OctaneBench
The benchmark allows testing a suitable NVIDIA your GPU with OctaneRender and RTX On. This provides a level playing field by ensuring that everyone uses the same version, the same scenes, and the same settings. The workload from the Reviewers Guide used in the OctaneRender demo is rather unrepresentative for the PC-wide test, so I prefer to use the benchmark tool distributed by otoy itself. The differences between the similarly fast cards are somewhat larger here. I’ll save the really big VRAM bombs for later, the Tensor cores are scurrying around here for now.
V-Ray 5 Benchmark
With V-Ray plugins you create photorealistic images and animations and they work seamlessly with 3ds Max, Cinema 4D, Houdini, Maya, Nuke, Revit, Rhino, SketchUp and Unreal. The V-Ray Benchmark is a free standalone application that allows you to test how fast your system renders. It is easy to perform and fast, and includes three rendering engine tests:
V-Ray – CPU compatible
V-Ray GPU CUDA – GPU and CPU compatible
V-Ray GPU RTX – RTX GPU compatible
Three self-created test scenes are also included to put each V-Ray 5 render engine through its paces. In CUDA mode, the differences between the cards are the biggest, so I preferred this one. By the way, the test system doesn’t look that bad in direct comparison to the publicly available Chaos scores, and it is in a good position because no overclocked card was used.
Blender 3.2.1
Blender Benchmark 3’s score refers to the hardware’s pure ability to process samples during Cycles rendering, i.e. the number of samples per minute that a GPU can compute. The higher this number is, the better (for me). I deliberately chose OptiX because the speed gains are striking, even though all the big cards are already quite close. The performance drop on the RTX 3080 Ti and below is probably simply due to too little memory.
However, in order not to test OptiX as NVIDIA-only, I also leave the paths of the reviewer guide a bit here and rather render my own igoBOT. So that the memory is shoveled up as threatened, there are the tin journeyman eight times! We see the same effect as with Blender Benchmark 3, only much more pronounced. The result shows the seconds the system needs for rendering including denoising. By the way, I am testing CUDA vs. OpenCL so that the AMD cards can also play along.
I think Jensen will still be happy about the moment when CUDA plays into the hands of the green camp so grandiosely, even if OpenCL could gain some ground in Blender in the meantime. If you want it AMD-only, you’d better grab Pro Render as an extra plugin.
29 Antworten
Kommentar
Lade neue Kommentare
Veteran
1
Veteran
Urgestein
1
1
Urgestein
Urgestein
1
Urgestein
Urgestein
1
Neuling
Veteran
1
Mitglied
Alle Kommentare lesen unter igor´sLAB Community →