Overall, Dell has simplified its PC portfolio but made it easier for customers to choose the right system for their workloads by providing the latest chips from AMD, Intel, Nvidia, and Qualcomm. On-Premise AI Deployment Gains Ecosystem Momentum Following the theme of choice, organizations need the flexibility to run AI workloads on-premises and in the cloud. Dell is making significant strides in enabling on-premise AI deployments with major software partners. The company announced it is the first provider to bring Cohere capabilities on-premises, combining Cohere's generative AI models with Dell's secure, scalable infrastructure for turnkey enterprise solutions. Similar partnerships with Mistral and Glean were also announced, with Dell facilitating their first on-premise deployments. Additionally, Dell is supporting Google's Gemini on-premises with Google Distributed Cloud. To simplify model deployment, Dell now offers customers the ability to choose models on Hugging Face and deploy them in an automated fashion using containers and scripts. Enterprises increasingly recognize that while public cloud AI has its place, a hybrid AI infrastructure approach could deliver better economics and security for production workloads. The imperative for scalable yet efficient AI infrastructure at the edge is a growing need. As Michael Dell said during his Dell Technologies World keynote, “Over 75% of enterprise data will soon be created and processed at the edge, and AI will follow that data; it's not the other way around. The future of AI will be decentralized, low latency, and hyper-efficient.” Dell's ability to offer robust hybrid and fully on-premises solutions for AI is proving to be a significant advantage as companies increasingly seek on-premises support and even potentially air-gapped solutions for their most sensitive AI workloads. Key industries adopting the Dell AI Factory include finance, retail, energy, and healthcare providers. Scaling AI Requires a Focus on Energy Efficiency Simplifying AI also requires product innovations that deliver cost-effective, energy-efficient technology. As AI workloads drive unprecedented power consumption, Dell has prioritized energy efficiency in its latest offerings. The company introduced the Dell PowerCool Enclosed Rear Door Heat Exchangerwith Dell Integrated Rack Controller. This new cooling solution captures nearly 100% of the heat coming from GPU-intensive workloads. This innovation lowers cooling energy requirements for a rack by 60%, allowing customers to deploy 16% more racks with the same power infrastructure. Dell's new systems are rated to operate at 32 to 37 degrees Celsius, supporting significantly warmer temperatures than traditional air-cooled or water-chilled systems, further reducing power consumption for cooling. The PowerEdge XE9785L now offers Dell liquid cooling for flexible power management. Even if a company isn't aiming for a specific sustainability goal, every organization wants to improve energy utilization. Early Adopter Use Cases Highlight AI's Opportunity With over 200 product enhancements to its AI Factory in just one year, Dell Technologies is positioning itself as a central player in the rapidly evolving enterprise AI infrastructure market. It offers the breadth of solutions and expertise organizations require to successfully implement production-grade AI systems in a secure and scalable fashion. However, none of this technology matters if enterprises can't find a way to create business value by adopting it. Fortunately, examples from the first wave of enterprise early adopters highlight ways AI can deliver meaningful returns in productivity and customer experience. Let's look at two use cases presented at Dell Tech World.Michael Dell, CEO of Dell Technologies, interviews Larry Feinsmith, the Managing Director and Head ... More of Global Tech Strategy, Innovation & Partnerships at JPMorgan Chase at Dell Technologies WorldDell Technologies The Power of LLMs in Finance at JPMorgan Chase JPMorgan Chase took the stage to make AI real from a customer’s perspective. The financial firm uses Dell's compute hardware, software-defined storage, client, and peripheral solutions. Larry Feinsmith, the Managing Director and Head of Global Tech Strategy, Innovation & Partnerships at JPMorgan Chase, said, "We have a hybrid, multi-cloud, multi-provider strategy. Our private cloud is an incredibly strategic asset for us. We still have many applications and data on-premises for resiliency, latency, and a variety of other benefits." Feinsmith also spoke of the company's Large Language Modelstrategy. He said, “Our strategy is to use a constellation of models, both foundational and open, which requires a tremendous amount of compute in our data centers, in the public cloud, and, of course, at the edge. The one constant thing, whether you're training models, fine-tuning models, or finding a great use case that has large-scale inferencing, is that they all will drive compute. We think Dell is incredibly well positioned to help JPMorgan Chase and other companies in their AI journey.” Feinsmith noted that using AI isn't new for JPMorgan Chase. For over a decade, JPMorgan Chase has leveraged various types of AI, such as machine learning models for fraud detection, personalization, and marketing operations. The company uses what Feinsmith called its LLM suite, which over 200,000 people at JPMorgan Chase use today. The generative AI application is used for QA summarization and content generation using JPMorgan Chase's data in a highly secure way. Next, it has used the LLM suite architecture to build applications for its financial advisors, contact center agents, and any employee interacting with its clients. Its third use case highlighted changes in the software development area. JPMorgan Chase rolled out code generation AI capabilities to over 40,000 engineers. It has achieved as much as 20% productivity in the code creation and expects to leverage AI throughout the software development life cycle. Going forward, the financial firm expects to use AI agents and reasoning models to execute complex business processes end-to-end.Seemantini Godbole, EVP and Chief Digital and Information Officer at Lowe's shares the retailers AI ... More strategy at Dell Technologies WorldDell Technologies. How AI Makes It Easier For Employees to Serve Customers at Lowe's Lowe's Home Improvement Stores provided another example of how companies are leveraging Dell Technology and AI to transform the customer and employee experience. Seemantini Godbole, EVP and Chief Digital and Information Officer at Lowe's, shared insights on designing the strategy for AI when she said, "How should we deploy AI? We wanted to do impactful and meaningful things. We did not want to die a death of 1000 pilots, and we organized our efforts across how we sell, how we shop, and how we work. How we sell was for our associates. How we shop is for our customers, and how we work is for our headquarters employees. For whatever reason, most companies have begun with their workforce in the headquarters. We said, No, we are going to put AI in the hands of 300,000 associates." For example, she described a generative AI companion app for store associates. "Every store associate now has on his or her zebra device a ChatGPT-like experience for home improvement.", said Godbole. Lowe's is also deploying computer vision algorithms at the edge to understand issues such as whether a customer in a particular aisle is waiting for help. The system will then send notifications to the associates in that department. Customers can also ask various home improvement questions, such as what paint finish to use in a bathroom, at Lowes.com/AI. Designing A World Where AI Infrastructure Delivers Human Opportunity Michael Dell said, "We are entering the age of ubiquitous intelligence, where AI becomes as essential as electricity, with AI, you can distill years of experience into instant insights, speeding up decisions and uncovering patterns in massive data. But it’s not here to replace humans. AI is a collaborator that frees your teams to do what they do best, to innovate, to imagine, and to solve the world's toughest problems." While there are many AI deployment challenges ahead, the customer examples shared at Dell Technologies World provide a glimpse into a world where AI infrastructure and services benefits both customers and employees. The challenge now is to do this sustainably and ethically at scale. #how #dells #infrastructure #updates #deliver
    WWW.FORBES.COM
    How Dell’s AI Infrastructure Updates Deliver Choice, Control And Scale
    Dell Technologies World focuses its keynote on Inventing the Future with AIDell Technologies Dell Technologies unveiled a significant expansion of its Dell AI Factory platform at its annual Dell Technologies World conference today, announcing over 40 product enhancements designed to help enterprises deploy artificial intelligence workloads more efficiently across both on-premises environments and cloud systems. The Dell AI Factory is not a physical manufacturing facility but a comprehensive framework combining advanced infrastructure, validated solutions, services, and an open ecosystem to help businesses harness the full potential of artificial intelligence across diverse environments—from data centers and cloud to edge locations and AI PCs. The company has attracted over 3,000 AI Factory customers since launching the platform last year. In an earlier call with industry analysts, Dell shared research stating that 79% of production AI workloads are running outside of public cloud environments—a trend driven by cost, security, and data governance concerns. During the keynote, Michael Dell provided more color on the value of Dell’s AI factory concept. He said, "The Dell AI factory is up to 60% more cost effective than the public cloud, and recent studies indicate that about three-fourths of AI initiatives are meeting or exceeding expectations. That means organizations are driving ROI and productivity gains from 20% to 40% in some cases. Making AI Easier to Deploy Organizations need the freedom to run AI workloads wherever makes the most sense for their business, without sacrificing performance or control. While IT leaders embraced the public cloud for their initial AI services, many organizations are now looking for a more nuanced approach where the company can control over their most critical AI assets while maintaining the flexibility to use cloud resources when appropriate. Over 80 percent of the companies Lopez Research interviewed said they struggled to find the budget and technical talent to deploy AI. These AI deployment challenges have only increased as more AI models and AI infrastructure services were launched. Silicon Diversity and Customer Choice A central theme of Dell's AI Factory message is how Dell makes AI easier to deploy while delivering choice. Dell is offering customers choice through silicon diversity in its designs, but also with ISV models. The company announced it has added Intel to its AI Factory portfolio with Intel Gaudi 3 AI accelerators and Intel Xeon processors, with a strong focus on inferencing workloads. Dell also announced its fourth update to the Dell AI Platform with AMD, rolling out two new PowerEdge servers—the XE9785 and the XE9785L—equipped with the latest AMD Instinct MI350 series GPUs. The Dell AI Factory with NVIDIA combines Dell's infrastructure with NVIDIA's AI software and GPU technologies to deliver end-to-end solutions that can reduce setup time by up to 86% compared to traditional approaches. The company also continues to strengthen its partnership with NVIDIA, announcing products that leverage NVIDIA's Blackwell family and other updates launched at NVIDIA GTC. As of today, Dell supports choice by delivering AI solutions with all of the primary GPU and AI accelerator infrastructure providers. Client-Side AI Advancements At the edge of the AI Factory ecosystem, Dell announced enhancements to the Dell Pro Max in a mobile form factor, leveraging Qualcomm’s AI 100 discrete NPUs designed for AI engineers and data scientists who need fast inferencing capabilities. With up to 288 TOPs at 16-bit floating point precision, these devices can power up to a 70-billion parameter model, delivering 7x the inferencing speed and 4x the accuracy over a 40 TOPs NPU. According to Dell, the Pro Max Plus line can run a 109-billion parameter AI model. The Dell Pro Max Plus targets AI engineers and data scientist with with the ability to run a ... More 109-billion parameter modelDell Technologies The Pro Max and Plus launches follow Dell's previous announcement of AI PCs featuring Dell Pro Max with GB 10 and GB 300 processors powered by NVIDIA's Grace Blackwell architecture. Overall, Dell has simplified its PC portfolio but made it easier for customers to choose the right system for their workloads by providing the latest chips from AMD, Intel, Nvidia, and Qualcomm. On-Premise AI Deployment Gains Ecosystem Momentum Following the theme of choice, organizations need the flexibility to run AI workloads on-premises and in the cloud. Dell is making significant strides in enabling on-premise AI deployments with major software partners. The company announced it is the first provider to bring Cohere capabilities on-premises, combining Cohere's generative AI models with Dell's secure, scalable infrastructure for turnkey enterprise solutions. Similar partnerships with Mistral and Glean were also announced, with Dell facilitating their first on-premise deployments. Additionally, Dell is supporting Google's Gemini on-premises with Google Distributed Cloud. To simplify model deployment, Dell now offers customers the ability to choose models on Hugging Face and deploy them in an automated fashion using containers and scripts. Enterprises increasingly recognize that while public cloud AI has its place, a hybrid AI infrastructure approach could deliver better economics and security for production workloads. The imperative for scalable yet efficient AI infrastructure at the edge is a growing need. As Michael Dell said during his Dell Technologies World keynote, “Over 75% of enterprise data will soon be created and processed at the edge, and AI will follow that data; it's not the other way around. The future of AI will be decentralized, low latency, and hyper-efficient.” Dell's ability to offer robust hybrid and fully on-premises solutions for AI is proving to be a significant advantage as companies increasingly seek on-premises support and even potentially air-gapped solutions for their most sensitive AI workloads. Key industries adopting the Dell AI Factory include finance, retail, energy, and healthcare providers. Scaling AI Requires a Focus on Energy Efficiency Simplifying AI also requires product innovations that deliver cost-effective, energy-efficient technology. As AI workloads drive unprecedented power consumption, Dell has prioritized energy efficiency in its latest offerings. The company introduced the Dell PowerCool Enclosed Rear Door Heat Exchanger (eRDHx) with Dell Integrated Rack Controller (IRC). This new cooling solution captures nearly 100% of the heat coming from GPU-intensive workloads. This innovation lowers cooling energy requirements for a rack by 60%, allowing customers to deploy 16% more racks with the same power infrastructure. Dell's new systems are rated to operate at 32 to 37 degrees Celsius, supporting significantly warmer temperatures than traditional air-cooled or water-chilled systems, further reducing power consumption for cooling. The PowerEdge XE9785L now offers Dell liquid cooling for flexible power management. Even if a company isn't aiming for a specific sustainability goal, every organization wants to improve energy utilization. Early Adopter Use Cases Highlight AI's Opportunity With over 200 product enhancements to its AI Factory in just one year, Dell Technologies is positioning itself as a central player in the rapidly evolving enterprise AI infrastructure market. It offers the breadth of solutions and expertise organizations require to successfully implement production-grade AI systems in a secure and scalable fashion. However, none of this technology matters if enterprises can't find a way to create business value by adopting it. Fortunately, examples from the first wave of enterprise early adopters highlight ways AI can deliver meaningful returns in productivity and customer experience. Let's look at two use cases presented at Dell Tech World.Michael Dell, CEO of Dell Technologies, interviews Larry Feinsmith, the Managing Director and Head ... More of Global Tech Strategy, Innovation & Partnerships at JPMorgan Chase at Dell Technologies WorldDell Technologies The Power of LLMs in Finance at JPMorgan Chase JPMorgan Chase took the stage to make AI real from a customer’s perspective. The financial firm uses Dell's compute hardware, software-defined storage, client, and peripheral solutions. Larry Feinsmith, the Managing Director and Head of Global Tech Strategy, Innovation & Partnerships at JPMorgan Chase, said, "We have a hybrid, multi-cloud, multi-provider strategy. Our private cloud is an incredibly strategic asset for us. We still have many applications and data on-premises for resiliency, latency, and a variety of other benefits." Feinsmith also spoke of the company's Large Language Model (LLM) strategy. He said, “Our strategy is to use a constellation of models, both foundational and open, which requires a tremendous amount of compute in our data centers, in the public cloud, and, of course, at the edge. The one constant thing, whether you're training models, fine-tuning models, or finding a great use case that has large-scale inferencing, is that they all will drive compute. We think Dell is incredibly well positioned to help JPMorgan Chase and other companies in their AI journey.” Feinsmith noted that using AI isn't new for JPMorgan Chase. For over a decade, JPMorgan Chase has leveraged various types of AI, such as machine learning models for fraud detection, personalization, and marketing operations. The company uses what Feinsmith called its LLM suite, which over 200,000 people at JPMorgan Chase use today. The generative AI application is used for QA summarization and content generation using JPMorgan Chase's data in a highly secure way. Next, it has used the LLM suite architecture to build applications for its financial advisors, contact center agents, and any employee interacting with its clients. Its third use case highlighted changes in the software development area. JPMorgan Chase rolled out code generation AI capabilities to over 40,000 engineers. It has achieved as much as 20% productivity in the code creation and expects to leverage AI throughout the software development life cycle. Going forward, the financial firm expects to use AI agents and reasoning models to execute complex business processes end-to-end.Seemantini Godbole, EVP and Chief Digital and Information Officer at Lowe's shares the retailers AI ... More strategy at Dell Technologies WorldDell Technologies. How AI Makes It Easier For Employees to Serve Customers at Lowe's Lowe's Home Improvement Stores provided another example of how companies are leveraging Dell Technology and AI to transform the customer and employee experience. Seemantini Godbole, EVP and Chief Digital and Information Officer at Lowe's, shared insights on designing the strategy for AI when she said, "How should we deploy AI? We wanted to do impactful and meaningful things. We did not want to die a death of 1000 pilots, and we organized our efforts across how we sell, how we shop, and how we work. How we sell was for our associates. How we shop is for our customers, and how we work is for our headquarters employees. For whatever reason, most companies have begun with their workforce in the headquarters. We said, No, we are going to put AI in the hands of 300,000 associates." For example, she described a generative AI companion app for store associates. "Every store associate now has on his or her zebra device a ChatGPT-like experience for home improvement.", said Godbole. Lowe's is also deploying computer vision algorithms at the edge to understand issues such as whether a customer in a particular aisle is waiting for help. The system will then send notifications to the associates in that department. Customers can also ask various home improvement questions, such as what paint finish to use in a bathroom, at Lowes.com/AI. Designing A World Where AI Infrastructure Delivers Human Opportunity Michael Dell said, "We are entering the age of ubiquitous intelligence, where AI becomes as essential as electricity, with AI, you can distill years of experience into instant insights, speeding up decisions and uncovering patterns in massive data. But it’s not here to replace humans. AI is a collaborator that frees your teams to do what they do best, to innovate, to imagine, and to solve the world's toughest problems." While there are many AI deployment challenges ahead, the customer examples shared at Dell Technologies World provide a glimpse into a world where AI infrastructure and services benefits both customers and employees. The challenge now is to do this sustainably and ethically at scale.
    0 Комментарии 0 Поделились 0 предпросмотр
  • Researchers Expose New Intel CPU Flaws Enabling Memory Leaks and Spectre v2 Attacks

