AMD RX 9070 AI performance benchmark review vs 9070 XT, 7800 XT, Nvidia RTX 5070, 4070
Review
When you purchase through links on our site, we may earn an affiliate commission. Here’s how it works.
AMD RX 9070 AI performance benchmark review vs 9070 XT, 7800 XT, Nvidia RTX 5070, 4070
Sayan Sen
Neowin
@ssc_combater007 ·
May 24, 2025 14:40 EDT
Earlier this month, we shared the first part of our review of AMD's new RX 9070. It was about the gaming performance of the GPU, and we gave it a 7.5 out of 10. The 9070 XT, in contrast, received a full 10 out of 10.
The main reason for the lower score on the non-XT was the relatively high price and thus the poorer value it offered compared to the XT. We thought the price was much closer to the XT than it needed to be.
While the RX 9070 proved to be more power efficient than the XT, for a desktop gaming graphics card, value and performance typically take the front seat compared to something like power efficiency.
However, that may not be the case in terms of productivity which also takes into account things like power savings. Thus, similar to the one we did for the XT model, we are doing a dedicated productivity review for the RX 9070 as well where we compare to against the 9070 XT, 7800 XT, as well as Nvidia's 5070 and 4070.
AI performance is a very important metric in today's world and AMD also promised big improvements thanks to its underlying architectural improvements. We already had a taste of that with the XT model so now it's time to see how good the non-XT does here.
Before we get underway, this is a collaboration between Sayan Sen, and Steven Parker who lent us their test PC for this review. Speaking of which, here are the specs of the test PC:
Cooler Master MasterBox NR200P MAX
ASRock Z790 PG-ITX/TB4
Intel Core i7-14700K with Thermal Grizzly Carbonaut Pad
T-FORCE Delta RGB DDR57600MT/s CL362TB Kingston Fury Renegade SSD
Windows 11 24H2Drivers used for the 7800 XT, 9070 XT and 9070 were Adrenaline v24.30.31.03 / 25.3.1 RC, and for the Nvidia RTX 5070 and 4070, GeForce v572.47 was used.Sapphire Pulse 9070 XT, Nvidia 5070 FE, and Pulse 9070First up, we have Geekbench AI running on ONNX.
The RTX 5070 gets beaten by both the 9070 XT and 9070 in quantized and single precisionperformance. Similarly, the 4070 gets close to the 9070 in half-precisionperformance, but the latter is an enormous 30% faster in quantized score and nearly 12.2% better in single precision.
The reason for this beatdown is the amount of memory available to each card. The Nvidia GPUs have 12GB each and thus only do better in the FP16 precision tests since the other ones are more VRAM-intensive.
Next up, we move to UL Procyon suite starting with the Image generation benchmark.
We chose the Stable Diffusion XL FP16 test since this is the most intense workload available on Procyon suite. Similar to what we saw on Geekbench AI, the Nvidia GPUs to relatively better here as it is FP16 or half precision which means the used VRAM is lower.
So this is something to keep in mind again, if you wish to float32 AI workloads, it is likely that graphics cards with greater than 12 GB buffers would emerge as victors.
There is still a big improvement on the RX 9070 compared to the 7800 XT as we see a ~54% gain. This boost is due to improvements to the core architecture itself as VRAM capacities of both cards are the same at 16 Gigs.
Following image generation, we move to the text generation benchmark.
In this workload, we see the least impressive performance of the 9070 in terms of how much it improves over the 7800 XT. The former is up to ~7.25% faster here. The 9070 is also not as well-performing as the Nvidia 4070 in Phi and Mistral models, although it does do better in both the Llama tests.
Another odd result stood out here where the 5070 underperformed all the cards including the 7800 XT in Llama 2. We ran each test three times and considered the best score and so we are not exactly sure what happened here.
Wrapping up AI testing, we measured OpenCL throughput in Geekbench compute benchmark.
The RX 9070 did not fare well here at all even falling behind the 7800 XT and it is significantly slower than the three other cards. Interestingly, even the RTX 5070 could not beat the 4070 on OpenCL so perhaps this suggests that OpenCL optimization has not been a priority for either AMD or Nvidia this time. It could also be an issue with Geekbench itself.
Conclusion
We reach the end of our productivity performance review of the 9070 and we have to say we are fairly impressed but there is also a slight bit of disappointment. It is clear that the 9070 as well as the 9070 XT really shine when inferencing precision is higher, and that is due to the higher memory buffers they possess compared to the Nvidia 5070. But on FP16, the Nvidia cards pull ahead.
Still RNDA 4, including the RX 9070, see big boost over RDNA 3. As we noted in the image generation benchmark, which is an intense load, there is over a 50% gain.
So what do we make of the RX 9070 as a productivity hardware? We think it's a good card. If someone was looking for a GPU around that can do both gaming and crunch through some AI tasks this is a good card to pick up especially if you are dealing with single precision situations or some other VRAM-intense tasks. And we already know it is efficient so there's that too.
For those however looking for a GPU that can deal with more, AMD recently unveiled the Radeon AI PRO R9700 which is essentially a 32 GB refresh of the 9070 XT with some additional workstation-based optimizations.
Considering everything, we rate AMD's RX 9070 a 9 out of 10 for its AI performance. Price is less of a factor for those looking at productivity cases compared to ones considering the GPU for gaming, and as such, we felt it did quite decent overall and can be especially handy if you need more than 12 GB.
Purchase links: RX 9070 / XTAs an Amazon Associate we earn from qualifying purchases.
