I just got a NUC 5i5RYH and I kind of love it. ;-)
I can't find out the sweet spot for the cooling thou.
Any suggestions as for what temps I should aim for?
Let me give you some background on the solution that you have and how you should use the parameters...
First of all, let's talk about the processor. The processor has three important temperature points, Tcontrol, Tjmax and Ttrip.
These three temperature points can vary from one individual processor to another. In most* processors, the Tcontrol and Tjmax temperatures are accessible from hardware FSC solutions and from software (like the BIOS, which will implement the configuration of the FSC solution).Now, these temperature points are relative to a temperature reading that I call the package temperature. The processor takes readings from a number of digital thermal sensors (DTS) spread out across the processor die (there's one at each core's major hotspot, for example) and calculates, using a secret algorithm, a package temperature reading from them. This package temperature is then provided to external FSC solutions over the Platform Environmental Control Interface (PECI) bus.
(* - In the exception processors, these two temperatures are provided in the thermal specifications for the processor)
I do not recommend that you configure the fan to turn off below some temperature. When the fan is stopped, the processor only has the thermal mass of the heatsink to dissipate heat. In the NUCs, this mass is fairly small and you could end up seeing the fan fairly quickly cycling off and on (depending upon activity, of course). At the same time, this fan is also generating the airflow that is dissipating heat from all of the other components in the NUC. In the NUC design, air flows into the chassis through the bottom of the chassis (the opposite end of the chassis from the processor). If flows over the components on the opposite side of the board from the processor, around the edges of the board to the processor side and is then exhausted through the processor heatsink. Stopping the fan can cause component overheating. We don't have temperature sensors everywhere, so it is possible that, if the fan off temperature is determined only based upon the processor temperature, things could be overheating - and getting damaged - without the FSC solution even knowing it. Remember that the busyness and temperature of the processor can be independent of the busyness and temperature of other components; they can be busy when the processor, relatively speaking, is not.
Next, let's talk about the FSC solution itself. The NUCs rely on the FSC solutions that are built into the Super I/O (SIO) ICs that are utilized. These SIOs have varying FSC capabilities. The minimum subset of these capabilities that we can rely upon allows the speed of the processor's fan to be controlled by two temperature inputs, the processor temperature and one other. Within each NUC, we place thermal diodes (or thermistors) at a number of strategic points that we consider to be most likely the hottest. This typically means a diode is placed under the MSATA/M.2 device and a diode is placed under the memory DIMMs. Only one of these diodes can be used; you need to decide which one based upon the temperatures that you see. In most cases, we found this to be the one under the MSATA/M.2 and this this is picked as the default. The ability to control the processor fan with two temperature inputs is done using a "Hottest Of" algorithm. For each temperature, you define the independent fan speed control curve that is necessary. At any point in time, the necessary fan duty cycle will be calculated for each temperature based upon the control curve defined. Then, whichever of these duty cycle values is higher is fed to the actual fan speed controller for implementation.
Aside: many folks run applications - like SpeedFan or HWMonitor or AIDA64 (my preferred) - that can display a *lot* of temperatures. In addition to those temperatures monitored by the SIO, this can include temperatures from a lot of possible places -- the MSATA/M.2 SSD, the HDD/SSHD, each processor core, each memory DIMM (though this capability is rare in SODIMMs), etc. and etc. It should be understood that the hardware FSC solution does not have access to all of these temperatures; some of these readings can only be extracted through software (for example, via S.M.A.R.T. for SSDs/HDDs/SSHDs). The SIO can input temperatures from the processor, which are obtained across the PECI bus, and from a number of internal sensors, which are measuring temperatures at the thermal diodes (or thermistors) that were placed on the board. That's it. These are all that can be used for fan speed control.
OK, that's everything you need. If you want to tune the configuration, run tests that stress the processor, memory, disks (SSD and/or HHD/SSHD), etc. and use the resulting temperature information to determine which temperature input is best coupled with the processor temperature and what fan speed control curves are necessary for each. Simple, eh? (Oops, there's my Canadian roots slipping through ). Clear as mud?
In order to keep down the size of this response (and the amount of my time necessary to do so), I have been rather terse in some areas and abstract in my explanations for others. I hope this helps with your decision-making. Let me know if you have any questions or want more clarification on something.