I would say yes. Consider upping your fan speed by adjusting the minimum duty cycle.
My NUC 6 went South at your temps.
I have been running a 2 year old NUC at 24 Minimum Duty Cycle , 1900 RPM and a CPU temp of 60 degrees. My returned NUC took a higher RPM to keep it cool than does the old NUC,
The above is just a guess. With the 6 I went for lower RPM and higher temp and I paid for it.
Thanks for the input. Yesterday was my first full day of developing on this machine.
I had it set to Power Level: Maximum Performance and Cooling set to quiet after 6 hrs it was bouncing between 80 - 85 c at one point I noticed it jumped to 90c, the fans never went over 3500rpm.
I immediately rebooted into the bios and put it to Power Level: Balanced and the temp leveled off around 65c.
The performance at Maximum is fantastic, at Balanced it's still quick but leaves a little to be desired. I am kind of disappointed, as I'm not a hardware guru and assumed that the sky lake platform would deal with heat better.
I'm considering returning and going to my original plan of building a actual fanless i5 machine.
It is not so much the Skylake platform itself that is to blame for not dealing with the heat. The Skylake platform, just like any other integrated circuit has a maximum temperature it can take before it breaks. To mitigate this problem chip designers use ever lower voltages which lead to less losses in semiconductor switches and less heat. That would be perfect, but for the fact that we want ever faster systems. The lower voltages enables higher clock speeds, leading again to higher losses and more heat.
The real problem is the cooling capacity of the system the Skylake is in. Although the TDP of this NUC is only 15 watts, that heat is generated in a very tight spot, on the die. The cooling solution needs to take that heat off of the die ASAP and transport it to to the ambient air to get rid of it. The first step, off the die, is done via thermal paste and a copper structure, which conducts heat great. But then, to get this heat 'out' of this copper heatsink and out of the box, into the air, it needs a decent sized fan and / or a big copper-to-air surface. Unfortunately we want our NUCs to be as small as possible. And that is why there's always a tradeoff between size, noise and performance.
I had it set to maximum performance for about 5 hrs ,the fan never got over 3500 rpm and I noticed it hit 90 c, I immediately rebooted and set it to Balanced.
The fan stayed at 3000 rpm and temp was low 60's.
I will say I was very impressed with the performance on Maximum, and satisfied with the performance on Balanced just not impressed.
1 of 1 people found this helpful
I have noticed that the configuration for the fan speed control in the SY NUC is incorrect and causes your thermal situation when using Maximum Performance to be much higher than it should be. If you would like to address this, you need to do so from within Visual BIOS...
At power-on or reset, immediately being pressing the F2 key over and over until Visual BIOS starts up. Click on Advanced and then Cooling. Then, on the right-hand side of the display, make the following changes:
- Set the Fan Control Mode to Custom.
- Set the Minimum Duty Cycle (%) to 25.
- Set the Primary Temperature Sensor to Processor [this is usually the default].
- For this sensor, set the Minimum Temperature (°C) to 60 and the Duty Cycle Increment (%/°C) to 3.
- Set the Secondary Temperature Sensor to Memory [this is usually the default].
- For this sensor, set the Minimum Temperature (°C) to 55 and the Duty Cycle Increment (%/°C) to 3.
Make sure, when you exit Visual BIOS, that you save the configuration.
Ok, these setting will ensure that the Processor and Memory will stay within their proper operating ranges even if you configure Windows for Maximum Performance. Yes, the blower will run at higher speeds (and the processor will run somewhat hotter), but this is the penalty you pay for performance from this small a system.
To add to some of the earlier comments, the issue with systems this small is their requirements for cooling. Using a blower is forced by the very small space available for the cooling solution. The problem with a blower is that its airflow generation capability is less efficient at any particular speed (compared to normal fans anyway). It's the price you pay. Unless you are willing to pay a significantly higher amount for a special chassis, you are going to give up significant performance to run fanless - and this expensive chassis will be significantly larger to get the thermal mass and surface area necessary to dissipate the heat being generated.
Hope this helps,
Personally I do not see a reason to set the minimum temp so (relatively) low. In my system (same hardware as TS) I set the min duty% to 20, the min CPU temp to 70 deg and the step to 3%/deg.
This makes for a very silent NUC that is mostly running in the fifties, only to go up in the sixties when I actually do something beyond staring at a web page. Without the extra fan noise. If I really give the CPU something to do (running a test that stresses 2 cores to a 100% ) the temp shoots up into the mid-seventies and stays there. The fan speed follows more slowly and cycles between 2600 - 3100 rpm. This situation continues as long as I let the test run, in other words, the system does not get any hotter. If I then close all vents the fan speed increases to just above 3600 rpm. CPU temp increases to 79 degrees. As soon as I open the vents again the system goes back to the cycling described above.