    May 16, 2025Ravie LakshmananHardware Security / Vulnerability

    Researchers at ETH Zürich have discovered yet another security flaw that they say impacts all modern Intel CPUs and causes them to leak sensitive data from memory, showing that the vulnerability known as Spectre continues to haunt computer systems after more than seven years.
    The vulnerability, referred to as Branch Privilege Injection, "can be exploited to misuse the prediction calculations of the CPUin order to gain unauthorized access to information from other processor users," ETH Zurich said.
    Kaveh Razavi, head of the Computer Security Groupand one of the authors of the study, said the shortcoming affects all Intel processors, potentially enabling bad actors to read the contents of the processor's cache and the working memory of another user of the same CPU.

    The attack leverages what's called Branch Predictor Race Conditionsthat emerge when a processor switches between prediction calculations for two users with different permissions, opening the door to a scenario where an unprivileged hacker could exploit it to bypass security barriers and access confidential information from a privileged process.
    Intel has issued microcode patches to address the vulnerability, which has been assigned the CVE identifier CVE-2024-45332.
    "Exposure of sensitive information caused by shared microarchitectural predictor state that influences transient execution in the indirect branch predictors for some Intel Processors may allow an authenticated user to potentially enable information disclosure via local access," Intel said in an advisory released on May 13.
    The disclosure comes as researchers from the Systems and Network Security Groupat Vrije Universiteit Amsterdam detailed a category of self-training Spectre v2 attacks codenamed Training Solo.
    "Attackers can speculatively hijack control flow within the same domainand leak secrets across privilege boundaries, re-enabling classic Spectre v2 scenarios without relying on powerful sandboxed environments like eBPF," VUSec said.

    The hardware exploits, tracked as CVE-2024-28956 and CVE-2025-24495, can be used against Intel CPUs to leak kernel memory at up to 17 Kb/s, with the study finding that they could "completely break the domain isolation and re-enable traditional user-user, guest-guest, and even guest-host Spectre-v2 attacks."

    CVE-2024-28956- Indirect Target Selection, which affects Intel Core 9th-11th, and Intel Xeon 2nd-3rd, among others.
    CVE-2025-24495- Lion Cove BPU issue, which affects Intel CPUs with Lion Cove core

    While Intel has shipped microcode updates for these defects, AMD said it has revised its existing guidance on Spectre and Meltdown to explicitly highlight the risk from the use of classic Berkeley Packet Filter.

    Found this article interesting? Follow us on Twitter  and LinkedIn to read more exclusive content we post.

