Blender Benchmark
We are using the Blender Open Data Blender Benchmark to benchmark how long it takes to render Blender scenes. The version of Blender being used is 2.93.1. OpenCL is only supported on the AMD Radeon RX 6900 XT. For the RTX 3090, we can run in CUDA and NVIDIA OptiX. We will include both of these results in the graphs compared to the Radeon RX 6900 XT in OpenCL. The graphs are shown in time to complete, with lower being faster and better. We have graphed out each scene in the benchmark.
In the bmw27 scene, it takes the Radeon RX 6900 XT 37 seconds to render, while the RTX 3090 with CUDA is much faster at 18 seconds, and with OptiX it’s even faster at only 9 seconds.
The classroom scene is interesting, the AMD Radeon RX 6900 XT is faster than the RTX 3090. The Radeon RX 6900 XT in OpenCL renders it in 71 seconds, beating the RTX 3090 with CUDA which is 79 seconds. However, when we use the OptiX renderer the RTX 3090 more than half that time and is much faster.
In the fishy_cat scene, it takes the Radeon RX 6900 XT 62 seconds to render, while the RTX 3090 with CUDA takes 40 seconds. Turn on OptiX and it goes down to 18 seconds.
The koro scene takes 63 seconds on the Radeon RX 6900 XT and 58 seconds on the RTX 3090 with CUDA. They are close until you enable OptiX and the RTX 3090 is much faster.
In the pavillon_barcelona scene, the Radeon RX 6900 XT with OpenCL takes 127 seconds to render, while it is only 99 seconds on the RTX 3090 with CUDA. Turn on OptiX and that goes down to 44 seconds.
The victor scene takes a while to render out, with the AMD Radeon RX 6900 XT in OpenCL it’s a full 3 minutes and 48 seconds. That time is cut down to just 3 minutes with the RTX 3090 under CUDA. However enable OptiX and that time is cut way down to just 1 minute and 11 seconds, a much improved time.