OT: Monitors and GPU Temps
Wow, I learned something new today...
I have a desktop with a GTX 1070 and GTX 1080ti. And I generally connect my 3 monitors to the 1070. And I noticed that in idle, the 1070 temps were a lot higher than the 1080ti. Like 20C higher or more. The 1080ti was in the low 30's, and fans off, but the 1070 was almost at 50C with fans on.
So it occurred to me maybe the reason was the 1070 was heating up because it had to feed the monitors. So today I swapped the monitors to the 1080ti, and what do you know....the 1070 was running at 31C, and the 1080ti started heating up to just over 60C, and then the fans went on. Even with GPU-Z showing 0 utilization on the GPU, it's still warming up to feed the monitors. A lot. And apparently this 1080ti is set to turn the fans on at 60C.
Not sure this means much, just kinda interesting.
Comments
The one of the marketing bits for my GTX 1080 ti is that under light loads it will run silent. I did not like dle temperature being that high, so I changed up the fan profile EVGA's monitoring software. Now it idles around 30c with both fans spinning a little under 2000rpm.
Yeah, it looks like my 1070 is idling (while powering all 3 monitors) at about 48C, which is 27C above an ambient of 21C. And that's with the fan running at 30% speed. I guess it runs like that continuously because of the monitors. I just never noticed.
Even if it was 60C I wouldn't be concerned. Heck, since the manufacturer doesn't even turn the fans of the 1080ti on until 60C I think it's clear that it has a lot of room to go before you need to worry.
I know, I just don't like my components running that warm. I would rather all that generated heat be removed from the computer, rather than having the computer do a mad dash to cool componenets when load is applied.
Yes - and no. The fan profiles are set for gaming, when it is unusual to run the gpu at 95+ utilization. With DAZ and Iray, 95% is 'normal' and once the gpu gets ahead of the fans they have troble bringing the temperature down, so the card starts slowing the clock speed to keep the teperature from exceeding the recommended value.
It's always a good idea to set up custom fan profiles and start then fans at a much lower gpu usage, ramping up as the usage goes up.
Right, but I posted something recently from Tom's Hardware saying that typically it isn't until 100C + that performance throttling occurs on CPU's, and I assume it's similar with GPU's. But since the fans are only on at 30% speed at 60C, it seems clear that you're not even close to any danger zone.
I have a 980ti and a 1080ti; they both throttle at 80+ degrees. The 980ti is limited at 83 degrees and I have seen the clocks drop back to near idle speed for extended periods of time (30+ minutes); the 1080ti throttles at 85 degrees. In both cases a render that should take 30 to 40 minutes ends up taking hours.
Yeah, you're probably right. I guess it's not a straightforward thing that applies across the board. Probably depends on whether you're at base clock or boost, and whether it's throttling down below the rated speed or not. Maybe in the mid 80s it only throttles back to base clock? And maybe up in the much higher numbers it throttles below the base clock speed?
And it probably depends on the individual GPU's and a bunch of stuff.
Though I'm surprised the throttling in your case results in such a huge increase in render times. You'd think if it throttles by, say, 10-20% then render times would increase by the same amount.
Heck, I dunno.
Yeah - I'm on my laptop at a wifi hotspot, or I'd drop my fan profiles and fire up one of my long-running renders to see just how far down the clocks go (quite a bit below base, that's for sure). With the fan profiles I use I end up running the 980ti at 70 to 74 degrees and the 1080ti at 62 to 65 degrees (the 980ti is above the 1080ti) and the gpu on both is running at 97% and full clock speeds.