Currently I have a Titan Z so I have (in theory) 2 x 2880 CUDA cores, 2 x 6Gb of GDDR and 384bit x 2 bus width. I have maybe in a couple months the possibility to get a 980Ti that has only half of the specs of the Titan Z (Titan Z is a dual GPU…)
And now comes the question: for making calculations in Unreal Engine for GPU lightmass (obviously RT is discarded as it’s not supported), what card is better or has the best performance?
And another one: do you know if at some point UE will have multiGPU support? Because in that case it could take advantage of both CPU’s for “rendering”, righ?
Just for you to know, both cards come from a friend that works in VFX and is making hardware upgrades so I have them for free. And no, I have no money to buy a graphic card (just in case )
The best to do is try it. It’s most likely going to function somewhat well for most things in the engine. A few things that are probably going to be slower are material editor updates to the preview render, large landscape foliage and material modifying, and potentially processing shaders at first opening of a project or after certain settings changes and restart of a project. But I discovered a way to drastically reduce that and could share a link to the video of it if interested. Lightmass calculations wouldn’t be too slow, most likely, even for larger scale projects. There’s also level streaming and perf optimization. If using GPU Lightmass, I don’t know how speeds are for that, and it’s not currently supported, apparently, in the latest version (4.25.3). It’s supposedly set to be supported in 4.26. Lots of issues were encountered in the latest 4.25 versions, especially lighting / shadowing noise.
)