NVIDIA Fermat OptiX-RTX ray tracing benchmark - NVIDIA TITAN V | GTX 1080 Ti | GTX 1080 | GTX 980 Ti

Hi, I’ve managed to set up NVIDIA’s Fermat physically based rendering system (NVlabs GitHub repo) and created a little benchmark from it, here are my numbers: http://boostclock.com/show/000219/gpu-rendering-nv-fermat-gtx980ti-gtx1080-gtx1080ti-titanv.html

Any feedback is appreciated, thanks!

Lovely!

Any plans to check out new consumer level RTX GPU’s (2080/2080 Ti)? Would love to see performance comparison with 1080/1080 TI in optix.

Glad you liked it!

Sure, I’ll test it with a 2080 in the coming days. Plus, I have other Optix Prime benchmark in the pipeline as well.

Unfortunateky, Optix5.2 is not released to the public just yet, but an optix dll is shipped with the GeForce driver pack, this makes things a bit unclear, I’ll see how it goes.

OptiX Prime does not take advantage of RT Cores so Fermat is not a good benchmark for Turing.

m_nyers is correct that since OptiX 5.2 is not yet publicly available it is not possible to benchmark Turing with OptiX.

Regarding the optix dll in the driver, we are implementing OptiX in the driver. We are not finished so the optix dll in the driver does not yet support the full OptiX API. However that dll is what was used for the demos at Siggraph.

Thanks for the information!

Is it worth using OptiX Prime, will Prime get any updates, or is it essentially “deprecated” with regard to the RT Core support and RTX acceleration?

Hi, updated results with RTX 2080. No Optix update available for Turing, so this is CUDA GPGPU ray tracing without RT Cores (Fermat, ProtoRay).

http://boostclock.com/show/000230/gpu-rendering-nv-gigarays-gtx980ti-gtx1080-gtx1080ti-rtx2080.html

Any feedback is appreciated, thanks!