V-Ray vs Karma - Lags while rendering

[Nevoreko]

Hi,

https://we.tl/t-LvV5sWhIAT

You will find in this package a video and a simple hip file with an instancer, we noticed how laggy Houdini was while rendering heavy instancers, the whole scene gets unworkable unless we stop hydra rendering.
Notice the fps counter in the video, it also translate to the overhaul houdini interface laggyness, while using karma everything is butter smooth but gets sluggish as soon as we use openGL (expected) or V-Ray...

Note that this is a simple example scene, but on heavy layout it gets much worse and you don't need to move in the viewport to see the issue.

Maybe OpenGL keeps running in the background? Maybe karma keeps a cpu thread "alive" to handle user operations outside of rendering?
Whatever the magic sauce is, would there be any way to improve this area to be on par with karma?

This Houdini forum post about openGL/Vulkan mentions "Background updates and rendering on a separate thread (much like Karma in the viewport)", it seems like it's what we might be missing with V-Ray
https://www.sidefx.com/forum/topic/9...=1#post-404597



V-Ray 6.2.05.22651
Houdini 20.0.590

Thanks,
Marvin

[bdancer]

> Maybe OpenGL keeps running in the background?

Not like it's running, but we use needsdepth=true (https://www.sidefx.com/docs/hdk/_h_d...s_d_hydra.html).

> This Houdini forum post about openGL/Vulkan ... it seems like it's what we might be missing with V-Ray

This has nothing to do with Hydra renderer delegates.

We'll check if it's a bug or smth of course for sure.

[Nevoreko]

I was only referring to the "Background updates and rendering on a separate thread (much like Karma in the viewport)" part of that post that mention that it's like Karma
So from what I understand, it sounds like V-Ray is "limited" to the performance of openGL? If that's the case there might be better ways to handle that to hit Karma's performance?