Just installed my intel nuc hades canyon i7-8809g(Product code: BOXNUC8i7HVK2). First time i started it up it sounded like a vacuum-cleaner, or a graphics Card with fans on 100%. installed core temp and see the temps are a little over 50 doing nothing and just by browsing in Edge they rocket to above 70degrees, gives no sense? not reasonable that the temps raise like this? Also playing battlefield 4 for 10 minutes made one core reach a 100 degrees and another 99 degrees. Am I getting the right readings? Read a comment by a guy on reddit, that said the primary temp is controlled by a sensor directly in the cores and that makes it run like this, every time the cpu boosts it will activate the fans and make the temp raise like a lot within short time and down again. Any people have a solution to this? Not exactly a usable computer like it is now, i have the newest bios installed, 0053. Just disabled turbo boost and fans are not crazy now, temps are stil around 55degrees just writing this, but i am not sure about changeing primary temp sensor before i hear something here, guess intel would already have changed this in a bios update if this was the case. Will just keep the turbo boost off until i hear something here.
Also tried "quiet-mode" for fans in bios, did not help a lot cause of the temps jumping around as they do with turbo boost on. Right now running with fans in balanced and turbo mode off and seems quiet. But would be awesome with a solution where i can use the turbo boost on the cpu.
Hope somebody can help with this
See comment i found on reddit below
Text below is from Reddit:
I'm pretty sensitive to fan noise, so I had to handle this part, maybe this will be useful to you...
The problem is the temperature probe used to control the fans - the CPU is meant to boost until it hits a temperature limit, but the probe used for the fan control is the one right at the core(s).
Any boost will cause the temperature to spike almost immediately, and consequently cause the fans to ramp up to max. However, since there is heat transfer delay between outside the CPU and the core, that increased cooling will actually have zero effect on short-term boost, and next to nothing on long-term boost.
Thankfully it's possible to change the sensor used to a different one in the BIOS. Using any other sensor, the fans will stop their infuriating ramp-up/down behaviour.
To be safe, I set the minimum fan speed to the highest inaudible setting. During gaming and benchmarking it never ramped up again and also never hit core temperatures above 70C as far as I could see; while being effectively noiseless and still able to turbo significantly.
- Intel® NUC 8th Gen
That guy on Reddit is correct that the fans are primarily reacting to the processor's package temperature. The fan speed control decision-making is being done by firmware running on a microcontroller built into the Super I/O (SIO) chip. The SIO gets its processor package temperature readings over the PECI bus. The PECI bus interface (to the various DTS) inside the processor has a smoothing circuit that reduces the short-term volatility of readings received. Still, he is right that these processors heat up far more than expected, for longer than expected, for seemingly benign operations. Intel will need to answer for that.
Alright, that's as far as I will go associating with that guy. Being as nice as I can possibly be, I have to say, ONLY AN ABSOLUTE IDIOT WOULD REMOVE THE PROCESSOR TEMPERATURE FROM THE FAN SPEED CONTROL DECISION-MAKING. None of the other sensors correlate well with the temperature of the processor. They are on the opposite side of the board from the processor and are designed to measure area temperatures that will (loosely) correlate with the temperatures of the DRAM and SSD devices (whose temperature readings cannot be accessed by the microcontroller). While some heat from the processor and the AMD GPU will breed through the board and affect these sensor readings, this will only mean a few degrees of difference. The net result will be that the fan is not responding to processor temperatures at all. This change is thus a recipe for disaster. The only thing protecting the processor from thermal overrun would be the TCC circuit, which throttles processor performance in an (not always successful) attempt to protect the processor from thermal overrun damage. Please, don't do it.
Ok, down off my soapbox (and a half-hour layover to lower my blood pressure).
Next, let's talk about the preset algorithms that Intel provides in the BIOS. I am sorry, but they are useless. There is no tangible difference between them. I have told the NUC development team (of which I was a part before I retired) that these need to be improved, but they have not done so. As a result, I always use a custom configuration. I have detailed them here in the forums a number of times for various NUC models. I actually have a new set that I am using to make my HV even quieter. It is more aggressive than I have used previously (being relatively conservative), but folks (thank you, Leon) suggested I try it. Give this a try and see what you think:
Using the AIDA64 stress test, I did not see processor readings go above 91 and, even though the fans did accelerate, they never reached anywhere near the levels seen previously. I'm happy with it.
Hope this helps you as well,
Thanks for the very thorough reply and entertaining :D. Still surprised have much noise the fans made and that they started all the time without any workload. But maybe your settings will change that, I hope. Cause i really like it now, when it is quiet. Been quiet since I disabled the turbo boost, haven't heard the fans once. But of cause i wanna take advantage of the turbo boost, i kinda payed for it :)
I am using this fan configuration on a newly setup NUC7i7BNH as I have been seeing the "Core temperature above threshold, cpu clock throttled" while running Fedora. So far it has seems to significantly reduce the occurrence counts.
I also now have Splunk recording the core temps once a minute. It should be interesting to track the temps over time. My use case is as a general house server running Minecraft, Splunk, Plex, etc, so it does have some significant load spikes at times.
In the short time that this config has been running, I have had some processes that use 100% of two threads. During that the temp CPU temp shot up to 91-93C, but then quickly drops back down to a idle temp of 53C with occasional spikes to the low 70s. This does appear to be working well.
Also how often do you dust out the heatsink and blower?
Oh, I do this religiously, once every year (or so)...😉
Seriously, I always recommend doing this once a month. With the 4x4 NUCs, you can do it with a can of compressed air without even opening the chassis. The KY and HN/HV NUCs are a bit more work, however.
My biggest problem is that, with the number of test systems I have set up, I spend quite a bit of time outside blowing them out. The neighbors see me doing this -- and then come the calls to fix their PCs too...😣