    SHARE




    #researchers #expose #new #intel #cpu
    Researchers Expose New Intel CPU Flaws Enabling Memory Leaks and Spectre v2 Attacks
    May 16, 2025Ravie LakshmananHardware Security / Vulnerability Researchers at ETH Zürich have discovered yet another security flaw that they say impacts all modern Intel CPUs and causes them to leak sensitive data from memory, showing that the vulnerability known as Spectre continues to haunt computer systems after more than seven years. The vulnerability, referred to as Branch Privilege Injection, "can be exploited to misuse the prediction calculations of the CPUin order to gain unauthorized access to information from other processor users," ETH Zurich said. Kaveh Razavi, head of the Computer Security Groupand one of the authors of the study, said the shortcoming affects all Intel processors, potentially enabling bad actors to read the contents of the processor's cache and the working memory of another user of the same CPU. The attack leverages what's called Branch Predictor Race Conditionsthat emerge when a processor switches between prediction calculations for two users with different permissions, opening the door to a scenario where an unprivileged hacker could exploit it to bypass security barriers and access confidential information from a privileged process. Intel has issued microcode patches to address the vulnerability, which has been assigned the CVE identifier CVE-2024-45332. "Exposure of sensitive information caused by shared microarchitectural predictor state that influences transient execution in the indirect branch predictors for some Intel Processors may allow an authenticated user to potentially enable information disclosure via local access," Intel said in an advisory released on May 13. The disclosure comes as researchers from the Systems and Network Security Groupat Vrije Universiteit Amsterdam detailed a category of self-training Spectre v2 attacks codenamed Training Solo. "Attackers can speculatively hijack control flow within the same domainand leak secrets across privilege boundaries, re-enabling classic Spectre v2 scenarios without relying on powerful sandboxed environments like eBPF," VUSec said. The hardware exploits, tracked as CVE-2024-28956 and CVE-2025-24495, can be used against Intel CPUs to leak kernel memory at up to 17 Kb/s, with the study finding that they could "completely break the domain isolation and re-enable traditional user-user, guest-guest, and even guest-host Spectre-v2 attacks." CVE-2024-28956- Indirect Target Selection, which affects Intel Core 9th-11th, and Intel Xeon 2nd-3rd, among others. CVE-2025-24495- Lion Cove BPU issue, which affects Intel CPUs with Lion Cove core While Intel has shipped microcode updates for these defects, AMD said it has revised its existing guidance on Spectre and Meltdown to explicitly highlight the risk from the use of classic Berkeley Packet Filter. Found this article interesting? Follow us on Twitter  and LinkedIn to read more exclusive content we post. SHARE     #researchers #expose #new #intel #cpu
    THEHACKERNEWS.COM
    Researchers Expose New Intel CPU Flaws Enabling Memory Leaks and Spectre v2 Attacks
    May 16, 2025Ravie LakshmananHardware Security / Vulnerability Researchers at ETH Zürich have discovered yet another security flaw that they say impacts all modern Intel CPUs and causes them to leak sensitive data from memory, showing that the vulnerability known as Spectre continues to haunt computer systems after more than seven years. The vulnerability, referred to as Branch Privilege Injection (BPI), "can be exploited to misuse the prediction calculations of the CPU (central processing unit) in order to gain unauthorized access to information from other processor users," ETH Zurich said. Kaveh Razavi, head of the Computer Security Group (COMSEC) and one of the authors of the study, said the shortcoming affects all Intel processors, potentially enabling bad actors to read the contents of the processor's cache and the working memory of another user of the same CPU. The attack leverages what's called Branch Predictor Race Conditions (BPRC) that emerge when a processor switches between prediction calculations for two users with different permissions, opening the door to a scenario where an unprivileged hacker could exploit it to bypass security barriers and access confidential information from a privileged process. Intel has issued microcode patches to address the vulnerability, which has been assigned the CVE identifier CVE-2024-45332 (CVSS v4 score: 5.7). "Exposure of sensitive information caused by shared microarchitectural predictor state that influences transient execution in the indirect branch predictors for some Intel Processors may allow an authenticated user to potentially enable information disclosure via local access," Intel said in an advisory released on May 13. The disclosure comes as researchers from the Systems and Network Security Group (VUSec) at Vrije Universiteit Amsterdam detailed a category of self-training Spectre v2 attacks codenamed Training Solo. "Attackers can speculatively hijack control flow within the same domain (e.g., kernel) and leak secrets across privilege boundaries, re-enabling classic Spectre v2 scenarios without relying on powerful sandboxed environments like eBPF," VUSec said. The hardware exploits, tracked as CVE-2024-28956 and CVE-2025-24495, can be used against Intel CPUs to leak kernel memory at up to 17 Kb/s, with the study finding that they could "completely break the domain isolation and re-enable traditional user-user, guest-guest, and even guest-host Spectre-v2 attacks." CVE-2024-28956 (CVSS v4 score: 5.7) - Indirect Target Selection (ITS), which affects Intel Core 9th-11th, and Intel Xeon 2nd-3rd, among others. CVE-2025-24495 (CVSS v4 score: 6.8) - Lion Cove BPU issue, which affects Intel CPUs with Lion Cove core While Intel has shipped microcode updates for these defects, AMD said it has revised its existing guidance on Spectre and Meltdown to explicitly highlight the risk from the use of classic Berkeley Packet Filter (cBPF). Found this article interesting? Follow us on Twitter  and LinkedIn to read more exclusive content we post. SHARE    
    0 Комментарии 0 Поделились 0 предпросмотр
  • Intel hopes its foundry business will break even around 2027

    TL;DR: Intel's new CEO, Lip-Bu Tan, made it clear he isn't backing away from the company's push to become a top-tier foundry. Now, Intel has shared new details about how – and when – it expects chip manufacturing to become a profitable part of the business.
    At a recent investor conference, Intel Chief Financial Officer David Zinsner stated that he expects the company's foundry business to break even by 2027. Within a few years, Chipzilla should turn a profit and begin building the trust needed to attract new customers.
    Intel's chip business isn't in the best shape right now. The company has embraced a multi-foundry approach, outsourcing part of its wafer production to Taiwanese foundry TSMC while developing new advanced manufacturing nodes internally. Intel has scheduled mass production of the recently unveiled 18A and 14A nodes for 2027.
    Tom's Hardware notes that Intel will first use the 18A nodeto manufacture Panther Lake client PC processors, with new consumer CPUs launching later this year. Intel will also apply the same technology to "Clearwater Forest" Xeon processors and some undisclosed third-party products. Nevertheless, the company sees the 18A node as a promising proof of concept to attract external customers.

    Zinsner expects larger third-party volumes to come from the 14A manufacturing node. Intel is currently partnering with potential customers, but the process remains challenging. Some clients leave after a few test chips, while others stick around without committing to significant production volumes. Intel still needs to prove it can operate as a reliable foundry business.
    Intel will use High-NA EUV lithography for the 14A node, initially increasing costs. The company hopes its advanced capabilities will eventually outweigh the expense. The foundry unit should also benefit from increased internal production, with both Panther Lake and Nova Lake processors set to be built entirely in-house.
    // Related Stories

    Intel believes its foundry business needs only a few billion dollars in additional external revenue to break even. The 14A node could see broader adoption among third-party customers, while more mature nodes like Intel 16 and partnerships with companies such as Tower and UMC will help diversify revenue sources.
    #intel #hopes #its #foundry #business
    Intel hopes its foundry business will break even around 2027
    TL;DR: Intel's new CEO, Lip-Bu Tan, made it clear he isn't backing away from the company's push to become a top-tier foundry. Now, Intel has shared new details about how – and when – it expects chip manufacturing to become a profitable part of the business. At a recent investor conference, Intel Chief Financial Officer David Zinsner stated that he expects the company's foundry business to break even by 2027. Within a few years, Chipzilla should turn a profit and begin building the trust needed to attract new customers. Intel's chip business isn't in the best shape right now. The company has embraced a multi-foundry approach, outsourcing part of its wafer production to Taiwanese foundry TSMC while developing new advanced manufacturing nodes internally. Intel has scheduled mass production of the recently unveiled 18A and 14A nodes for 2027. Tom's Hardware notes that Intel will first use the 18A nodeto manufacture Panther Lake client PC processors, with new consumer CPUs launching later this year. Intel will also apply the same technology to "Clearwater Forest" Xeon processors and some undisclosed third-party products. Nevertheless, the company sees the 18A node as a promising proof of concept to attract external customers. Zinsner expects larger third-party volumes to come from the 14A manufacturing node. Intel is currently partnering with potential customers, but the process remains challenging. Some clients leave after a few test chips, while others stick around without committing to significant production volumes. Intel still needs to prove it can operate as a reliable foundry business. Intel will use High-NA EUV lithography for the 14A node, initially increasing costs. The company hopes its advanced capabilities will eventually outweigh the expense. The foundry unit should also benefit from increased internal production, with both Panther Lake and Nova Lake processors set to be built entirely in-house. // Related Stories Intel believes its foundry business needs only a few billion dollars in additional external revenue to break even. The 14A node could see broader adoption among third-party customers, while more mature nodes like Intel 16 and partnerships with companies such as Tower and UMC will help diversify revenue sources. #intel #hopes #its #foundry #business
    WWW.TECHSPOT.COM
    Intel hopes its foundry business will break even around 2027
    TL;DR: Intel's new CEO, Lip-Bu Tan, made it clear he isn't backing away from the company's push to become a top-tier foundry. Now, Intel has shared new details about how – and when – it expects chip manufacturing to become a profitable part of the business. At a recent investor conference, Intel Chief Financial Officer David Zinsner stated that he expects the company's foundry business to break even by 2027. Within a few years, Chipzilla should turn a profit and begin building the trust needed to attract new customers. Intel's chip business isn't in the best shape right now. The company has embraced a multi-foundry approach, outsourcing part of its wafer production to Taiwanese foundry TSMC while developing new advanced manufacturing nodes internally. Intel has scheduled mass production of the recently unveiled 18A and 14A nodes for 2027. Tom's Hardware notes that Intel will first use the 18A node (1.8nm) to manufacture Panther Lake client PC processors, with new consumer CPUs launching later this year. Intel will also apply the same technology to "Clearwater Forest" Xeon processors and some undisclosed third-party products. Nevertheless, the company sees the 18A node as a promising proof of concept to attract external customers. Zinsner expects larger third-party volumes to come from the 14A manufacturing node. Intel is currently partnering with potential customers, but the process remains challenging. Some clients leave after a few test chips, while others stick around without committing to significant production volumes. Intel still needs to prove it can operate as a reliable foundry business. Intel will use High-NA EUV lithography for the 14A node, initially increasing costs. The company hopes its advanced capabilities will eventually outweigh the expense. The foundry unit should also benefit from increased internal production, with both Panther Lake and Nova Lake processors set to be built entirely in-house. // Related Stories Intel believes its foundry business needs only a few billion dollars in additional external revenue to break even. The 14A node could see broader adoption among third-party customers, while more mature nodes like Intel 16 and partnerships with companies such as Tower and UMC will help diversify revenue sources.
    0 Комментарии 0 Поделились 0 предпросмотр
  • John Carmack suggests the world could run on older hardware – if we optimized software better

