Large GPU Memory Workflows
So the one thing this card offers that say the RTX 3080 does not is the large amount of VRAM. This enables to GPU to not only run, but perform better in certain applications and scenarios.
The first test we are going to run is OctaneRender. OctaneRender is the world’s first and fastest unbiased, spectrally correct GPU render engine, delivering quality and speed unrivaled by any production renderer on the market. We render a scene using both the RTX 3080 and RTX 3090. The scene actually won’t render at all on the RTX 3080 as it does not have enough VRAM. We can get past this by changing the “Out-of-core” settings and using some of our system memory. Even doing this on the RTX 3080, you can see it gets smoked by the RTX 3090.
The RTX 3090 was able to render the scene in just 37 seconds, this is compared to the 5 minutes and 19 seconds of the RTX 3080. Just think of how much time will be saved using the RTX 3090 instead of the RTX 3080.
Our next test is Black Magic DaVinci Resolve 16. Here we take an 8K R3D RED CAMERA clip. On both the RTX 3080 and RTX 3090 the clip will play back just fine, but when we apply a Motion Blur effect it actually exceeds the amount of memory available on the RTX 3080 so it will give you an error.
The RTX 3090 plays back the clip no problem, even with the effect applied.
Finally we have Blender. Blender is a rendering engine. We load up a scene and try and render it with both the RTX 3080 and RTX 3090. While the scene loads on each card it is only able to render on the RTX 3090. On the RTX 3080 Blender loads scene assets into GPU memory for final frame rendering, but it will exceed 10GB and will either cause Blender to crash or show a rendering error.