Render times and possible upgrades ?
in The Commons
Hey guys,
I wanted some opinions on a good option for a new graphic card that can improve my render times. Right now I'm using an i7-6700k, 16gb ram and a GTX 980 TI which is decent setup I think, but my Iray renders still seem slow. I wanted to boost the performance in some way. I was looking into maybe switching cards to a GTX 1070 or 1080, but I'm not sure if that will actually help that much.
Any thoughts or tips ?
Thanks in advance to anyone who replies, I appreciate it!

Comments
If your motherboard and power supply can handle it, then you would be best adding a GTX 1070 or 1080 to your box, not replacing the very good card you already have. GTX 1080 has similar performance to a GTX 980 TI, although it has an extra 2GB of VRAM. Both cards working together on your renders will give you the best speed up.
...actually the 1080 draws 70w less power than the 980Ti (180w vs 250w) and while it has about 250 fewer CUDA cores it does offer 2 GB more VRAM meaning bigger scenes could be rendered.
The 8 GB 1070 draws 10w less power than my old 1 GB 460 does (150w for the 1070 vs 160w for the 460).
This would be a good idea, but as far as I know, you can't SLI two different cards, so my 980 TI with a 1080 wouldn't work

I tried looking for another 980 TI but it has since been discontinued, so either I buy one used somewhere, get lucky and find it in some store or would have to get two new cards....which is kind of expensive and not in my plans, so idk
Iray doesn't use SLI anyway, it uses each card separately. Note, however, that while the 1070 has more memory than the 980Ti it isn't faster - even the 1080, from the benchmarks I've seen, is pretty close to the 980Ti as long as the sceen fits into RAM on both.
You don't say how slow slow is - if it's really slow you might want to check that the GPU is being used in the first place, if the scene is large then it may not fit into memory. In that case yoiu might be better trying to optimise memory use - check for excessive levels of SubD (remember that if you have items using the surface setting for Displacement Subdivision that applies to the whole model - and each level quadruples the polygon count), try using lower resolution versions of the textures (there's at least one script to help with this).
If it is fitting into memory check for many shiny surfaces in enclosed spaces which can let the light paths bounce around for ages (there is a max path length setting in Render Settings, under optimisation, which can help to tame that), and for translucent surfaces that allow you to partially see through them. Extensive areas that rely only on bounce light will also slow a render as it takes a long time for them to get enough rays to converge - adding more basic light to the scene and adjusting the Tone Mapping settings may help there.
As Richard said, no SLI is needed for DAZ. I think SLI is only important for games.
A 1070 might indeed be a little slower than a 980ti, but if you have a 980ti and a 1070 running together, well you might get quite a big leap in speed. Or you could just get another 980ti and double them up. As long as you have an open slot for one and enough power for them, no problem. Also, Nvidia is just about to release a 1080ti next month, so there's that.
But really, a 980ti is slow? I feel like such a peasant now.