    In context: Google researcher and reverse engineer "LaurieWired" recently posed a thought-provoking thread on X: What would happen after a CPU manufacturing apocalypse? How would the tech world respond to a future without newer, faster processors? Programming and optimization legend John Carmack offered an equally compelling answer.
    LaurieWired proposes the idea of a "Zero Tape-out Day", an event causing manufacturers to stop producing new silicon designs. Considering the existing supply, the researcher predicts skyrocketing computer prices, stalled cloud capacity, and a ticking clock on electromigration slowly degrading the most advanced chips built on smaller nodes – all within the first year after Z-Day.
    Conditions would deteriorate even further in the following years, with a booming black market for processors and Xeon CPUs valued more than gold. Computing technology could regress by decades as older systems built on larger nodes prove far more resilient to electromigration.
    People would mod classic processors like the Motorola 68000 to operate for thousands of years without significant gate wear. More advanced systems – such as the iMac G3s sold between 1998 and 2003 – would become workstations for the elite, while the proles use repurpose hardware from Gameboys, Macintosh SEs, and Commodore 64s.
    LaurieWired suggests that 30 years after Z-Day, the world would become a dystopia where computing resembles the 1970s or 1980s. The modern internet would vanish, replaced by sneakernet data exchanges on SSDs and efforts to safeguard valuable desktop hardware from confiscation.
    Former id Software developer John Carmack decided to weigh in on the thought experiment. Having created the Doom graphics engine in just 28 hours on "vintage hardware," his expertise provided some perspective. Carmack said that a significant part of the modern world could run on outdated hardware if software optimization were a priority for developers.
    // Related Stories

    The god-tier coder suggests that developers could transition all interpreted, microservice-based products to monolithic, native codebases. Programmers would abandon modern development patterns and seek more efficient approaches, such as those used during earlier computing eras when there was no internet to push patches.
    Such a paradigm reset would force post-apocalyptic coders to make ancient hardware hum through software optimization. Carmack also acknowledges that innovative new products would become much rarer without ultra-cheap and scalable computing.
    While framed within the context of LaurieWired's thought experiment, Carmack's ideas hold practical relevance in today's computing landscape. For example, would Microsoft still impose strict hardware requirements if it prioritized optimizing Windows 11? It's a question worth considering. Similarly, how much could the gaming industry benefit from better optimization?
    #john #carmack #suggests #world #could
    John Carmack suggests the world could run on older hardware – if we optimized software better
    In context: Google researcher and reverse engineer "LaurieWired" recently posed a thought-provoking thread on X: What would happen after a CPU manufacturing apocalypse? How would the tech world respond to a future without newer, faster processors? Programming and optimization legend John Carmack offered an equally compelling answer. LaurieWired proposes the idea of a "Zero Tape-out Day", an event causing manufacturers to stop producing new silicon designs. Considering the existing supply, the researcher predicts skyrocketing computer prices, stalled cloud capacity, and a ticking clock on electromigration slowly degrading the most advanced chips built on smaller nodes – all within the first year after Z-Day. Conditions would deteriorate even further in the following years, with a booming black market for processors and Xeon CPUs valued more than gold. Computing technology could regress by decades as older systems built on larger nodes prove far more resilient to electromigration. People would mod classic processors like the Motorola 68000 to operate for thousands of years without significant gate wear. More advanced systems – such as the iMac G3s sold between 1998 and 2003 – would become workstations for the elite, while the proles use repurpose hardware from Gameboys, Macintosh SEs, and Commodore 64s. LaurieWired suggests that 30 years after Z-Day, the world would become a dystopia where computing resembles the 1970s or 1980s. The modern internet would vanish, replaced by sneakernet data exchanges on SSDs and efforts to safeguard valuable desktop hardware from confiscation. Former id Software developer John Carmack decided to weigh in on the thought experiment. Having created the Doom graphics engine in just 28 hours on "vintage hardware," his expertise provided some perspective. Carmack said that a significant part of the modern world could run on outdated hardware if software optimization were a priority for developers. // Related Stories The god-tier coder suggests that developers could transition all interpreted, microservice-based products to monolithic, native codebases. Programmers would abandon modern development patterns and seek more efficient approaches, such as those used during earlier computing eras when there was no internet to push patches. Such a paradigm reset would force post-apocalyptic coders to make ancient hardware hum through software optimization. Carmack also acknowledges that innovative new products would become much rarer without ultra-cheap and scalable computing. While framed within the context of LaurieWired's thought experiment, Carmack's ideas hold practical relevance in today's computing landscape. For example, would Microsoft still impose strict hardware requirements if it prioritized optimizing Windows 11? It's a question worth considering. Similarly, how much could the gaming industry benefit from better optimization? #john #carmack #suggests #world #could
    WWW.TECHSPOT.COM
    John Carmack suggests the world could run on older hardware – if we optimized software better
    In context: Google researcher and reverse engineer "LaurieWired" recently posed a thought-provoking thread on X: What would happen after a CPU manufacturing apocalypse? How would the tech world respond to a future without newer, faster processors? Programming and optimization legend John Carmack offered an equally compelling answer. LaurieWired proposes the idea of a "Zero Tape-out Day" (Z-Day), an event causing manufacturers to stop producing new silicon designs. Considering the existing supply, the researcher predicts skyrocketing computer prices, stalled cloud capacity, and a ticking clock on electromigration slowly degrading the most advanced chips built on smaller nodes – all within the first year after Z-Day. Conditions would deteriorate even further in the following years, with a booming black market for processors and Xeon CPUs valued more than gold. Computing technology could regress by decades as older systems built on larger nodes prove far more resilient to electromigration. People would mod classic processors like the Motorola 68000 to operate for thousands of years without significant gate wear. More advanced systems – such as the iMac G3s sold between 1998 and 2003 – would become workstations for the elite, while the proles use repurpose hardware from Gameboys, Macintosh SEs, and Commodore 64s. LaurieWired suggests that 30 years after Z-Day, the world would become a dystopia where computing resembles the 1970s or 1980s. The modern internet would vanish, replaced by sneakernet data exchanges on SSDs and efforts to safeguard valuable desktop hardware from confiscation. Former id Software developer John Carmack decided to weigh in on the thought experiment. Having created the Doom graphics engine in just 28 hours on "vintage hardware," his expertise provided some perspective. Carmack said that a significant part of the modern world could run on outdated hardware if software optimization were a priority for developers. // Related Stories The god-tier coder suggests that developers could transition all interpreted, microservice-based products to monolithic, native codebases. Programmers would abandon modern development patterns and seek more efficient approaches, such as those used during earlier computing eras when there was no internet to push patches. Such a paradigm reset would force post-apocalyptic coders to make ancient hardware hum through software optimization. Carmack also acknowledges that innovative new products would become much rarer without ultra-cheap and scalable computing. While framed within the context of LaurieWired's thought experiment, Carmack's ideas hold practical relevance in today's computing landscape. For example, would Microsoft still impose strict hardware requirements if it prioritized optimizing Windows 11? It's a question worth considering. Similarly, how much could the gaming industry benefit from better optimization?
    0 Комментарии 0 Поделились 0 предпросмотр
  • Accessing texture data efficiently

