Performance Limit - Thermal randomly triggers

[Necrophagist98]

Hello! Lately I've been noticing that the Performance Limit - Thermal says "Yes" sometimes even though my temperatures are fine. I have a 2080 and my maximum temperature is usually 77 degrees, whereas the temp limit is 83C. By that logic, I absolutely cannot be thermal throttling. The weird thing is that sometimes it says "Yes" even when it's much much cooler. Like a few times my GPU temp was in the 60s and when I exited my game I saw that the max value for perf limit - thermal was "yes". I purposefully lowered my fan speeds and let the gpu go up to 81 degrees but that time the thermal limit said "No". What's even weirder is that yesterday I was idle all day. I literally didn't do anything on my computer all day but at some point the thermal limit triggered to "Yes" along with the voltage limits (I keep hwinfo on all day every day). What could this be? I have read that it might be a driver issue which explains why I started seeing this problem for the first time one month ago when I updated my drivers after not updating them for almost 2 years. I managed to log this in GPU-Z and the thermal limit triggered exactly for 1 second and immediately stopped. However, hwmonitor was on at the same time and didn't detect anything. How can it possibly throttle for one second and then continue to not throttle even though the temperature stays the same/gets higher? Any opinions would be greatly appreciated! Thank you!

 

[Kiriakos-GR]

It is wise to change hwinfo64 sampling rate to something higher than 1000ms (1 per second) , and to disable monitoring of non essential sensors.
I have select only 20 sensors as active monitor, and disabled everything else, sampling 10000ms when I am testing at gaming session. .

It is not unusual to see spikes at 1000ms with all sensors on.

 

[Martin]

Such spikes are not the fault of HWiNFO, those are really reported by the GPU driver. Several users have observed such spikes on recent NVIDIA GPUs for no apparent reason (temperatures below high threshold). There are multiple threads about this on this forum.

 

[Necrophagist98]

Such spikes are not the fault of HWiNFO, those are really reported by the GPU driver. Several users have observed such spikes on recent NVIDIA GPUs for no apparent reason (temperatures below high threshold). There are multiple threads about this on this forum.

I realized that when I saw that GPU-Z detected the spike at the same time as HWiNFO even though HWmonitor didn't. Thank you very much for replying. Since there are no clock drops or performance issues, I will assume it's simply a driver bug or something.

 

[Kiriakos-GR]

Every one using HWiNFO under enormous system stress due gaming.
By default HWiNFO this is going to log over 150 parameters in the unit of time (this is extreme as number).
No all computer systems are equal in performance (hardware).
Sudden spikes this translates that the PC cannot handle super fast logging of so many sensors.

HWiNFO it might started as small project, now its a Big one because entire PC includes enormous amount of sensor.
I might say something new, due this opportunity as conversation = minimum system requirements for all sensor logging at 1000ms interval.
Some systems they might not be able to complete 150 logging parameters within 1000ms = they need more time.

 

[Deleted member 19650]

Such spikes are not the fault of HWiNFO, those are really reported by the GPU driver. Several users have observed such spikes on recent NVIDIA GPUs for no apparent reason (temperatures below high threshold). There are multiple threads about this on this forum.

Hello! This is my first post in this forum.

I have been following this topic for a while and after some time testing trying to trigger this Thermal limit indicator on NVIDIA drivers I got a theory!

It is known that nvidia gpus boost 3.0 push the frequency a bit higher depending on the temperatures. In my case (3070 FE) at 60ºC my GPU drops its frequency by 15mhz. I have read that this should affects the frecuency every 10ºC but my gpu doesn´t get any higher than 60ish. I don´t know if this happens from 40C to 50C. Not in my case.

My theory is: the fact of exceeding 60ºC (or whatever temp the gpu boost 3.0 considers), which drops the frequency by 15mhz, triggers that thermal limit flag on nvidia drivers.

I´m not saying this is what is really happening, but is what make sense from my testing. I would like to know if someone else could replicate this on other hardware.

Regards!

 

[Martin]

That's an interesting theory... basically it would make sense as such clock reduction could be considered as a performance limiting factor.

 

[umeng2002]

I don't wish to necro a thread, but I've noticed the same thing moving from an RTX 2070 to a RTX 4070 Ti maxing out board power and voltage. The "GPU Temperature" never gets above 70 or 75, but the hot spot does creep into the low 90s in certain titles. Sometimes the Thermal Performance Limit is triggered and sometimes it's not. I can't really determine what hotspot temp triggers it or if it is as simple as hot spot trigger.

A polling rate of 500ms causes slightly higher hot spot temps. I don't think I remember seeing MSI Afterburner indicate a temp limit reached.

 

[umeng2002]

Edit: Since this forum's edit function doesn't work.

It did go off on Thermal Performance in MSI Afterburner, briefly like a flicker. 98% of the time, it's off on the power and voltage limits. All of this is with the "GPU Temperature" under 75 C and the "Hot Spot Temperature" under 95 C. It seems that there is no specific thermal threshold in either parameter which triggers the Thermal Performance Limit tag.

 

[umeng2002]

Yeah, it's definitely hidden driver stuff. I have the same hot spot temperatures, Core Temperature, and performance; but when I put the voltage increase to +50% instead of +100% in MSI Afterburner, it never triggers the Thermal Performance Limit flag.