I am using Intel Desktop Utilities (IDU) 3.0 to evaluate temperatures. At normal (read low/no) load, the CPU is at 26C and the VR (voltage regulator) is at 47C. Why the big difference between them if the VR is integrated into the CPU? (It is integrated, isn't it?) IDU gives an upper threshold of 64C for the VR, but I have a feeling that is for boards where the VR is discrete from the CPU (?). IDU gives 97C as the threshold for the CPU. Is my reasoning correct and should I just up the VR threshold to the CPU threshold? The difference at idle still bothers me. What CPU and VR temperatures are others seeing? I am using the stock Intel heatsink/fan on a DQ57TM motherboard.
The voltage regulator is not inside the processor, it is next to it on the motherboard so that the fan Heat Sink can cool it down a bit. The temperatures that you are reaching are normal and you should not worry. If you want to lower your voltage regulator temperature you can increase your chassis airflow. (one fan at the front pulling cool air inside and another one in the back pushing it out) Make sure that there are no cables near the processor.
BigThor wrote: "The voltage regulator is not inside the processor, it is next to it on the motherboard so that the fan Heat Sink can cool it down a bit. The temperatures that you are reaching are normal and you should not worry. If you want to lower your voltage regulator temperature you can increase your chassis airflow. (one fan at the front pulling cool air inside and another one in the back pushing it out) Make sure that there are no cables near the processor."
Someone on one of these Intel boards said it was integral to the CPU. That they are only a few degrees apart under load is uncanny, IMO, if the VR is on the MB, separate from the CPU. I'm still wondering what limits I should set in IDU and why the limit was 64C as installed. (Aside, I have a 120mm rear fan and 92mm front fan, but the motherboard controls their speed (they are PWM), but I'm all for quiet if the temps are acceptable. Peculiarly, the fans never change speed at full load and "high" temps and I'm wondering if the CPU is throttling back rather than increasing the fan speeds because the temps do level off).
The VR voltage level decision-making is indeed happening within the processor, but the actual voltage generation (for lack of a better word) is happening in circuitry on the motherboard (a diode sensor in the vicinity of these components is measuring their temperature). None of these temperatures you've mentioned seem high enough to be concerning. There is no processor throttling going on; the processor cores would need to reach their Tjmax temperature (somewhere in the vicinity of 100C, through it varies from one individual processor to another) before the TCC would kick in and throttling take place. The temperature is stabilizing without you hearing any major jumps in fan speed because Intel(R) Quiet System Technology is designed to minimize acoustics while still achieving the necessary cooling...
A common reason for the VR temperature to be higher is because of the use of a cooling solution that does not push air down onto the motherboard to (actively) cool these components. In these cases, unless there is an issue with airflow (through bad cable routing, etc.), the VR components will still be sufficiently cooled by normal chassis airflow; they will just be at higher temperatures than the processor itself. Even if you have a solution (like the boxed CPU heatsink-fan unit) that blows air downwards, it is still possible for the VR temperature to be higher than the processor, so don't go thinking that there is something necessarily wrong in this case...
The threshold for generating an alert for this sensor can be changed within IDU. You could adjust it higher if you wish -- but don't go so high that the alert capability is useless. I see that, in some of the newer boards, an alert level of 80 has been used for this sensor. This seems like a reasonable level to me, but what I would do is adjust it up a little bit at a time...
I was assuming to some degree that IDU defaults were "reasonable", but after the VR temps triggering an alarm because the default was 64C I feel "a bit light" on the knowledge I'm supposed to have to set up these boards or that the defaults shouldn't be there to begin with so it would require the integrator to think about what the alarm thresholds should be. It would seem to me that there should be a setting for an integrator to set according to what motherboard was being used. At least that would be a good starting point and maybe something used by 80% of integrators, as opposed to having the CPU default at 97C and the VR default at 64C. Is the VR so close to the CPU that the CPU is dumping heat into the VR or does the VR generate it's own heat?
If you are using the boxed CPU cooling solution and you have a chassis that is compliant with the Thermally Advantaged Chassis specification and you have all fans installed per this specification and you have not created any airflow problems with cables, etc., then these settings are perfectly fine and reasonable - and have been tested as being so. If you vary from this - using a poorer cooling solution (one that doesn't direct airflow downwards, for example) or not following the recommendatons for #fans, airflow, etc., then you are going to see variances...
