I have one 1080TI. What would a second one do for me using GPU render? My scenes are around 20MB, for the most part. In laymen’s terms, would I benefit from a second one?
Are they really 20Mb?
GPU memory is not additive, speed is though. So in terms of memory, no! In terms of speed, you’ll almost half your render time.
i suspect bobby missed a 0 of there (or two?)
i assume youre talking about the max file size? if so its pretty irrelevant. how much gpu ram do you use when rendering is the thing you need to look at. for example a 50km plane with a forestpack preset on it could be a 1mb max file, but could use 128GB of ram to render.
anway as stated above, if youve got 11gb gpu ram available now, youl still have 11gb, adding a second card, but youll render twice as fast.
if it was a 2080, 2080 TI, or (drool) Titan RTX you could use nvlink and in that case, you (almost) double the ram as it shares the ram of the two cards.
Yes, 20Gb, my bad. So, getting a second 1080TI would benefit me, regardless if my scene is larger than available v-ram?
The 1080TI is 1/2 the price of what I purchased my first one at. The 2080TI is priced at what my 1080TI was.
Your 1080Ti under win 10 will fit only 9-10gb.
If youre scene exceed that, no matter if you have 1-2 or 11 1080Tis stacked, you won’t be able to render.
GPU memory it is not additive!
To sum all the above, whatever you can fit in one Ti, it’s exactly what you can fit in an infinite number of them. So if you CAN NOT fit your scenes in one Ti, you won’t be able to fit them in 3-4 etc.
BTW… are you using TexOnDemand??
This will save you a lot of GPU memory.
Dealing with geometry usually it is not a problem, since you can fit hundreds of millions of polys under 1080Ti, maps are the ones that will kill your memory.
I never heard of TexOnDemand. I’ll look into it, thanks!
You are welcome!
2080 and 2080ti have nvlink. You need to have SLI working to use the nvlink with them. Nvlink allows stacking of memory, so 2 x 2080ti will be almost 22GB (with some overhead). There is no nvlink on 1080ti. Keep in mind that the nvlink bridge costs about 79$ and is purchased separately.
Both of the GPUs need to be nvlink compatible in order to use nvlink.
Best,
Blago.
That’s awesome!
I have V-Ray NEXT, and I am not seeing that. Maybe it’s auto now.
That whole box is grayed out.
I got it, thanks!
Would this card be worth it? How much faster than my 1080TI?
Take a look at this.
Although bare in mind that RT cores are not fully (or better to say, at all) utilised yet in V-Ray GPU!
@savage309 ??
So, it’s fast! I see that the top runners have multiple cards. I don’t know what you mean by cores not being utilized.
Yes it is fast! Yes you can stack as many as your setup allows you to (temps, pci lanes, space, psu, etc)
Nvidia 20x series, Turing architecture was release on 4thQ. of 2018 which introduced RT (RayTrace) Cores.
These cores though are not yet utilised almost from any software so far (at least in public releases).
What RT can do? You can have an idea here:
https://www.nvidia.com/en-us/design-…-architecture/
and here:
https://www.chaosgroup.com/blog/ray-…project-lavina
last but not least here:
https://www.chaosgroup.com/blog/what…endering-v-ray
Plus a bunch of other videos on youtube showing some impressive RealTime raytracing.