Users out there running either a single or dual 1080 Ti, is it more than enough power / speed for 3D rendering?
Whenever developers pickup the RTX RT capability, would it out perform the 1080 Ti significantly?
What do you guys think the RTX would bring to the table for 3D users?
I can afford either a single 2080 Ti, dual 1080 Ti, or a 1080 Ti and save money.
Replies
I noticed the 1080 Ti has 11 GB vRAM, while the 1070 Ti has 8 GB vRAM, I don't know how much of a difference is that since the "best" pc I had was running a 2GB vRAM card.
Here's chaosgroup's initial writeup on their experience with RTX. Quite promising, but it still might be some time before we see stable usage in public builds. It's important to note that they were included with Nvidia's conversation and perhaps some of the development of RTX, hard to say who elese was involved. I haven't sat down to look at renderer benchmarks from the 2080ti yet, so I don't know how much I could recommend that over dual (or even single) 1080Tis yet, given the current price. It's faster and on a new architecture, but I don't believe it's a huge leap over the 1080Ti and IMO this release seems rushed, and I really don't like this new "Gigarayzzzzz" buzzword they started throwing around. Given that RTX is primarily focused on realtime, we could expect IPRs to move a little quicker at the very least, and as extra power at the simplest.
As for 1080 vs dual 1070, firstly, the VRAM is not shared, that's only possible now with NVLink. So whatever you're doing needs to fit in each cards VRAM or else the renderer fails, unless it can do out-of-core. For price vs raw power in rendering, I'd suggest the dual 1070Tis; the clock speeds are lower, but you're getting much more CUDA than a single 1080Ti. The only 3D app that will really use your card in regular non-rendering work is Substance Painter.
Thank you for all that info, the V-Ray post cleared many questions I had, also, the NVLink sounds really good!