search

NVlabs / nvdiffrecmc

Official code for the NeurIPS 2022 paper "Shape, Light, and Material Decomposition from Images using Monte Carlo Rendering and Denoising".
Other
363 stars 28 forks source link

How about the speed on V100? #18

Open aoliao12138 opened 1 year ago

aoliao12138 commented 1 year ago

Nice work! I am curious about the rendering speed on the V100. Will the speed be much slower since there are no RTX cores on V100? Thank you very much!

jmunkberg commented 1 year ago

Thanks @aoliao12138 ,

Yes, we have noticed significant slowdown on V100 due to the lack of RTX cores, as we cast many rays to compute lighting in the hemisphere. One alternative may be to run nvdiffrec, https://github.com/NVlabs/nvdiffrec/, which is a similar reconstruction algorithm with a simpler shading model, which does not require ray tracing.