Why you can trust Tom's Hardware Our expert reviewers spend hours testing and comparing products and services so you can choose the best for you. Find out more about how we test.
Nvidia RTX 4070 Super AI Performance
GPUs are also used with professional applications, AI training and inferencing, and more. Along with our usual professional tests, we've added Stable Diffusion benchmarks on the various GPUs. AI is a fast-moving sector, and it seems like 95% or more of the publicly available projects are designed for Nvidia GPUs. Those Tensor cores aren't just for DLSS, in other words. Intel has XMX (Xe Matrix eXtensions) for a similar performance boost, while AMD has "AI Accelerators" in RDNA 3 to help boost FP16 performance. Let's start with our AI testing and then hit the professional apps.
We're using Automatic1111's Stable Diffusion version for the Nvidia cards, with the TensorRT extensions enabled. Intel has an OpenVINO fork of Automatic1111's repository, which shows about a 50% improvement over our previous testing. Finally, AMD has a DirectML fork of Automatic1111, though there are some limitations. Specifically, the DirectML fork seems tuned specifically for RDNA 3 GPUs. In theory, it should also work with RDNA 2, Intel, and Nvidia GPUs, but performance tends to be far worse (RDNA 2), or else it doesn't work at all (last time we tried Intel and Nvidia GPUs).
As we saw with the gaming results, the 4070 Super lands right between the 4070 and 4070 Ti. This time it's closer to splitting the difference, with 12% more performance than the 4070 and 11% less performance than the 4070 Ti — that's at 512x512. For 768x768, it's 11% faster and 10% slower to those same GPUs.
AMD's GPUs can in theory run larger LLMs as they have more VRAM, but Nvidia offers far more AI compute performance than AMD. The 4070 Super is 40% and 30% faster at 512x512 and 768x768, when compared with AMD's current top GPU, the 7900 XTX.
Nvidia RTX 4070 Super Professional Workloads
We use SPECviewperf 2020 v3 to assess professional workstation graphics performance on the various GPUs. Nvidia only offers full performance tuning in its drivers for professional GPUs like the RTX 4500 Ada Generation (which also uses AD104 silicon), while AMD tends to be more forthcoming with driver tuning for its consumer cards.
As a result, in the overall view of performance using the geometric mean of all eight tests, the 4070 Super only manages to match the RX 7600, and falls far behind the 7900 XT. Nvidia would point users interested in better performance — particularly in the Medical and SNX workflows — to its workstation offerings, but of course the RTX 4500 Ada costs over three times as much as the 4070 Super.
For GPU-accelerated 3D rendering, the only app that currently supports all three GPU vendors is Blender. It now leverages the ray tracing hardware to boost performance, and that generally puts Nvidia in the pole position. We're using Blender Benchmark, version 3.6.0, though version 4.0.0 is also available. We've found some GPUs, including all of the Nvidia models, perform better with version 3.6.0, and it's not clear whether 4.0.0 executes the same workload or if it does extra calculations to deliver a better result.
Given the use of ray tracing hardware, it's no surprise that Nvidia's GPUs come out far ahead of AMD's offerings. This is basically like the path tracing games results, and the 4070 Super is 84% faster than AMD's 7900 XTX.
Nvidia also has RT hardware acceleration for Otoy OctaneRender and V-Ray, but the other GPU vendors currently lack hardware acceleration for these benchmarks. As such, we once again see the 4070 Super landing between the 4070 and 4070 Ti, but closer to the Ti in overall ray tracing performance.
Nvidia RTX 4070 Super Content Creation Summary
Professional workloads running on a consumer GPU are a bit of a wild card situation. Certain applications do really well — like ray tracing and AI — while others end up with limited driver tuning. AMD loses in the RT and AI results but wins in the SPECviewperf results, while the reverse is true of the Nvidia 4070 Super.
If you're more of a hobbyist trying to learn how to use some of these professional applications, it's possible to get started using consumer hardware. Once you become a working professional doing CAD/CAM or other work, however, the cost of the GPU hardware often becomes far less of a concern. Most companies working in these industries provide the necessary hardware, with the higher cost being far outweighed by the increase in productivity and throughput.
- MORE: Best Graphics Cards
- MORE: GPU Benchmarks and Hierarchy
- MORE: All Graphics Content
