Gpu slow down
VEGA
Posts: 86
Hello , I have a question if adding a second gpu gtx 1070, currently have gtx 1080, will slow down my 1080 ? Or I should only add the same gpu?
Daz 3D is part of 
Connect
DAZ Productions, Inc.
224 S 200 W, Suite #250
Salt Lake City, UT 84101
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2024 Daz Productions Inc. All Rights Reserved.
Comments
No it will not slow it down, becuase they are both running at about the same speed. It's only when you try mixing an older gen card like a 780ti with a 1080ti that you could see the 1080ti slowing down to match the megahertz of the slower 780ti,
So even with card like titan x (Maxwell) it would slow down but with 1070 it wouldnt ?
My understanding is the faster card slows down to accomodate the slower card.
I Have a system that runs a 1080ti and a 7series Titan Black.
The 1080ti runs at 1880mhz and the Titan black runs at around 1150mhz.
I monitor card speeds & Temps with gpuZ and afterburner. Neither monitoring app shows the 1080ti slowing down to match the speed of the slower card.
S.K.
I don't know why this rumor keeps coming up.
Iray treats each card separately - there is no interaction. I have a 980TI and a 1080TI - these numbers are from gpu-z and a render last night.
980TI Core clock: 1227 MHz and Memory clock 3304 MHz
1080TI Core clock 1898 MHz and Memory clock 5005 MHz
Both card running at 97% gpu utilization.
No sign of the 1080TI slowing down.
Good to know.
Interesting question, and I'm not sure there's a clear answer. When you get down to a software programming level, I think it comes down to how well Studio and Iray interact to all of the hundreds of individual rendering tasks to the two processors. And that gets real complicated, making sure you assign instructions/tasks with the right timing so the GPU's aren't sitting there waiting some of the time, and you're real efficient in how you separate the tasks between GPU's. Once you add a second GPU, the programming gets a lot more complicated.
Does a second GPU slow down the clock speed of the faster GPU? Probably not (other than maybe some PCI bus speed issues), but that doesn't necessarily mean the faster GPU is handling the same number of instructions per second as it was before, or that it isn't sitting there waiting for the other GPU due to some inefficiencies in how the instruction are programmed between GPU's.
Consider that when you add a second, identical GPU you don't actually halve the rendering time. It's not quite that good. I'm guessing that's due to both GPU's not performing as efficiently as they were on their own.
I have a 1080ti and a 1070, and honestly I'm not sure how I'd even determine if the 1080ti is performing as well in parallel with the 1070 as it would on its own.
There is no question, there is some interference with multiple GPUs but it is relatively minor. As near as I can tell, the Update Interval in Progressive Render Settings is the key player here. Every <n> seconds (default is 5) the Iray resident code retrieves the updated image from the GPUs and merges it into the visible render window. This is why two identical cards will not result in halving the render time.
Well if you look at the benchmark times for one 1070 it's around 3 minutes. And 2 x 1070 is 1.75 minutes, rather than the 1.5 minutes you'd expect from halving the render time. That's an extra 15 seconds longer than expected. It might seem like a significant amount to some for a 1.5 minute render. Though most of these times are ballpark, but at least it seems reasonable to assume the second card might have a significant negative effect. But that applies for adding any additional card I suppose.
Though clearly adding a 1080ti to a 1070 doesn't drop the render time to the same number as only 2 x 1070 (in fact it's much faster, as expected: 1 min vs. 1.75 min). So yeah, the 1070 doesn't necessarily cripple the 1080ti to drop to the 1070's performance level. But it seems reasonable to assume that any second card might diminish relative performance of the first card.
And my fuzzy memory is that there are some configurations of card vintages and PCI bus configs that also result in a significant diminishing of performance when adding the second card, which I suppose is what the OP is really asking. I thought there was some configuration that would drag the faster card down to the slower cards performance....
Or is it just my bad memory again?
EDIT: Oh, wait...it was my stupid kicking in again...I was thinking of the lower VRAM GPU being the limiting factor, dragging down the higher VRAM card. In any case, I do think that dual GPU's of different models, as well as the programming issues I already mentioned, might make a second GPU drag down the first one's relative performance.