GPU distributed rendering

[danio]

Hi all,

We're currently considering adding some Titan Z's to one of the nodes of our CPU based render farm to explore GPU rendering as an option. Is there any downside to doing VRay RT GPU rendering over a network? Any features disabled, caveats or major slowdowns (aside from the typical transfering the scene issues.) Do assets still cache?

Thanks!
Dan

[savage309]

Hi all,

We're currently considering adding some Titan Z's to one of the nodes of our CPU based render farm to explore GPU rendering as an option. Is there any downside to doing VRay RT GPU rendering over a network? Any features disabled, caveats or major slowdowns (aside from the typical transfering the scene issues.) Do assets still cache?

Thanks!
Dan

Automatic transfer of UDIM textures with RT GPU is not working. This is a bug and is already fixed for the next SP.
Beside that, because the GPUs produce a lot of data and the network might become an issue. However, if you increase the Rays Per Pixel / Ray Bundle Size options from the V-Ray RT settings, the client will give the servers bigger chunks of work, and the data that has to be transferred over will be reduced. So in case the network becomes bottleneck, just do that.

Keep in mind that Titan Z has in practice 6GB of memory (not 12), because it is dual GPU. If you happen to need more, probably you should look at Titan X (it should not be much slower as well).

[danio]

Ok, great! Thank you for the information. We're going to do some exploring but this seems like a good option.

And I misspoke, the Titan X is the product we're looking at. Most of our scenes are at least 4-6 gb so we don't want to push it.

Best,
Dan