As the i5-6260U is rated for a 100 degrees junction temp and assuming the temp sensor is on-die I see no real reason to let the fan start earlier, even taking into account that in summer ambient might be 10 degrees hotter. And then there is (probably, I do not know this for certain in this CPU) thermal throttling to lower the load when really necessary.
But maybe I'm mistaken and these settings are lining my NUC up for a premature dead :-)
First of all, the specified minimum duty cycle for the blowers used in the NUC is 25%. Using a lower setting is not helping you acoustically and could lower the life of the fan if it has to struggle to run at lower duty cycle settings.
As for the minimum temperature setting, you have to decide what slope (rate of increase) you want the fan to have. There are two schools of thought on this (the field of psycho-acoustics studies the phenomenon involved). First some background. You want the fan to reach 100% duty cycle at or before the control temperatures for the processor and memory. The processors have a Tjmax (Maximum Junction Temperature, the point where processor throttling occurs to protect the processor from thermal damage) temperature around 100°C and a Tcontrol temperature in the vicinity of 80°C (you want fan at 100% by then). Most DIMMs have a Maximum Operating Temperature (MOT) around 75°C (again, you want the fan at 100% by then). Ok, two schools of thought:
- By choosing a lower minimum temperature, the rate at which the fan needs to increase (per degree) is lower. The advantage of a slower rate of increase is that it is less noticeable and less likely to become a source of irritation. As well, by applying cooling earlier in the process, you typically do not have to utilize the fan at high speeds (unless, of course, you sustain the load long enough to completely saturate the thermal solution). The disadvantage of this is approach is that the fan is speeding up - and producing more noise - at a lower temperature. Depending upon the individual acoustics of the particular fan, this can be a source of irritation.
- By choosing a higher minimum temperature, the rate at which the fan needs to increase (per degree) is higher. The advantage of the higher minimum temperature is that the fan is not sped up until this higher temperature threshold is reached and, as a result, you have a wider range of (temperature) operation at the minimum duty cycle where the fan is not noticeable. The disadvantages of this approach are that (a) once you reach the minimum temperature, the rate at which the fan speed has to increase is higher, much more noticeable and much more likely to become a source of irritation and (b) at higher temperature levels, the rate of thermal increase is often faster. Since the affects of the fan speed increase can take some time to manifest, temperatures may overshoot the control temperature. If this happens often enough, long-term damage to the processor could occur.
Which of these schools of thought you adhere to is your choice. Since I want to minimize the overall acoustics, I usually choose the first.
As a final note, remember that the airflow generated by this blower is also responsible for the cooling of all of the other components on the board, even those on the bottom...
Another note: While the Tjmax threshold is designed to protect the processor from thermal damage, damage can still occur over time at levels above Tcontrol. Secondly, remember that the heat from the processor is travelling through the board and into other components - and these components don't have the thermal dissipation capabilities (i.e. no heatsink-fan unit like the processor) and can be damaged at levels much, much below that sustainable by the processor. Consequently, I caution you not to rely on the processor's automatic throttling for protection; by the time it kicks in, damage can have occurred elsewhere...
Point taken. I was not about to remove the fan and rely on the thermal cutout of the CPU for protection :-)
Without wanting to start a holy war, but to make sure I do understand you correctly (some tunnel view is not beyond me :-) what components are we talking about that "cant take the heat"? Each and every discrete component in a computer can take way beyond 100 deg. Industrial IC's can take 85 degrees. Assemblies like the wireless module are specced to at least 80 degrees working.
The CPU is the only real, 15W heat source in this system. Its temperature is never going to be above a 100 deg on-die, blocked by thermal throttling. Stress testing my NUC revealed that I could not raise the temperature of the CPU over 76 degrees no matter how long I left it. Assuming that I do not remove or block the fan and assuming a normal Tambient I think that it is unlikely that heat from the CPU transferred through pins/tracks/pcb will ever ruin other components on the board.
I totally agree with you that heat is the biggest opportunity in a system the size of a NUC in terms of performance versus size, but at the same time one should not overestimate the problem.
Your lucky, with mine set to maximum and the os set to maximum I can hit 80 - 90 c no problem.. with only a Sql Server Dev, a couple of VS 2015 instances open (not overly large projects) and a few browser tabs IE or Google doesn't matter.
I put mine back on balanced in both the OS and Bios, fan is always spinning at or above 3000 and temps range between 70 and 85 at min.
The only way I can stay below 70's is to keep both on balanced.