I really don't understand why folks are so mystified when they see this result. You have a 15W processor (28W for Core i7 version) whose heat needs to be dissipated and this requires the efficient transfer of heat from the processor and chipset to the chassis, enough mass in the chassis to buffer this heat and enough surface area to dissipate this heat at a reasonable rate. Achieving this while running in more-normal circumstance should not be an issue. When you are throwing something like Prime95 at it - and do so for an extended period of time - however, you can easily overcome the thermal mass and dissipation capability available. All I can say is, don't do it - not unless you have an expectation that this is exactly the kind of results you are going to see.
While I am at it, let's talk about some related issues. The NUC chassis are carefully designed so that the heatsink-blower unit will also create airflow over both surfaces of the board, thereby providing the cooling necessary for other components in the design. Despite this capability, it still proved necessary to provide thermal foam to dissipate heat generated by the M.2 SSD through the chassis. Something to also note is that, at the same time that the processor SOC is dissipating heat into its heat spreader, it is also dissipating some heat in the opposite direction and this heat is travelling through the board and can affect the temperatures - and cooling requirements - of components on the other side of the board. The same goes for some of the voltage regulation components. Bottom line, airflow over the opposite side of the board is actually necessary.
I would hope that the designers of the fanless chassis would take these factors into consideration and carefully test for cooling remaining adequate. In the absence of any feedback, I often wonder whether they even think about these factors at all. In the case of this Akasa chassis, I did see that there does not seem to be any way for the heat from the M.2 SSD to be dissipated directly into the chassis (well, other than by radiation).
Oops, missed your questions at the bottom...
- The AY and BN NUCs utilize a brand new Embedded Controller (EC) device (which, BTW, replaces the legacy SIO device). Plain and simple, none of the existing Monitoring and Fan Speed Control applications (things like AIDA64, SpeedFan, HWMONITOR, HWiNFO32/64, etc. on Windows and the LM-Sensors stack on Linux) have support developed for this EC device as yet. An additional complication is that, to also support the operation of the Ring and Power Button LEDs, custom firmware needed to be developed for this EC. This may mean that it will take longer for this support to be developed than otherwise. I am looking into this issue and working on some sample code that will demonstrate how to extract this information for display (and also how to modify the fan speed control programming at runtime). Once I have this working, the application developers will be able to use it to develop their own support. Stay tuned...
- Off the top of my head, I am not sure of the answer to this question. I will need to do some research. Others can chime in on this as well...
The thermal capacity problem does exist but it's irrelevant in this specific case. If the CPU was running at its lowest frequency say 500 MHz or something like that for the sake of argument, and the system still over heats, then you can claim that the thermal removal capacity of the fanless system is insufficient for this CPU. But this is definitely not the case. As I said the system sits quite happily at 2800 MHz, in fact it had been doing so for over 12 hours without any problem. All is needed is a new thermal throttling program in the BIOS that can correctly handle the fanless characteristics. When I run Prime95 for extended period of time the BIOS should automatically slow the CPU down to 2800 MHz or below instead of letting it over heat then finally force an OS shut down.
Speaking specifically of the Akasa case, are you implying that all previous versions of NUCs had the same problem of over heating but no one had been making any noise? There was a guy posted a message here a while ago just to tell people he’s very happy with his Akasa case. He mentioned specifically he had been running Prime95 for over 2 hours. He also mentioned that there was thermal throttling after extended period of time which is the expected behaviour.
It’s interesting to note that when Akasa first released Plato X7 earlier this year it’s claimed that the 28W NUC7i7 is supported just like the previous generations. But Akasa very quickly dropped the support for NUC7i7. Based on this and the uneventful history I would rather suppose the problem has something to do with the new Kaby Lake CPU? Or let’s put it this way, the new Plato X7 case is just as heavy as the old ones and they didn’t replace aluminium with plastics, so what can possibly go wrong?
Ok, your response was good for a laugh if nothing else. I've stopped and climbed back into my chair and now must figure out how to explain this to you without having to spend hours doing so.
Yes, your system is sitting quite happily when not much is going on, but it is failing when running Prime95. You seem to want to blame the BIOS for this. The BIOS has very little to do with it. The thermal throttling capability is implemented in the processor and the BIOS has absolutely no control over its operation. If the processor temperature reaches its maximum junction temperature (Tjmax, which is somewhere around 100c, but there could be a processor-specific variance of a few degrees plus or minus), it automatically begins throttling its performance. The BIOS is not involved. If the processor temperature continues to rise above Tjmax, however, it will eventually reach the Thermal Trip point and the processor will automatically shut itself off to protect itself. This all happens automatically; the BIOS is not involved. There is no such thing as a "thermal throttling program" in the BIOS; never has been one; never will be one. This goes for all board manufacturers.