    Learn about the benefits and trade-offs of different ways to access the underlying texture pixel data in your Unity project.Pixel data describes the color of individual pixels in a texture.
    Unity provides methods that enable you to read from or write to pixel data with C# scripts.You might use these methods to duplicate or update a texture (for example, adding a detail to a player’s profile picture), or use the texture’s data in a particular way, like reading a texture that represents a world map to determine where to place an object.There are several ways of writing code that reads from or writes to pixel data.
    The one you choose depends on what you plan to do with the data and the performance needs of your project.This blog and the accompanying sample project are intended to help you navigate the available API and common performance pitfalls.
    An understanding of both will help you write a performant solution or address performance bottlenecks as they appear.For most types of textures, Unity stores two copies of the pixel data: one in GPU memory, which is required for rendering, and the other in CPU memory.
    This copy is optional and allows you to read from, write to, and manipulate pixel data on the CPU.
    A texture with a copy of its pixel data stored in CPU memory is called a readable texture.
    One detail to note is that RenderTexture exists only in GPU memory.The memory available to the CPU differs from that of the GPU on most hardware.
    Some devices have a form of partially shared memory, but for this blog we will assume the classic PC configuration where the CPU only has direct access to the RAM plugged into the motherboard and the GPU relies on its own video RAM (VRAM).
    Any data transferred between these different environments has to pass through the PCI bus, which is slower than transferring data within the same type of memory.
    Due to these costs, you should try to limit the amount of data transferred each frame.Sampling textures in shaders is the most common GPU pixel data operation.
    To alter this data, you can copy between textures or render into a texture using a shader.
    All these operations can be performed quickly by the GPU.In some cases, it may be preferable to manipulate your texture data on the CPU, which offers more flexibility in how data is accessed.
    CPU pixel data operations act only on the CPU copy of the data, so require readable textures.
    If you want to sample the updated pixel data in a shader, you must first copy it from the CPU to the GPU by calling Apply.
    Depending on the texture involved and the complexity of the operations, it may be faster and easier to stick to CPU operations (for example, when copying several 2D textures into a Texture2DArray asset).The Unity API provides several methods to access or process texture data.
    Some operations act on both the GPU and CPU copy if both are present.
    As a result, the performance of these methods varies depending on whether the textures are readable.
    Different methods can be used to achieve the same results, but each method has its own performance and ease-of-use characteristics.Answer the following questions to determine the optimal solution:Can the GPU perform your calculations faster than the CPU?What level of pressure is the process putting on the texture caches? (For example, sampling many high-resolution textures without using mipmaps is likely to slow down the GPU.)Does the process require a random write texture, or can it output to a color or depth attachment? (Writing to random pixels on a texture requires frequent cache flushes that slow down the process.)Is my project already GPU bottlenecked? Even if the GPU is able to execute a process faster than the CPU, can the GPU afford to take on more work without exceeding its frame time budget?If both the GPU and the CPU main thread are near their frame time limit, then perhaps the slow part of a process could be performed by CPU worker threads.How much data needs to be uploaded to or downloaded from the GPU to calculate or process the results?Could a shader or C# job pack the data into a smaller format to reduce the bandwidth required?Could a RenderTexture be downsampled into a smaller resolution version that is downloaded instead?Can the process be performed in chunks? (If a lot of data needs to be processed at once, there’s a risk of the GPU not having enough memory for it.)How quickly are the results required? Can calculations or data transfers be performed asynchronously and handled later? (If too much work is done in a single frame, there is a risk that the GPU won’t have enough time to render the actual graphics for each frame.)By default, texture assets that you import into your project are nonreadable, while textures created from a script are readable.Readable textures use twice as much memory as nonreadable textures because they need to have a copy of their pixel data in CPU RAM.
    You should only make a texture readable when you need to, and make them nonreadable when you are done working with the data on the CPU.To see if a texture asset in your project is readable and make edits, use the Read/Write Enabled option in Texture Import Settings, or the TextureImporter.isReadable API.To make a texture nonreadable, call its Apply method with the makeNoLongerReadable parameter set to “true” (for example, Texture2D.Apply or Cubemap.Apply).
    A nonreadable texture can’t be made readable again.All textures are readable to the Editor in Edit and Play modes.
    Calling Apply to make the texture nonreadable will update the value of isReadable, preventing you from accessing the CPU data.
    However, some Unity processes will function as if the texture is readable because they see that the internal CPU data is valid.Performance differs greatly across the various ways of accessing texture data, especially on the CPU (although less so at lower resolutions).
    The Unity Texture Access API examples repository on GitHub contains a number of examples showing performance differences between various APIs that allow access to, or manipulation of, texture data.
    The UI only shows the main thread CPU timings.
    In some cases, DOTS features like Burst and the job system are used to maximize performance.Here are the examples included in the GitHub repository:SimpleCopy: Copying all pixels from one texture to anotherPlasmaTexture: A plasma texture updated on the CPU per frameTransferGPUTexture: Transferring (copying to a different size or format) all pixels on the GPU from a texture to a RenderTextureListed below are performance measurements taken from the examples on GitHub.
    These numbers are used to support the recommendations that follow.
    The measurements are from a player build on a system with a 3.7 GHz 8-core Xeon® W-2145 CPU and an RTX 2080.These are the median CPU times for SimpleCopy.UpdateTestCase with a texture size of 2,048.Note that the Graphics methods complete nearly instantly on the main thread because they simply push work onto the RenderThread, which is later executed by the GPU.
    Their results will be ready when the next frame is being rendered.Results1,326 ms – foreach(mip) for(x in width) for(y in height) SetPixel(x, y, GetPixel(x, y, mip), mip)32.14 ms – foreach(mip) SetPixels(source.GetPixels(mip), mip)6.96 ms – foreach(mip) SetPixels32(source.GetPixels32(mip), mip)6.74 ms – LoadRawTextureData(source.GetRawTextureData())3.54 ms – Graphics.CopyTexture(readableSource, readableTarget)2.87 ms – foreach(mip) SetPixelData(mip, GetPixelData(mip))2.87 ms – LoadRawTextureData(source.GetRawTextureData())0.00 ms – Graphics.ConvertTexture(source, target)0.00 ms – Graphics.CopyTexture(nonReadableSource, target)These are the median CPU times for PlasmaTexture.UpdateTestCase with a texture size of 512.You’ll see that SetPixels32 is unexpectedly slower than SetPixels.
    This is due to having to take the float-based Color result from the plasma pixel calculation and convert it to the byte-based Color32 struct.
    SetPixels32NoConversion skips this conversion and just assigns a default value to the Color32 output array, resulting in better performance than SetPixels.
    In order to beat the performance of SetPixels and the underlying color conversion performed by Unity, it is necessary to rework the pixel calculation method itself to directly output a Color32 value.
    A simple implementation using SetPixelData is almost guaranteed to give better results than careful SetPixels and SetPixels32 approaches.Results126.95 ms – SetPixel113.16 ms – SetPixels3288.96 ms – SetPixels86.30 ms – SetPixels32NoConversion16.91 ms – SetPixelDataBurst4.27 ms – SetPixelDataBurstParallelThese are the Editor GPU times for TransferGPUTexture.UpdateTestCase with a texture size of 8,196:Blit – 1.584 msCopyTexture – 0.882 msYou can access pixel data in various ways.
    However, not all methods support every format, texture type, or use case, and some take longer to execute than others.
    This section goes over recommended methods, and the following section covers those to use with caution.CopyTexture is the fastest way to transfer GPU data from one texture into another.
    It does not perform any format conversion.
    You can partially copy data by specifying a source and target position, in addition to the width and height of the region.
    If both textures are readable, the copy operation will also be performed on the CPU data, bringing the total cost of this method closer to that of a CPU-only copy using SetPixelData with the result of GetPixelData from a source texture.Blit is a fast and powerful method of transferring GPU data into a RenderTexture using a shader.
    In practice, this has to set up the graphics pipeline API state to render to the target RenderTexture.
    It comes with a small resolution-independent setup cost compared to CopyTexture.
    The default Blit shader used by the method takes an input texture and renders it into the target RenderTexture.
    By providing a custom material or shader, you can define complex texture-to-texture rendering processes.GetPixelData and SetPixelData (along with GetRawTextureData) are the fastest methods to use when only touching CPU data.
    Both methods require you to provide a struct type as a template parameter used to reinterpret the data.
    The methods themselves only need this struct to derive the correct size, so you can just use byte if you don’t want to define a custom struct to represent the texture’s format.When accessing individual pixels, it’s a good idea to define a custom struct with some utility methods for ease of use.
    For example, an R5G5B5A1 format struct could be made up out of a ushort data member and a few get/set methods to access the individual channels as bytes.The above code is an example from an implementation of an object representing a pixel in the R5G5B5A5A1 format; the corresponding property setters are omitted for brevity.SetPixelData can be used to copy a full mip level of data into the target texture.
    GetPixelData will return a NativeArray that actually points to one mip level of Unity’s internal CPU texture data.
    This allows you to directly read/write that data without the need for any copy operations.
    The catch is that the NativeArray returned by GetPixelData is only guaranteed to be valid until the user code calling GetPixelData returns control to Unity, such as when MonoBehaviour.Update returns.
    Instead of storing the result of GetPixelData between frames, you have to get the correct NativeArray from GetPixelData for every frame you want to access this data from.The Apply method returns after the CPU data has been uploaded to the GPU.
    The makeNoLongerReadable parameter should be set to “true” where possible to free up the memory of the CPU data after the upload.The RequestIntoNativeArray and RequestIntoNativeSlice methods asynchronously download GPU data from the specified Texture into (a slice of) a NativeArray provided by the user.Calling the methods will return a request handle that can indicate if the requested data is done downloading.
    Support is limited to only a handful of formats, so use SystemInfo.IsFormatSupported with FormatUsage.ReadPixels to check format support.
    The AsyncGPUReadback class also has a Request method, which allocates a NativeArray for you.
    If you need to repeat this operation, you will get better performance if you allocate a NativeArray that you reuse instead.There are a number of methods that should be used with caution due to potentially significant performance impacts.
    Let’s take a look at them in more detail.These methods perform pixel format conversions of varying complexity.
    The Pixels32 variants are the most performant of the bunch, but even they can still perform format conversions if the underlying format of the texture doesn’t perfectly match the Color32 struct.
    When using the following methods, it’s best to keep in mind that their performance impact significantly increases by varying degrees as the number of pixels grows:GetPixelGetPixelBilinearSetPixelGetPixelsSetPixelsGetPixels32SetPixels32GetRawTextureData and LoadRawTextureData are Texture2D-only methods that work with arrays containing the raw pixel data of all mip levels, one after another.
    The layout goes from largest to smallest mip, with each mip being “height” amount of “width” pixel values.
    These functions are quick to give CPU data access.
    GetRawTextureData does have a “gotcha” where the non-templated variant returns a copy of the data.
    This is a bit slower, and does not allow direct manipulation of the underlying buffer managed by Unity.
