I just noticed something weird in HWInfo64. It says my GPU thermal limit is 65, and it’s throttling. Uh, what? 65 is comfy cool; low even. Anyone know why the thermal limit would be reporting so low?
I don’t notice any performance hit in-game, so I’m skeptical it’s actually throttling. I have an EVGA 3050 XC Black.
65 Celsius is 149 Fahrenheit.
Google says the thermal limit of the EVGA 3050 XC Black is 93 C (199 F). As I said, 65 C is comfy cool.
Just running Grounded now with RTSS overlay, and when it goes above 65 C, the framerates don’t drop at all. Certainly not a dramatic thermal-throttling kind of drop. (I assume; I’ve never thermal throttled anything, to my knowledge.)
Here’s a thread from the HWInfo forums describing almost exactly the same situation (throttling reported despite a max temp of 68°) but with a 3070 TI.
Martin’s best guess is that it’s just a reporting bug in the 30-series Nvidia drivers. (Martin is the guy who wrote HWInfo.)
That’s unsatisfying, but I guess it means everything’s fine. Weird and annoying.
Anyone else with a 30-series card see any similar behavior?
I have a 3080 (sneak brag) – how do you want me to test it?
Using whatever reporting software you prefer, what does it say the GPU thermal limit is?
If you don’t have any reporting software, HWinfo64 is free.
I’m not so much interested in whether you get reports that it throttled or not; it’s the super low thermal limit that seems really weird to me. So you don’t even need to stress test it or run a game or anything. As shown in the screenshot in the OP, I get the 65° thermal limit report at idle.
Yeah, GPU thermal limits should be in the 90+ range. I’d assume a bug in the monitoring software. You can always run it in a benchmark like 3DMark and it should give you a rough comparison and place on the bell curve for similar systems.
You overclock; have you ever thermal throttled anything? I only ask because my assumption is that if you thermal throttle anything, your frame rates will drop like a rock in an extremely obvious manner. Do I have that right?
Well, my GPU is underclocked, not over, but yes if the chip (CPU or GPU) throttles, it’s pretty noticeable since it throttles hard to try to bring temps down. It doesn’t go from 5GHz to 4.85GHz, it goes right to 2.1GHz.
NVIDIA GeForce overlay has a temperature target of 83 degrees and caps out at 91. So, that seems to be in order.
Sweet, checking out GeForce Overlay (which I’ve never used before) is reporting 90 for my max, so yeah, appears to be the HWInfo is just reading the wrong sensor for thermal limit. And then falsely flagging thermal throttling when that low limit is reached.
Interestingly, GeForce Overlay is saying my target is 65. Looks like HWInfo is reading the target instead of the limit. (I don’t even know what target is.)
My guess is that the target is when the fans will really kick up to keep it at that temperature.
According to this Reddit topic on the Linux gaming sub, “target” is a user defined thermal limit. Looks like I may want to kick that up a bit, since supposedly your clock does throttle when you exceed that. Weird that I don’t notice any frame rate drop.
Think I might kick mine up to 75° and see what happens.
So finally got back on the computer and tried it. Raising the target in Nvidia Overlay from 65 to 75 warned me about overclocking (?) but once I accepted, the thermal limit in HWInfo now shows 75.
RTX 3070 here. My GeForce Experience Overlay shows a temp target of 83 C, which I never set manually so it must have been the default for me. I built my PC for good cooling, so even with a very beefy benchmarking programming I couldn’t get the GPU hotter than 74 C. (I was looking to see if throttling happened above the temp target. If it’s useful to you, I can lower my target setting and test again.)
Possible that YOU lowered it without realizing when you were trying to silence your 3050? Like assumed that setting a lower limit would result in less fan noise?
It’s usually the opposite. You run your GPU a bit hot so the fans can spin slower (or less often) for less noise. Trying to keep your GPU cool requires an aggressive fan profile and a lot of noise.
100% possible. I was using afterburner, and I don’t remember any Target setting, but I will definitely check again in a little bit. The only things I changed in afterburner were the CPU power, which I jacked all the way down, and I do see that setting in the Nvidia overlay correctly reflected as all the way down. (77%, I think it said.) That’s a different setting than the target in Nvidia overlay. But maybe afterburner pulled the target down silently because I pulled the power down, like it has them paired but doesn’t mention it. Or maybe even it does mention it and I was just too obtuse to notice.
I also then lowered the fan curve in afterburner, but I can’t imagine that would have done it. (Lowering the power dropped temps by like 5°, then lowering the fan curve raised it by 6° or 7°, so I ended up essentially back to where I started except very quiet.)
I’ll play around in afterburner and see if raising the power level changes the target as reported by HWinfo.
One thing I noticed after raising the target in Nvidia overlay: My frame rates are lower by almost 10. That seems weird to me, and I don’t love it. I may end up leaving the Target back down at 65° to get my 8 FPS back. The actual max temperature reported did not change; it still says 68.something, so it doesn’t appear the card wants to run much hotter anyway.
Yep, it was me. Simply re-applying my afterburner settings dropped the thermal max back down to 65. And now in afterburner I see the “Temp Limit” setting, clear as day, set all the way to the minimum with a value of 65. Just me being oblivious, nothing to see here.
Here’s a screenshot of my actual afterburner settings:
Looks like I tuned down everything, not just power limit. heh. Seems to have been running fine for months, though, so I’m just going to leave it as is.
Hah, suspected that was it, only because that was your first time playing with Afterburner and trying to undervolt, play with the fan curve, etc.