Well, the alert threshold for the processor is defaulting to (in your particular case) 97 because this is where your particular processor told IDU that the TCC (Thermal Control Circuit) was going to begin asserting and throttling performance if the active cooling solution is not providing adequate cooling (which could be for any of the reasons stated above and many, many more (poor thermal compound, poor attachment, etc. and etc.)).
Yes, as air travels through the fan and heatsink, the heat from the processor is being absorbed by this air, but the speed of this air is selected to ensure that there is adequate dissipation for the VR components as well (which *do* generate significant heat).
Ah, so the defaults come from the CPU! OK on that.OK on the curiosity I had about the VR generating its own heat and the CPU heat dumping. I am still bothered by the VR temp default alarm threshold and I'll explain why. First, the system I am evaluating:
Intel components: TQ57TM, core i3 530, stock Intel fan/heatsink
Case: Mini-tower, CAG 1.1 side air duct, add-in card side vent, 120 mm PWM rear fan, 92 mm PWM front fan, no cabling or other airflow hinderances (the case is mostly empty).
Rear Fan Specs: 53 CFM @ 1200 RPM (QST runs this at idle at 400 RPM)
Front Fan Specs: 56 CFM @ 2500 RPM (QST runs this at idle at 470 RPM)
CPU Fan: QST runs this at 1120 RPM
Using the fan/temp monitoring in the BIOS, the CPU temp is 49C and the VR is 60C after a quick look. That only gives 4 more degrees before the alert occurs. And I am testing in a room that is below 70F ambient. If this was mid summer and the ambient was, say, 90F, well, that 64C triggering on the VR seems like too low of a threshold.
I agree. According to the Intel(R) QST configuration, the control temperature for this sensor is much higher than this health threshold. I would presume that, when they established the setting for the one, they missed doing so for the other (which should correspond with it). I reported this issue to the boards team and requested that this be fixed...
OK. I feel comfy upping the threshold now (though, of course, out-of-the-box settings should be useful or be delivered with appropriate guidelines/warnings/engineering design info). It surely won't be my final choice of thresholds, but at least it will allow me to deliver PCs without losing sleep wondering about whether there is a fire hazard out there or the client being bothered unnecessarilly by false alarms. Since the CPU is "way up there/hot" at 97C, I'm considering 85C for the VR on the high end and 80C if I decide to be conservative. Probably the latter since the board doesn't perform control against this parameter like it does for CPU temp (is that correct?).
i860 shortly after booting in a 3U rack case with 2 fans at the back and vents galore at the front. 250W PSU (which is plenty, don't argue.)
ambient room temp 27'c.
PCH 59, Mem 41, chassis fans about 2000rpm,
VR 71 processor 40 (952rpm) after a while it went down to about 59.
starting Burn In Test (v6 pro)
3 minutes, cpu 72, vr 79,
5 minutes, cpu 82 (fan now 1020), vr 90! case fans still 2000rpm <-- maybe I should get 4pin fans?? Mem 44
10 minutes, cpu 80 (fan now 1171), vr 90, mem 45
(at this point it appears stable. fan exhaust is 35'c)
Let me clear one confusion. In a 4-wire fan, the raw, pulse-width modulated (PWM) output signal - which specifies, via duty cycle, the desired fan speed (as a percentage) - is delivered to the fan and the fan determines how to convert this into a particular (RPM) fan speed. In the 3-wire fan, however, there is no capability (no wire; the 3 available are used for power, ground and tach signal) to deliver the PWM signal to the fan. Instead, a circuit on the motherboard is responsible for making the conversion. There are typically two ways that this is implemented. One is the voltage variance control methodology, which scales down the voltage delivered to the fan as the duty cycle drops below 100%. The other is the pulsed power control methodology, which sends power (at a full 12V) to the fan in pulses. What is the performance difference between 3-wire and 4-wire fans? Well, there might be small differences in how the 4-wire fan and the 3-wire fan's control circuit (on the motherboard) respond to PWM signals, but that is usually compensated for by the fan controller's control based upon temperature. A 3-wire fan and a 4-wire fan, if the same size and same blade pitch, will be capable of generating the same airflow and thus the same cooling capacity. Thus, for a case like yours, you would look to use fans that generate higher airflow (through being physically larger, having different blade pitch, having higher maximum speed, etc. Whether it is 3-wire or 4-wire is typically not a factor - except when it comes to the motherboard's capabilites (i.e. whether providing support for 3-wire, 4-wire or both)...