    GetPixelData does not have this quirk and can only return a NativeArray pointing to the underlying buffer that remains valid until user code returns control to Unity.ConvertTexture is a way to transfer the GPU data from one texture to another, where the source and destination textures don’t have the same size or format.
    This conversion process is as efficient as it gets under the circumstances, but it’s not cheap.
    This is the internal process:Allocate a temporary RenderTexture matching the destination texture.Perform a Blit from the source texture to the temporary RenderTexture.Copy the Blit result from the temporary RenderTexture to the destination texture.Answer the following questions to help determine if this method is suited to your use case:Do I need to perform this conversion?Can I make sure the source texture is created in the desired size/format for the target platform at import time?Can I change my processes to use the same formats, allowing the result of one process to be directly used as an input for another process?Can I create and use a RenderTexture as the destination instead? Doing so would reduce the conversion process to a single Blit to the destination RenderTexture.The ReadPixels method synchronously downloads GPU data from the active RenderTexture (RenderTexture.active) into a Texture2D’s CPU data.
    This enables you to store or process the output from a rendering operation.
    Support is limited to only a handful of formats, so use SystemInfo.IsFormatSupported with FormatUsage.ReadPixels to check format support.Downloading data back from the GPU is a slow process.
    Before it can begin, ReadPixels has to wait for the GPU to complete all preceding work.
    It’s best to avoid this method as it will not return until the requested data is available, which will slow down performance.
    Usability is also a concern because you need GPU data to be in a RenderTexture, which has to be configured as the currently active one.
    Both usability and performance are better when using the AsyncGPUReadback methods discussed earlier.The ImageConversion class has methods to convert between Texture2D and several image file formats.
    LoadImage is able to load JPG, PNG, or EXR (since 2023.1) data into a Texture2D and upload this to the GPU for you.
    The loaded pixel data can be compressed on the fly depending on Texture2D’s original format.
    Other methods can convert a Texture2D or pixel data array to an array of JPG, PNG, TGA, or EXR data.These methods are not particularly fast, but can be useful if your project needs to pass pixel data around through common image file formats.
    Typical use cases include loading a user’s avatar from disk and sharing it with other players over a network.There are many resources available to learn more about graphics optimization, related topics, and best practices in Unity.
    The graphics performance and profiling section of the documentation is a good starting point.You can also check out several technical e-books for advanced users, including Ultimate guide to profiling Unity games, Optimize your mobile game performance, and Optimize your console and PC game performance.You’ll find many more advanced best practices on the Unity how-to hub.Here’s a summary of the key points to remember:When manipulating textures, the first step is to assess which operations can be performed on the GPU for optimal performance.
    The existing CPU/GPU workload and size of the input/output data are key factors to consider.Using low level functions like GetRawTextureData to implement a specific conversion path where necessary can offer improved performance over the more convenient methods that perform (often redundant) copies and conversions.More complex operations, such as large readbacks and pixel calculations, are only viable on the CPU when performed asynchronously or in parallel.
    The combination of Burst and the job system allows C# to perform certain operations that would otherwise only be performant on a GPU.Profile frequently: There are many pitfalls you can encounter during development, from unexpected and unnecessary conversions to stalls from waiting on another process.
    Some performance issues will only start surfacing as the game scales up and certain parts of your code see heavier usage.
    The example project demonstrates how seemingly small increases in texture resolution can cause certain APIs to become a performance issue.Share your feedback on texture data with us in the Scripting or General Graphics forums.
    Be sure to watch for new technical blogs from other Unity developers as part of the ongoing Tech from the Trenches series.
    Source: https://unity.com/blog/engine-platform/accessing-texture-data-efficiently
    #accessing #texture #data #efficiently
    Accessing texture data efficiently
    Learn about the benefits and trade-offs of different ways to access the underlying texture pixel data in your Unity project.Pixel data describes the color of individual pixels in a texture. Unity provides methods that enable you to read from or write to pixel data with C# scripts.You might use these methods to duplicate or update a texture (for example, adding a detail to a player’s profile picture), or use the texture’s data in a particular way, like reading a texture that represents a world map to determine where to place an object.There are several ways of writing code that reads from or writes to pixel data. The one you choose depends on what you plan to do with the data and the performance needs of your project.This blog and the accompanying sample project are intended to help you navigate the available API and common performance pitfalls. An understanding of both will help you write a performant solution or address performance bottlenecks as they appear.For most types of textures, Unity stores two copies of the pixel data: one in GPU memory, which is required for rendering, and the other in CPU memory. This copy is optional and allows you to read from, write to, and manipulate pixel data on the CPU. A texture with a copy of its pixel data stored in CPU memory is called a readable texture. One detail to note is that RenderTexture exists only in GPU memory.The memory available to the CPU differs from that of the GPU on most hardware. Some devices have a form of partially shared memory, but for this blog we will assume the classic PC configuration where the CPU only has direct access to the RAM plugged into the motherboard and the GPU relies on its own video RAM (VRAM). Any data transferred between these different environments has to pass through the PCI bus, which is slower than transferring data within the same type of memory. Due to these costs, you should try to limit the amount of data transferred each frame.Sampling textures in shaders is the most common GPU pixel data operation. To alter this data, you can copy between textures or render into a texture using a shader. All these operations can be performed quickly by the GPU.In some cases, it may be preferable to manipulate your texture data on the CPU, which offers more flexibility in how data is accessed. CPU pixel data operations act only on the CPU copy of the data, so require readable textures. If you want to sample the updated pixel data in a shader, you must first copy it from the CPU to the GPU by calling Apply. Depending on the texture involved and the complexity of the operations, it may be faster and easier to stick to CPU operations (for example, when copying several 2D textures into a Texture2DArray asset).The Unity API provides several methods to access or process texture data. Some operations act on both the GPU and CPU copy if both are present. As a result, the performance of these methods varies depending on whether the textures are readable. Different methods can be used to achieve the same results, but each method has its own performance and ease-of-use characteristics.Answer the following questions to determine the optimal solution:Can the GPU perform your calculations faster than the CPU?What level of pressure is the process putting on the texture caches? (For example, sampling many high-resolution textures without using mipmaps is likely to slow down the GPU.)Does the process require a random write texture, or can it output to a color or depth attachment? (Writing to random pixels on a texture requires frequent cache flushes that slow down the process.)Is my project already GPU bottlenecked? Even if the GPU is able to execute a process faster than the CPU, can the GPU afford to take on more work without exceeding its frame time budget?If both the GPU and the CPU main thread are near their frame time limit, then perhaps the slow part of a process could be performed by CPU worker threads.How much data needs to be uploaded to or downloaded from the GPU to calculate or process the results?Could a shader or C# job pack the data into a smaller format to reduce the bandwidth required?Could a RenderTexture be downsampled into a smaller resolution version that is downloaded instead?Can the process be performed in chunks? (If a lot of data needs to be processed at once, there’s a risk of the GPU not having enough memory for it.)How quickly are the results required? Can calculations or data transfers be performed asynchronously and handled later? (If too much work is done in a single frame, there is a risk that the GPU won’t have enough time to render the actual graphics for each frame.)By default, texture assets that you import into your project are nonreadable, while textures created from a script are readable.Readable textures use twice as much memory as nonreadable textures because they need to have a copy of their pixel data in CPU RAM. You should only make a texture readable when you need to, and make them nonreadable when you are done working with the data on the CPU.To see if a texture asset in your project is readable and make edits, use the Read/Write Enabled option in Texture Import Settings, or the TextureImporter.isReadable API.To make a texture nonreadable, call its Apply method with the makeNoLongerReadable parameter set to “true” (for example, Texture2D.Apply or Cubemap.Apply). A nonreadable texture can’t be made readable again.All textures are readable to the Editor in Edit and Play modes. Calling Apply to make the texture nonreadable will update the value of isReadable, preventing you from accessing the CPU data. However, some Unity processes will function as if the texture is readable because they see that the internal CPU data is valid.Performance differs greatly across the various ways of accessing texture data, especially on the CPU (although less so at lower resolutions). The Unity Texture Access API examples repository on GitHub contains a number of examples showing performance differences between various APIs that allow access to, or manipulation of, texture data. The UI only shows the main thread CPU timings. In some cases, DOTS features like Burst and the job system are used to maximize performance.Here are the examples included in the GitHub repository:SimpleCopy: Copying all pixels from one texture to anotherPlasmaTexture: A plasma texture updated on the CPU per frameTransferGPUTexture: Transferring (copying to a different size or format) all pixels on the GPU from a texture to a RenderTextureListed below are performance measurements taken from the examples on GitHub. These numbers are used to support the recommendations that follow. The measurements are from a player build on a system with a 3.7 GHz 8-core Xeon® W-2145 CPU and an RTX 2080.These are the median CPU times for SimpleCopy.UpdateTestCase with a texture size of 2,048.Note that the Graphics methods complete nearly instantly on the main thread because they simply push work onto the RenderThread, which is later executed by the GPU. Their results will be ready when the next frame is being rendered.Results1,326 ms – foreach(mip) for(x in width) for(y in height) SetPixel(x, y, GetPixel(x, y, mip), mip)32.14 ms – foreach(mip) SetPixels(source.GetPixels(mip), mip)6.96 ms – foreach(mip) SetPixels32(source.GetPixels32(mip), mip)6.74 ms – LoadRawTextureData(source.GetRawTextureData())3.54 ms – Graphics.CopyTexture(readableSource, readableTarget)2.87 ms – foreach(mip) SetPixelData(mip, GetPixelData(mip))2.87 ms – LoadRawTextureData(source.GetRawTextureData())0.00 ms – Graphics.ConvertTexture(source, target)0.00 ms – Graphics.CopyTexture(nonReadableSource, target)These are the median CPU times for PlasmaTexture.UpdateTestCase with a texture size of 512.You’ll see that SetPixels32 is unexpectedly slower than SetPixels. This is due to having to take the float-based Color result from the plasma pixel calculation and convert it to the byte-based Color32 struct. SetPixels32NoConversion skips this conversion and just assigns a default value to the Color32 output array, resulting in better performance than SetPixels. In order to beat the performance of SetPixels and the underlying color conversion performed by Unity, it is necessary to rework the pixel calculation method itself to directly output a Color32 value. A simple implementation using SetPixelData is almost guaranteed to give better results than careful SetPixels and SetPixels32 approaches.Results126.95 ms – SetPixel113.16 ms – SetPixels3288.96 ms – SetPixels86.30 ms – SetPixels32NoConversion16.91 ms – SetPixelDataBurst4.27 ms – SetPixelDataBurstParallelThese are the Editor GPU times for TransferGPUTexture.UpdateTestCase with a texture size of 8,196:Blit – 1.584 msCopyTexture – 0.882 msYou can access pixel data in various ways. However, not all methods support every format, texture type, or use case, and some take longer to execute than others. This section goes over recommended