In HWiNFO, the GPU thermal limit was displayed as “83.0 C”
The temperature increased during rendering work as expected.
I have the following questions:
I am a bit worry about the working temperature. Is 83 deg C too high for GTX-1660?
Is the GPU hot spot temperature 97 deg C too high or acceptable? Is a difference between average temperature and hot spot temperature of 12 deg C normal?
Is it better to decrease the thermal limit to a lower value eg. 80 or 75 deg C?
nvidia Control Panel provides no option to display nor adjust the thermal limit. How can I change this value?
Is it a stock card? By that I mean, is it a Nvidia card straight from factory, or is it a version from another supplier, such as EVGA, MSI etc… If it is stock, then the temperatures you’re seeing aren’t crazy, as stock cards often have pretty limited cooling compared with others.
Do you have a fan profile setup at-all?
You can normally find out the maximum operating temp for your card. Most cards can handle 90+ degrees, but it’s not ideal for long periods. For reference, my 1070 goes to about 55c rendering in Cycles. Some games will send it more towards 60+, but if I disable my fan profile, then yeah, it’ll quickly reach 80+.
It higher on 2 degrees of your initial temperature limit.
I have GTX 1060 and it’s never been higher than ~60C as far as I remember that.
Change what? Your temperature limit or to reduce its index? Dunno what may the actual cause of such high temperatures in your own case, but suggest it is rather because of the scene or because of the problems with your GPU drivers and/or GPU itself.
Temperature over 80°C should be worrying. I think you should attempt to get it lower somehow. Cycles rendering seems to be really more intense on CPU and GPU than most other tasks. I have a PC in the office that is perfectly stable with anything else but Cycles rendering overheats it in a couple of hours and it restarts after a while when using CPU and GPU rendering together. Simply opening the case and leaving it open when rendering combined with increasing fan speed in the BIOS/UEFI solved the issue. I suppose, if I didn’t want to leave it with the case open I would try to install another fan in the case, or change the stock CPU cooler(and I think it might even help a bit with temperature for all other components in the case). I think it is even possible to upgrade GPU fans, but anything that you could do to get the temperature down even only inside the case might also help.
If a PC is not designed as a workstation computer with rendering in mind it is very well possible it might overheat when rendering. From my experience, it seems that most PCs and stock cooling solutions are not designed with rendering in mind. Higher temperatures risk damaging the parts like CPUs and GPUs and it even happens sometimes, although most hardware has some sort of safety mechanisms to prevent it so it’s not extremely likely, but it definitely shortens the life of the components and you might loose your renders and time when your computer shuts down because of overheating. It might not be enough to worry too much, because usually GPUs go out of date way sooner than the hardware wears down - you are probably not going to use your GPU 10 or 20 years, but it probably could work even 50 years fine, so this might not be as serious as it might seem, but still, you never know, and if you come up with something to improve the situation that does not require a lot of investment and effort(like an extra fan if you can fit it), I think you should do it just in case. Checking if all fans work OK(and also move the air in the right direction) and cleaning the dust is a good idea in any case.
My own GPU(RTX 2060) temprerature is usually 33°C, and it raises to around 65°C when rendering. Over 80°C would certainly worry me personally and I would attempt to do something about it, mainly to prevent possibility of renders crashing.
Similarly, for AMD GPUs, GPU Temperatures in the range of 65 to 75 °C are “normal” . Anything beyond these value means your GPU is overheating and you need to take care of it. The upper limit of Nvidia GPUs is 95 to 100 °C while it is 90 to 100 °C for AMD GPUs.
For 1660, the maximum temperature allowed for normal operation is 89 degrees. The average maximum temperature in use is around 75 degrees.
Try measuring the temperature with other software.
If the temperature is still high, please clean the dust on the graphics card.
If it’s a used graphics card for a long time, you need thermal grease redrawing.
For the 1060 I use, the maximum rendering temperature is around 67 degrees.
The CPU is up to 78 degrees.
The manual says higher temperatures are possible,
but higher temperatures inevitably lead to computational errors.
