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 4080 Super AI Performance
GPUs, particularly high-end solutions like the RTX 4080 Super, can also used with professional applications, AI training and inferencing, and more. Along with our usual professional tests, we have Stable Diffusion benchmarks on the various GPUs. AI is a fast-moving sector, with many of the publicly available projects designed specifically to run on Nvidia GPUs. That may be changing, slowly, but Nvidia remains the leader in AI right now.
For our AI testing, we're using Automatic1111's Stable Diffusion project 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 that's tuned for the RDNA 3 GPUs we're testing.
Compared to what we already had in the form of vanilla RTX 4080, the RTX 4080 Super once again doesn't really change the status quo as far as performance goes. It's 2% faster in Stable Diffussion at 768x768, and we measured a 6% gain at 512x512 but that's likely an extra 2% margin of error (because there's no reason for the 4080 Super to be more than 5% ahead of the 4080).
The RTX 4090 of course reigns supreme, with 45% higher performance at 768x768. It also has 50% more memory, which means it's able to run larger models that simply don't fit into the 16GB on the 4080 Super and 4080. If you're serious about AI work, and a lot of companies are, it's easy enough to see why the 4090 prices have increased in the past year.
The RTX 4080 Super also easily outpaces anything from AMD in this discipline. It's 90% faster at 768x768 and 108% faster at 512x512 image generation. There may be workloads where AMD is more competitive, and like the 4090, the 7900 XTX also has 24GB of VRAM that can be useful for running larger models. But AI also needs a lot of compute, and that's an area where Nvidia's tensor cores push it far into the lead.
Nvidia RTX 4080 Super Professional Workloads
We use SPECviewperf 2020 v3 as one of our benchmarks for assessing professional workstation graphics performance. This is an area where Nvidia's consumer GPUs fall far behind at times, as it only offers full performance tuning in its drivers for professional GPUs like the RTX 5000 Ada Generation.
Depending on the application, the competition from AMD can be substantially faster, only moderately faster, or slightly behind — actually, the 3dsmax-07 test is the only one where the 4080 Super beat the 7900 XTX, and only by 2%. In the remaining seven tests, the 7900 XTX leads the 4080 Super by anywhere from 8% (maya-06) to nearly 200% (medical-03), with snx-04 being the outlier where AMD's GPU is over 1400% faster.
Naturally, the 4080 Super shows no real advantage over the 4080 once again. It's 2.6% faster overall, leading by 1.1–4.4 percent in the eight individual tests. And the 4090 delivers 25% higher performance overall, with a range from 3.7–40.3 percent. Most of the results are 20% or more, but creo-03 doesn't appear to scale much with the faster GPUs.
For GPU-accelerated 3D rendering, the only app that currently supports all three GPU vendors is Blender. It 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 Blender 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.
Like our ray tracing game results, the 4080 Super delivers a beatdown to AMD's top RX 7900 XTX. It's 156% faster overall, and 131–179 percent faster in the three individual tests. Nvidia also has hardware acceleration for other professional 3D rendering applications (see below).
The 4080 Super offers 3% higher performance than the vanilla 4080, so again nothing no write home about. The RTX 4090 meanwhile delivers around 30% more performance than the 4080 Super. For 3D rendering, time can directly equate to money, and the higher cost of the 4090 — or a professional GPU like the RTX 6000 Ada — can be justifiable.
Nvidia has RT hardware acceleration for Otoy OctaneRender and V-Ray, which do not have support for AMD or Intel GPUs. The 4080 Super occupies its normal spot, 1–4 percent ahead of the 4080 and 17–28 percent behind the beefier RTX 4090.
Nvidia RTX 4080 Super Content Creation Summary
As one of the fastest consumer GPUs around, it's more feasible for the RTX 4080 Super to be used as a semi-professional substitute that doesn't cost an arm and a leg. If you're mostly doing AI or 3D rendering, it's plenty fast, but Nvidia's pro GPUs can offer double the VRAM (or more) for larger workloads.
If you want a competent GPU for poking at AI, the 4080 Super also works well. It doesn't have as much memory as the 4090, but if you're not specifically looking to run the largest LLMs a consumer GPU can handle, it's a good solution for research and development purposes.
- MORE: Best Graphics Cards
- MORE: GPU Benchmarks and Hierarchy
- MORE: All Graphics Content
Nvidia denies cutting GPU supplies to China — chipmaker fires back at 'recent false rumors' aired on social media
Intel Arc B580 review: The new $249 GPU champion has arrived
Chinese Hygon 16-core chip trades blows with AMD Threadripper 1950X in Geekbench — Chinese chipmaker continues to leverage AMD's Zen 1 architecture