methods, and the following section covers those to use with caution.CopyTexture is the fastest way to transfer GPU data from one texture into another. It does not perform any format conversion. You can partially copy data by specifying a source and target position, in addition to the width and height of the region. If both textures are readable, the copy operation will also be performed on the CPU data, bringing the total cost of this method closer to that of a CPU-only copy using SetPixelData with the result of GetPixelData from a source texture.Blit is a fast and powerful method of transferring GPU data into a RenderTexture using a shader. In practice, this has to set up the graphics pipeline API state to render to the target RenderTexture. It comes with a small resolution-independent setup cost compared to CopyTexture. The default Blit shader used by the method takes an input texture and renders it into the target RenderTexture. By providing a custom material or shader, you can define complex texture-to-texture rendering processes.GetPixelData and SetPixelData (along with GetRawTextureData) are the fastest methods to use when only touching CPU data. Both methods require you to provide a struct type as a template parameter used to reinterpret the data. The methods themselves only need this struct to derive the correct size, so you can just use byte if you don’t want to define a custom struct to represent the texture’s format.When accessing individual pixels, it’s a good idea to define a custom struct with some utility methods for ease of use. For example, an R5G5B5A1 format struct could be made up out of a ushort data member and a few get/set methods to access the individual channels as bytes.The above code is an example from an implementation of an object representing a pixel in the R5G5B5A5A1 format; the corresponding property setters are omitted for brevity.SetPixelData can be used to copy a full mip level of data into the target texture. GetPixelData will return a NativeArray that actually points to one mip level of Unity’s internal CPU texture data. This allows you to directly read/write that data without the need for any copy operations. The catch is that the NativeArray returned by GetPixelData is only guaranteed to be valid until the user code calling GetPixelData returns control to Unity, such as when MonoBehaviour.Update returns. Instead of storing the result of GetPixelData between frames, you have to get the correct NativeArray from GetPixelData for every frame you want to access this data from.The Apply method returns after the CPU data has been uploaded to the GPU. The makeNoLongerReadable parameter should be set to “true” where possible to free up the memory of the CPU data after the upload.The RequestIntoNativeArray and RequestIntoNativeSlice methods asynchronously download GPU data from the specified Texture into (a slice of) a NativeArray provided by the user.Calling the methods will return a request handle that can indicate if the requested data is done downloading. Support is limited to only a handful of formats, so use SystemInfo.IsFormatSupported with FormatUsage.ReadPixels to check format support. The AsyncGPUReadback class also has a Request method, which allocates a NativeArray for you. If you need to repeat this operation, you will get better performance if you allocate a NativeArray that you reuse instead.There are a number of methods that should be used with caution due to potentially significant performance impacts. Let’s take a look at them in more detail.These methods perform pixel format conversions of varying complexity. The Pixels32 variants are the most performant of the bunch, but even they can still perform format conversions if the underlying format of the texture doesn’t perfectly match the Color32 struct. When using the following methods, it’s best to keep in mind that their performance impact significantly increases by varying degrees as the number of pixels grows:GetPixelGetPixelBilinearSetPixelGetPixelsSetPixelsGetPixels32SetPixels32GetRawTextureData and LoadRawTextureData are Texture2D-only methods that work with arrays containing the raw pixel data of all mip levels, one after another. The layout goes from largest to smallest mip, with each mip being “height” amount of “width” pixel values. These functions are quick to give CPU data access. GetRawTextureData does have a “gotcha” where the non-templated variant returns a copy of the data. This is a bit slower, and does not allow direct manipulation of the underlying buffer managed by Unity. GetPixelData does not have this quirk and can only return a NativeArray pointing to the underlying buffer that remains valid until user code returns control to Unity.ConvertTexture is a way to transfer the GPU data from one texture to another, where the source and destination textures don’t have the same size or format. This conversion process is as efficient as it gets under the circumstances, but it’s not cheap. This is the internal process:Allocate a temporary RenderTexture matching the destination texture.Perform a Blit from the source texture to the temporary RenderTexture.Copy the Blit result from the temporary RenderTexture to the destination texture.Answer the following questions to help determine if this method is suited to your use case:Do I need to perform this conversion?Can I make sure the source texture is created in the desired size/format for the target platform at import time?Can I change my processes to use the same formats, allowing the result of one process to be directly used as an input for another process?Can I create and use a RenderTexture as the destination instead? Doing so would reduce the conversion process to a single Blit to the destination RenderTexture.The ReadPixels method synchronously downloads GPU data from the active RenderTexture (RenderTexture.active) into a Texture2D’s CPU data. This enables you to store or process the output from a rendering operation. Support is limited to only a handful of formats, so use SystemInfo.IsFormatSupported with FormatUsage.ReadPixels to check format support.Downloading data back from the GPU is a slow process. Before it can begin, ReadPixels has to wait for the GPU to complete all preceding work. It’s best to avoid this method as it will not return until the requested data is available, which will slow down performance. Usability is also a concern because you need GPU data to be in a RenderTexture, which has to be configured as the currently active one. Both usability and performance are better when using the AsyncGPUReadback methods discussed earlier.The ImageConversion class has methods to convert between Texture2D and several image file formats. LoadImage is able to load JPG, PNG, or EXR (since 2023.1) data into a Texture2D and upload this to the GPU for you. The loaded pixel data can be compressed on the fly depending on Texture2D’s original format. Other methods can convert a Texture2D or pixel data array to an array of JPG, PNG, TGA, or EXR data.These methods are not particularly fast, but can be useful if your project needs to pass pixel data around through common image file formats. Typical use cases include loading a user’s avatar from disk and sharing it with other players over a network.There are many resources available to learn more about graphics optimization, related topics, and best practices in Unity. The graphics performance and profiling section of the documentation is a good starting point.You can also check out several technical e-books for advanced users, including Ultimate guide to profiling Unity games, Optimize your mobile game performance, and Optimize your console and PC game performance.You’ll find many more advanced best practices on the Unity how-to hub.Here’s a summary of the key points to remember:When manipulating textures, the first step is to assess which operations can be performed on the GPU for optimal performance. The existing CPU/GPU workload and size of the input/output data are key factors to consider.Using low level functions like GetRawTextureData to implement a specific conversion path where necessary can offer improved performance over the more convenient methods that perform (often redundant) copies and conversions.More complex operations, such as large readbacks and pixel calculations, are only viable on the CPU when performed asynchronously or in parallel. The combination of Burst and the job system allows C# to perform certain operations that would otherwise only be performant on a GPU.Profile frequently: There are many pitfalls you can encounter during development, from unexpected and unnecessary conversions to stalls from waiting on another process. Some performance issues will only start surfacing as the game scales up and certain parts of your code see heavier usage. The example project demonstrates how seemingly small increases in texture resolution can cause certain APIs to become a performance issue.Share your feedback on texture data with us in the Scripting or General Graphics forums. Be sure to watch for new technical blogs from other Unity developers as part of the ongoing Tech from the Trenches series. Source: https://unity.com/blog/engine-platform/accessing-texture-data-efficiently #accessing #texture #data #efficiently
    UNITY.COM
    Accessing texture data efficiently
    Learn about the benefits and trade-offs of different ways to access the underlying texture pixel data in your Unity project.Pixel data describes the color of individual pixels in a texture. Unity provides methods that enable you to read from or write to pixel data with C# scripts.You might use these methods to duplicate or update a texture (for example, adding a detail to a player’s profile picture), or use the texture’s data in a particular way, like reading a texture that represents a world map to determine where to place an object.There are several ways of writing code that reads from or writes to pixel data. The one you choose depends on what you plan to do with the data and the performance needs of your project.This blog and the accompanying sample project are intended to help you navigate the available API and common performance pitfalls. An understanding of both will help you write a performant solution or address performance bottlenecks as they appear.For most types of textures, Unity stores two copies of the pixel data: one in GPU memory, which is required for rendering, and the other in CPU memory. This copy is optional and allows you to read from, write to, and manipulate pixel data on the CPU. A texture with a copy of its pixel data stored in CPU memory is called a readable texture. One detail to note is that RenderTexture exists only in GPU memory.The memory available to the CPU differs from that of the GPU on most hardware. Some devices have a form of partially shared memory, but for this blog we will assume the classic PC configuration where the CPU only has direct access to the RAM plugged into the motherboard and the GPU relies on its own video RAM (VRAM). Any data transferred between these different environments has to pass through the PCI bus, which is slower than transferring data within the same type of memory. Due to these costs, you should try to limit the amount of data transferred each frame.Sampling textures in shaders is the most common GPU pixel data operation. To alter this data, you can copy between textures or render into a texture using a shader. All these operations can be performed quickly by the GPU.In some cases, it may be preferable to manipulate your texture data on the CPU, which offers more flexibility in how data is accessed. CPU pixel data operations act only on the CPU copy of the data, so require readable textures. If you want to sample the updated pixel data in a shader, you must first copy it from the CPU to the GPU by calling Apply. Depending on the texture involved and the complexity of the operations, it may be faster and easier to stick to CPU operations (for example, when copying several 2D textures into a Texture2DArray asset).The Unity API provides several methods to access or process texture data. Some operations act on both the GPU and CPU copy if both are present. As a result, the performance of these methods varies depending on whether the textures are readable. Different methods can be used to achieve the same results, but each method has its own performance and ease-of-use characteristics.Answer the following questions to determine the optimal solution:Can the GPU perform your calculations faster than the CPU?What level of pressure is the process putting on the texture caches? (For example, sampling many high-resolution textures without using mipmaps is likely to slow down the GPU.)Does the process require a random write texture, or can it output to a color or depth attachment? (Writing to random pixels on a texture requires frequent cache flushes that slow down the process.)Is my project already GPU bottlenecked? Even if the GPU is able to execute a process faster than the CPU, can the GPU afford to take on more work without exceeding its frame time budget?If both the GPU and the CPU main thread are near their frame time limit, then perhaps the slow part of a process could be performed by CPU worker threads.How much data needs to be uploaded to or downloaded from the GPU to calculate or process the results?Could a shader or C# job pack the data into a smaller format to reduce the bandwidth required?Could a RenderTexture be downsampled into a smaller resolution version that is downloaded instead?Can the process be performed in chunks? (If a lot of data needs to be processed at once, there’s a risk of the GPU not having enough memory for it.)How quickly are the results required? Can