The best way for cooling a graphics card is: “open access” to normal room temperatued air ( ~20 degree)…
…so if the flow is somehow retarded… → bad… or it’s just sucking hot air from the computer inside → bad… (some people do everthing to make their PC silent… using foam on the inside at the wrong places and slowing down the airflow… even closing the in- or outtakes with foam…)
Sometimes it already helps to slear all obstacles around or at or over the computer case… or to use a vacum cleaner to get rid of any dust… in the case… (pull the plug of the computer before opening !)…
Be careful with this. Cycles rendering is really tough on GPUs if they don’t have enough thermal headroom. Remember that GPUs are not specifically designed for these workloads; Cycles will have your GPU operating at 100% load for potentially many hours.
I have personally destroyed a 1080 in the past by rendering Cycles animations without paying attention to the sustained high temperatures. I would strongly suggest improving your cooling setup.
Personally I just set my GPU fan speeds to 100% manually before starting a render job I know that will cause high temperatures. Best to leave the room then because it gets annoyingly loud.
If nothing else works you can try undervolting. It is a rather tedious process but can help. Before doing it read up on it as it can damage your GPU if done incorrectly.
I mean changing the value of GPU thermal limit. I didn’t set this value and why it is set to 83 C (may be a default hardware value)? I think there must be some ways to change it. It seems that the computer or GPU will try to converge to / not to exceed the this thermal limit.
What do you mean by “2 degrees of your initial temperature limit”. Are you comparing the hot spot temperature (97 C in my case). If you are referring to the working temperature, 95 C - 83 C should equal 12 C, not 2 C.
I installed the nvidia GeForce Studio driver only. It has no setup for fan curve nor any fan setup apps installed. I am not sure whether the GPU fan profile can be changed in the UEFI BIOS or not and I am not dare to do so.
This time, the rendering was done by using the Cycle renderer. I have make some rendering using EEVEE before and the GPU temperature was not that high (at most 70 ~ 75 C for a rendering period of 2 to 3 hour. Actually, can be changed a bit by adjusting the rendered image PNG compression ratio).
From the screen capture, it can be seen that the GPU change (speed up in this case) the fan(s) speed to make it stay within the thermal limit. So I think if we can change the value of thermal limit manually, we can lower the working temperature to the desired value.
Also, I noticed that when the rendered image was being saved, the GPU will stop working and only the CPU handle the file saving process. The GPU temperature will start to fall until the rendering of the next frame begin. If we can add an intentional delay before starting the next frame, the GPU working temperature can be reduced further.
I have tried removing the computer side cover (entire left side, no extra fan). I help a little but will eventually reach a similar working temperature but took a longer time.
The computer and GPU has been used since 2019 and I hope to continue for 4 to 6 more years before buying a new system.
Perhaps the default profile for fan operation is a midpoint between silent and maximum. You install Asus GPU Tweak II utility and test some profile where the speed of the fans scales quickly:
I know I am late here but it might help. Gpus are safe up to 85-90°C. Improve airflow, clean dust, use custom fan curves or undervolt to reduce temps and press alt + 227 to insert π!
Depends on what “safe” means for you. Is it going to catch fire or explode? Probably not. But you might still want to lower the temperature as much as possible. And it is possible. Easily, without much effort and pretty much for free. I think the issue is most often the airflow in the case. I did nothing about cooling my PC, I just have a DIY case that is completely open and my GPU temperature is at most 65° C(Asus RTX 3060 now) even when rendering for hours. I did not spend a single cent on cooling and I have 0 case fans(and 0 case to be honest). It’s just completely open. Just open the case.
Just for the fun of it. You are definitely allowed to laugh at me But that’s my PC:
It’s a very cheap PC and yeah it’s not for everyone… But hey… no more than 65° C when rendering even for 12 hours! So remember that when laughing. Can you spot my on button? Yes…
It’s the first attempt and I enjoyed the project so I might still upgrade it. I think it could be smaller.
But that’s my PC: