I’m a 3D artist, building a new rig that will be used for alot of GPU simuation and rendering. I want to have multiple GPU’s, and am looking for some advice on how to avoid bottlenecks, and in general get the most out of the tech.
The 3D software I use is Side Effects Houdini, which allows you to specify in the env variables, which card to use for the viewport, and which card to use for OpenCL simulations. Houdini does not use Nvlink. As far as I know the cards are independent. You cannot use 2 cards on one OpenCL sim. You can only specify which single card you want to use.
So currently I have a 1070ti 8gb card, and would like to use this strictly for the viewport. And then I would like to get a Titan RTX to use for OpenCL simulations, as well as rendering in Redshift, and Unreal engine.
If I build a machine with those 2 cards, am I going to run into issues? ie: the power of the Titan gets under-utilized because of having an older, smaller card like the 1070ti.
Any advice on how to build a machine with multiple cards, involving a Titan RTX, and some smaller cards? The CPU I’m looking at is the Threadripper 3970x.
Thanks, and apologies if I’ve posted this in the wrong place, I’m not super savy with hardware, but really need some solid advice before I drop a bunch of cash on a new rig.