calculations or data transfers be performed asynchronously and handled later? (If too much work is done in a single frame, there is a risk that the GPU won’t have enough time to render the actual graphics for each frame.)By default, texture assets that you import into your project are nonreadable, while textures created from a script are readable.Readable textures use twice as much memory as nonreadable textures because they need to have a copy of their pixel data in CPU RAM. You should only make a texture readable when you need to, and make them nonreadable when you are done working with the data on the CPU.To see if a texture asset in your project is readable and make edits, use the Read/Write Enabled option in Texture Import Settings, or the TextureImporter.isReadable API.To make a texture nonreadable, call its Apply method with the makeNoLongerReadable parameter set to “true” (for example, Texture2D.Apply or Cubemap.Apply). A nonreadable texture can’t be made readable again.All textures are readable to the Editor in Edit and Play modes. Calling Apply to make the texture nonreadable will update the value of isReadable, preventing you from accessing the CPU data. However, some Unity processes will function as if the texture is readable because they see that the internal CPU data is valid.Performance differs greatly across the various ways of accessing texture data, especially on the CPU (although less so at lower resolutions). The Unity Texture Access API examples repository on GitHub contains a number of examples showing performance differences between various APIs that allow access to, or manipulation of, texture data. The UI only shows the main thread CPU timings. In some cases, DOTS features like Burst and the job system are used to maximize performance.Here are the examples included in the GitHub repository:SimpleCopy: Copying all pixels from one texture to anotherPlasmaTexture: A plasma texture updated on the CPU per frameTransferGPUTexture: Transferring (copying to a different size or format) all pixels on the GPU from a texture to a RenderTextureListed below are performance measurements taken from the examples on GitHub. These numbers are used to support the recommendations that follow. The measurements are from a player build on a system with a 3.7 GHz 8-core Xeon® W-2145 CPU and an RTX 2080.These are the median CPU times for SimpleCopy.UpdateTestCase with a texture size of 2,048.Note that the Graphics methods complete nearly instantly on the main thread because they simply push work onto the RenderThread, which is later executed by the GPU. Their results will be ready when the next frame is being rendered.Results1,326 ms – foreach(mip) for(x in width) for(y in height) SetPixel(x, y, GetPixel(x, y, mip), mip)32.14 ms – foreach(mip) SetPixels(source.GetPixels(mip), mip)6.96 ms – foreach(mip) SetPixels32(source.GetPixels32(mip), mip)6.74 ms – LoadRawTextureData(source.GetRawTextureData())3.54 ms – Graphics.CopyTexture(readableSource, readableTarget)2.87 ms – foreach(mip) SetPixelData(mip, GetPixelData(mip))2.87 ms – LoadRawTextureData(source.GetRawTextureData())0.00 ms – Graphics.ConvertTexture(source, target)0.00 ms – Graphics.CopyTexture(nonReadableSource, target)These are the median CPU times for PlasmaTexture.UpdateTestCase with a texture size of 512.You’ll see that SetPixels32 is unexpectedly slower than SetPixels. This is due to having to take the float-based Color result from the plasma pixel calculation and convert it to the byte-based Color32 struct. SetPixels32NoConversion skips this conversion and just assigns a default value to the Color32 output array, resulting in better performance than SetPixels. In order to beat the performance of SetPixels and the underlying color conversion performed by Unity, it is necessary to rework the pixel calculation method itself to directly output a Color32 value. A simple implementation using SetPixelData is almost guaranteed to give better results than careful SetPixels and SetPixels32 approaches.Results126.95 ms – SetPixel113.16 ms – SetPixels3288.96 ms – SetPixels86.30 ms – SetPixels32NoConversion16.91 ms – SetPixelDataBurst4.27 ms – SetPixelDataBurstParallelThese are the Editor GPU times for TransferGPUTexture.UpdateTestCase with a texture size of 8,196:Blit – 1.584 msCopyTexture – 0.882 msYou can access pixel data in various ways. However, not all methods support every format, texture type, or use case, and some take longer to execute than others. This section goes over recommended methods, and the following section covers those to use with caution.CopyTexture is the fastest way to transfer GPU data from one texture into another. It does not perform any format conversion. You can partially copy data by specifying a source and target position, in addition to the width and height of the region. If both textures are readable, the copy operation will also be performed on the CPU data, bringing the total cost of this method closer to that of a CPU-only copy using SetPixelData with the result of GetPixelData from a source texture.Blit is a fast and powerful method of transferring GPU data into a RenderTexture using a shader. In practice, this has to set up the graphics pipeline API state to render to the target RenderTexture. It comes with a small resolution-independent setup cost compared to CopyTexture. The default Blit shader used by the method takes an input texture and renders it into the target RenderTexture. By providing a custom material or shader, you can define complex texture-to-texture rendering processes.GetPixelData and SetPixelData (along with GetRawTextureData) are the fastest methods to use when only touching CPU data. Both methods require you to provide a struct type as a template parameter used to reinterpret the data. The methods themselves only need this struct to derive the correct size, so you can just use byte if you don’t want to define a custom struct to represent the texture’s format.When accessing individual pixels, it’s a good idea to define a custom struct with some utility methods for ease of use. For example, an R5G5B5A1 format struct could be made up out of a ushort data member and a few get/set methods to access the individual channels as bytes.The above code is an example from an implementation of an object representing a pixel in the R5G5B5A5A1 format; the corresponding property setters are omitted for brevity.SetPixelData can be used to copy a full mip level of data into the target texture. GetPixelData will return a NativeArray that actually points to one mip level of Unity’s internal CPU texture data. This allows you to directly read/write that data without the need for any copy operations. The catch is that the NativeArray returned by GetPixelData is only guaranteed to be valid until the user code calling GetPixelData returns control to Unity, such as when MonoBehaviour.Update returns. Instead of storing the result of GetPixelData between frames, you have to get the correct NativeArray from GetPixelData for every frame you want to access this data from.The Apply method returns after the CPU data has been uploaded to the GPU. The makeNoLongerReadable parameter should be set to “true” where possible to free up the memory of the CPU data after the upload.The RequestIntoNativeArray and RequestIntoNativeSlice methods asynchronously download GPU data from the specified Texture into (a slice of) a NativeArray provided by the user.Calling the methods will return a request handle that can indicate if the requested data is done downloading. Support is limited to only a handful of formats, so use SystemInfo.IsFormatSupported with FormatUsage.ReadPixels to check format support. The AsyncGPUReadback class also has a Request method, which allocates a NativeArray for you. If you need to repeat this operation, you will get better performance if you allocate a NativeArray that you reuse instead.There are a number of methods that should be used with caution due to potentially significant performance impacts. Let’s take a look at them in more detail.These methods perform pixel format conversions of varying complexity. The Pixels32 variants are the most performant of the bunch, but even they can still perform format conversions if the underlying format of the texture doesn’t perfectly match the Color32 struct. When using the following methods, it’s best to keep in mind that their performance impact significantly increases by varying degrees as the number of pixels grows:GetPixelGetPixelBilinearSetPixelGetPixelsSetPixelsGetPixels32SetPixels32GetRawTextureData and LoadRawTextureData are Texture2D-only methods that work with arrays containing the raw pixel data of all mip levels, one after another. The layout goes from largest to smallest mip, with each mip being “height” amount of “width” pixel values. These functions are quick to give CPU data access. GetRawTextureData does have a “gotcha” where the non-templated variant returns a copy of the data. This is a bit slower, and does not allow direct manipulation of the underlying buffer managed by Unity. GetPixelData does not have this quirk and can only return a NativeArray pointing to the underlying buffer that remains valid until user code returns control to Unity.ConvertTexture is a way to transfer the GPU data from one texture to another, where the source and destination textures don’t have the same size or format. This conversion process is as efficient as it gets under the circumstances, but it’s not cheap. This is the internal process:Allocate a temporary RenderTexture matching the destination texture.Perform a Blit from the source texture to the temporary RenderTexture.Copy the Blit result from the temporary RenderTexture to the destination texture.Answer the following questions to help determine if this method is suited to your use case:Do I need to perform this conversion?Can I make sure the source texture is created in the desired size/format for the target platform at import time?Can I change my processes to use the same formats, allowing the result of one process to be directly used as an input for another process?Can I create and use a RenderTexture as the destination instead? Doing so would reduce the conversion process to a single Blit to the destination RenderTexture.The ReadPixels method synchronously downloads GPU data from the active RenderTexture (RenderTexture.active) into a Texture2D’s CPU data. This enables you to store or process the output from a rendering operation. Support is limited to only a handful of formats, so use SystemInfo.IsFormatSupported with FormatUsage.ReadPixels to check format support.Downloading data back from the GPU is a slow process. Before it can begin, ReadPixels has to wait for the GPU to complete all preceding work. It’s best to avoid this method as it will not return until the requested data is available, which will slow down performance. Usability is also a concern because you need GPU data to be in a RenderTexture, which has to be configured as the currently active one. Both usability and performance are better when using the AsyncGPUReadback methods discussed earlier.The ImageConversion class has methods to convert between Texture2D and several image file formats. LoadImage is able to load JPG, PNG, or EXR (since 2023.1) data into a Texture2D and upload this to the GPU for you. The loaded pixel data can be compressed on the fly depending on Texture2D’s original format. Other methods can convert a Texture2D or pixel data array to an array of JPG, PNG, TGA, or EXR data.These methods are not particularly fast, but can be useful if your project needs to pass pixel data around through common image file formats. Typical use cases include loading a user’s avatar from disk and sharing it with other players over a network.There are many resources available to learn more about graphics optimization, related topics, and best practices in Unity. The graphics performance and profiling section of the documentation is a good starting point.You can also check out several technical e-books for advanced users, including Ultimate guide to profiling Unity games, Optimize your mobile game performance, and Optimize your console and PC game performance.You’ll find many more advanced best practices on the Unity how-to hub.Here’s a summary of the key points to remember:When manipulating textures, the first step is to assess which operations can be performed on the GPU for optimal performance. The existing CPU/GPU workload and size of the input/output data are key factors to consider.Using low level functions like GetRawTextureData to implement a specific conversion path where necessary can offer improved performance over the more convenient methods that perform (often redundant) copies and conversions.More complex operations, such as large readbacks and pixel calculations, are only viable on the CPU when performed asynchronously or in parallel. The combination of Burst and the job system allows C# to perform certain operations that would otherwise only be performant on a GPU.Profile frequently: There are many pitfalls you can encounter during development, from unexpected and unnecessary conversions to stalls from waiting on another process. Some performance issues will only start surfacing as the game scales up and certain parts of your code see heavier usage. The example project demonstrates how seemingly small increases in texture resolution can cause certain APIs to become a performance issue.Share your feedback on texture data with us in the Scripting or General Graphics forums. Be sure to watch for new technical blogs from other Unity developers as part of the ongoing Tech from the Trenches series.
    0 Комментарии 0 Поделились 0 предпросмотр
CGShares https://cgshares.com