To get the model into Rhino I exported the scene to the OBJ format, then imported that into Rhino. I had to fix up materials quite a bit. I made some minor corrections to the model as well (light fixtures were a bit off).
This scene I used to test my fix for multi-GPU rendering. Below you can see that I used an RTX A6000 and an RTX 5000. That is 1500 samples in 2 minutes and 8 seconds. The same scene with just the RTX A5000 would take around 3,5 minutes. This fix will be in Rhino 7.7.
Rolf, thanks, i’m lazy, I prefer a healthy disorder to a fake order!
We would have to define the whole frame (from where the light enters) … that part is only sketched … everything else seems good to me.
The composition of the scene is mine, but the credit goes to Vray. You can’t ask for better (even if it’s not the only great engine, there are many others …).
I did just now a rerender of this scene using Rhino Render with the multi-gpu using OptiX and that trims the rendering time down to 1m22s from 2m8s. That is only 64% the time it took multi-gpu CUDA.
For comparison, I rendered this scene earlier today on CPU with all cores selected (CPUx8), and rendering at resolution 800x1297 it took a whopping 1 hour, 50 minutes and 2 seconds…
fwiw: I wanted to know what V-ray would be capable of in 1min22s, on my relatively old hardware, so I quickly threw this scene together and this I think shows that with the right software, CPU rendering is still up to its task:
I rerendered with some more stats. It’s using the intel denoiser which is very fast. It kicks in after 4 samples/pixel (27 seconds) which gives already a usable image:
@nathanletwory that is one thing I have wanted to follow up, can you please make the denoiser kick in earler in the rendering? If I could I would have it on for every pass shown.
The denoisers in Rhino are all post-effects outside of my rendering. You’ll have to poke @DavidEranen, @andy and @johnc about that. FWIW, running the denoiser on every pass will slow down rendering quite a bit. Right now I believe denoising, if enabled, kicks in at about 5 seconds into the rendering and then once every second.
FWIW2 I tend to turn off all post effects, since they take up processing power. Even gamma, and I would also turn off tonemapping if possible at all. That will speed up rendering also quite a bit. I enable them once the rendering is done.
exactly that. I always need many test renders before the final result, this denoising is a huge time saver to more quickly evaluate the result closer to what the render is converging to. Compared to the time per pass there is hardly any penalty.