Tags
Report a problem with article
Follow @NeowinFeed
#amd #performance #benchmark #review #nvidia
AMD RX 9070 AI performance benchmark review vs 9070 XT, 7800 XT, Nvidia RTX 5070, 4070
Review
When you purchase through links on our site, we may earn an affiliate commission. Here’s how it works.
AMD RX 9070 AI performance benchmark review vs 9070 XT, 7800 XT, Nvidia RTX 5070, 4070
Sayan Sen
Neowin
@ssc_combater007 ·
May 24, 2025 14:40 EDT
Earlier this month, we shared the first part of our review of AMD's new RX 9070. It was about the gaming performance of the GPU, and we gave it a 7.5 out of 10. The 9070 XT, in contrast, received a full 10 out of 10.
The main reason for the lower score on the non-XT was the relatively high price and thus the poorer value it offered compared to the XT. We thought the price was much closer to the XT than it needed to be.
While the RX 9070 proved to be more power efficient than the XT, for a desktop gaming graphics card, value and performance typically take the front seat compared to something like power efficiency.
However, that may not be the case in terms of productivity which also takes into account things like power savings. Thus, similar to the one we did for the XT model, we are doing a dedicated productivity review for the RX 9070 as well where we compare to against the 9070 XT, 7800 XT, as well as Nvidia's 5070 and 4070.
AI performance is a very important metric in today's world and AMD also promised big improvements thanks to its underlying architectural improvements. We already had a taste of that with the XT model so now it's time to see how good the non-XT does here.
Before we get underway, this is a collaboration between Sayan Sen, and Steven Parker who lent us their test PC for this review. Speaking of which, here are the specs of the test PC:
Cooler Master MasterBox NR200P MAX
ASRock Z790 PG-ITX/TB4
Intel Core i7-14700K with Thermal Grizzly Carbonaut Pad
T-FORCE Delta RGB DDR57600MT/s CL362TB Kingston Fury Renegade SSD
Windows 11 24H2Drivers used for the 7800 XT, 9070 XT and 9070 were Adrenaline v24.30.31.03 / 25.3.1 RC, and for the Nvidia RTX 5070 and 4070, GeForce v572.47 was used.Sapphire Pulse 9070 XT, Nvidia 5070 FE, and Pulse 9070First up, we have Geekbench AI running on ONNX.
The RTX 5070 gets beaten by both the 9070 XT and 9070 in quantized and single precisionperformance. Similarly, the 4070 gets close to the 9070 in half-precisionperformance, but the latter is an enormous 30% faster in quantized score and nearly 12.2% better in single precision.
The reason for this beatdown is the amount of memory available to each card. The Nvidia GPUs have 12GB each and thus only do better in the FP16 precision tests since the other ones are more VRAM-intensive.
Next up, we move to UL Procyon suite starting with the Image generation benchmark.
We chose the Stable Diffusion XL FP16 test since this is the most intense workload available on Procyon suite. Similar to what we saw on Geekbench AI, the Nvidia GPUs to relatively better here as it is FP16 or half precision which means the used VRAM is lower.
So this is something to keep in mind again, if you wish to float32 AI workloads, it is likely that graphics cards with greater than 12 GB buffers would emerge as victors.
There is still a big improvement on the RX 9070 compared to the 7800 XT as we see a ~54% gain. This boost is due to improvements to the core architecture itself as VRAM capacities of both cards are the same at 16 Gigs.
Following image generation, we move to the text generation benchmark.
In this workload, we see the least impressive performance of the 9070 in terms of how much it improves over the 7800 XT. The former is up to ~7.25% faster here. The 9070 is also not as well-performing as the Nvidia 4070 in Phi and Mistral models, although it does do better in both the Llama tests.
Another odd result stood out here where the 5070 underperformed all the cards including the 7800 XT in Llama 2. We ran each test three times and considered the best score and so we are not exactly sure what happened here.
Wrapping up AI testing, we measured OpenCL throughput in Geekbench compute benchmark.
The RX 9070 did not fare well here at all even falling behind the 7800 XT and it is significantly slower than the three other cards. Interestingly, even the RTX 5070 could not beat the 4070 on OpenCL so perhaps this suggests that OpenCL optimization has not been a priority for either AMD or Nvidia this time. It could also be an issue with Geekbench itself.
Conclusion
We reach the end of our productivity performance review of the 9070 and we have to say we are fairly impressed but there is also a slight bit of disappointment. It is clear that the 9070 as well as the 9070 XT really shine when inferencing precision is higher, and that is due to the higher memory buffers they possess compared to the Nvidia 5070. But on FP16, the Nvidia cards pull ahead.
Still RNDA 4, including the RX 9070, see big boost over RDNA 3. As we noted in the image generation benchmark, which is an intense load, there is over a 50% gain.
So what do we make of the RX 9070 as a productivity hardware? We think it's a good card. If someone was looking for a GPU around that can do both gaming and crunch through some AI tasks this is a good card to pick up especially if you are dealing with single precision situations or some other VRAM-intense tasks. And we already know it is efficient so there's that too.
For those however looking for a GPU that can deal with more, AMD recently unveiled the Radeon AI PRO R9700 which is essentially a 32 GB refresh of the 9070 XT with some additional workstation-based optimizations.
Considering everything, we rate AMD's RX 9070 a 9 out of 10 for its AI performance. Price is less of a factor for those looking at productivity cases compared to ones considering the GPU for gaming, and as such, we felt it did quite decent overall and can be especially handy if you need more than 12 GB.
Purchase links: RX 9070 / XTAs an Amazon Associate we earn from qualifying purchases.
Tags
Report a problem with article
Follow @NeowinFeed
#amd #performance #benchmark #review #nvidia
·107 Vue