So, let's presume for a moment that the Akasa case can handle your processor. If this guy you mentioned has the same case and the same NUC board that you have and he can run Prime95 for two hours without issue and you can't run it for even 10 minutes, then one of two things is going on: (1) he has his processor performance configured correctly (dialed down from maximum to balanced in both the BIOS configuration and in the Windows Power Settings applet) but you do not, or (2) he has his board properly installed in the chassis and you do not. Actually, I should say one of three things, since both could be true.
Yes, I am saying that I believe that every one of the fanless designs is likely poor. That's a pretty harsh message, I know, but the shoe fits. Handling the processor is not the issue (well, it is in your case, but shouldn't be in general); you purchase the version of the chassis that supports your particular board and your particular processor and you should be good to go (well, unless you install it incorrectly and the processor's heat cannot be dissipated properly). No, what I am referring to are the other components on the board or connected to it (i.e. your M.2 SSD, your SODIMM(s) and your extra 2.5" HDD/SSD/SSHD if you indeed added one). In most designs that I have seen, these other components are simply roasting; they do not have any path for their heat to be properly dissipated. They do not have the airflow that is required to do so. Now, this doesn't mean that they are not going to work. Oh no, it's much more subtle than that. What it means is that they are not going to last anywhere near as long as they could have if their temperatures are maintained at lower levels.
Think about this: In the NUC design, even with the (what you call unacceptably loud) heatsink and blower installed, the M.2 SSD still needed to have a thermal foam strip included to help dissipate its heat into and through the chassis. In the pictures in this Akasa chassis' setup brochure, I do not see any thermal foam for the M.2 SSD. Please tell me I am wrong and it's there. If I am right, it means that the SSD's heat dissipation is only through the (almost negligible) air movement inside the chassis. If this is the case, I wouldn't be surprised to also find that the M.2 SSD is spending some portion of its time throttling its throughput in order to stay cool enough to continue functioning.
I don't want to waste my time talking to a charlatan...
1. The system is stable below 2800 MHz running Prime95 for more than 12 hours.
2. The Akasa case came with thermal block to attached to the M2 SSD.
A more serious point, this is an intel support forum not some sort of discussion board where people post their ramblings on things they are not competent to comment on, I expect expert advice not "opinions" to put quite politely.
First, I have to apologize; I completely misread your previous posting and thought you were saying this other guy had it working but you didn't...
Second, as an administrator for this forum, I have to admonish you for your disrespectful response. Don't do it again. Here are the Intel Communities House Rules that you are required to follow.
Re: "The system is stable below 2800 MHz running Prime95 for more than 12 hours."
>>> Good! You have a solution: Just stick with this configuration. I know that this means leaving performance on the floor, but there is simply no capability in the BIOS to do any kind of dynamic management of processor performance based upon temperature. If you would like to open a ticket formally requesting this as a feature, you need to do so through the alternate methods (forms, chat, phone or email) for contacting Intel Customer Support; you do not do so here in this user-to-user support community. If you need help with this, consult the appropriate page:
Re: "The Akasa case came with thermal block to attached to the M2 SSD."
>>> Good! As I said, the picture in their setup guide did not show this at all. Still, there are other components that do not get properly addressed (and this is fact, not opinion).
You do not crave for respect by posting incorrect and irrelevant information.
And let me be crystal clear I don't give a damn about rules on this forum, alright?
A quick note on your "there's no program controlling the CPU frequency, temperature" argument.
Well sure if you insist that a blower is not a fan you can also say CPU microcode is not a program.
You can also say there's not BIOS at all it's UEFI nowadays.
The fact is whatever is juggling with CPU thermal performance is not good enough.
And it can be fixed and it will be in the form of a new BIOS be it CPU microcode or BIOS fix.
BTW in the worst case the system just freeze without shutting down. The second worst is when the CPU shuts down the system immediately. In that case you get a screen black out. In my case it's the OS that's shutting the system down. The entire thermal throttling business is more sophisticated than T junction and T max.
I repeat, there is no mechanism in this board's BIOS to manage processor performance based upon temperature. I don't disagree that such a feature might be useful, however. As I said before, if you would like to submit a feature request, feel free to directly contact Intel Customer Support and open a ticket with them.
I am terminating this conversation.