This is more of a “Vray” question, but I’m running my renders on my GTX 3090 (24gb RAM) and notice that during the render the GPU never gets past 10% utilization.
I have my settings in Vray set to using the RTX engine for rendering. Is there some other setting that helps me maximize my GPU on renders? I’m new to this so just trying to understand if I’m one setting away from a different outcome.
You can download the V-Ray benchmark and run it, see if you see any differences in the activity monitor.
@MikeWayzovski Wouldn’t Vray want to hit my GPU as hard as it can for the render? It feels like if I’m at 10% GPU max there’s something (setting?) that’s not properly utilizing it. Or, it may be used in some way that isn’t reflected in the 10% report.
Here are my current settings. Is there anything else I could be doing to squeeze more power from my GPU?
I think it depends on the model.
Maybe it’s not demanding enough.
You’ve selected the RTX render engine which uses the RT cores, rather than just the other parts of the GPU that CUDA uses for example- so it may be as simple as windows just can’t report on that part of the GPU.
If you switch to CUDA is that different?
You have also Interactive on and set to High so probably part of the system is reserved for waiting for you to do something.
@Anssi If I have that setting enabled but I don’t actually run the interactive render window does it matter?
@Elmtec-Adam Do you have any sense for whether the GTX/Vray combo uses both CUDA and RTX cores?
Well, it is a 3090 with 24 gb memory. It’s just cruising. I have the same card and can’t believe the speed I see with Lumion. Software needs to keep up.
I’m not sure to be honest, I think CUDA and RTX use entirely different APIs to communicate with the hardware and I don’t know which ones V-Ray uses for both of those.
I imagine that the CUDA option doesn’t touch the RT cores, as that uses the Tensor cores normally. Perhaps a question for V-Ray.
CPU, CUDA and RTX rendering modes in V-Ray are not quite 100% feature equivalent, but the only way to see performance differences would be to render the same scene in each and take the time.
