CPU Temps - Is Anything "Normal?"

[whovous]

I am trying to decide if my cooling is good enough, or if I still have work to do.

This new build is in its early stages, and I don't have a lot of data to work with, but I can report the following:

1. I installed the stock cooler using only the provided thermal interface pad. Idle was 39C. I ran OCCT and after an hour had peak temps of 59C on Core 1 and 54C on Core 2. All results obtained using Core Temp 0.99.5.

2. I replaced the stock cooler with a Zalman 9700. I used Arctic Cooling MX-2 as the interface. Idle remained at 39C, but the OCCT numbers dropped to 51C and 46C, respectively, an eight degree improvement.

A few things confuse me here:

A. Is the 39C idle temp too high? Ambient room temps are in the low 20s.

B. Is it normal to get an eight degree improvement under load but no improvement at all at idle?

C. Is there any such thing as a "normal" temperature, or are temps likely to vary between otherwise identical systems due to the quality of the temp sensor on the board? Is there a good way to install an external temperature monitor without putting something between the CPU and the cooler?

D. Finally, what is up with the five degree difference in peak temps between the two cores? I get the same difference with the stock and Zalman coolers.

[Chike]

Your temps looks fine to me, at the same clock and 29c ambiant I get 57c.
Idle temperatures are not important, any core temp under 50c is considered under 50c the DTS sensore are most inaccurate at low temperature, and sometimes get stuck at a certain temperature they would not go below.
The DTS is not very accurate at higher temperatures either +-5c, to get more accurate and close to the actual readout you will need to calibrate them.

[whovous]

Thanks for the reply. How do I calibrate them?

[Acebmxer]

its a long read but tells you everything you need to know.techpowerup.com/realtemp/docs.php

[Chike]

Or the method described in Sticky: Core i7 and Core 2 Temperature Guide

[whovous]

Wow. I've long known that temperature readouts could be wildly unreliable, but I had no idea the problem was curable. It looks like my sensors are stuck and will not drop below 39C, but that with enough work I can get something resembling real numbers from them after all.

Thanks for some highly informative replies.

[Psycho101]

If your sensors are stuck at 39c then idle temperature calibration (as detailed in the Toms Hardware link) will be impossible.

[whovous]

Ah, heck. I went through the challenge of reading the whole thing on a cell phone on the subway yesterday, too. I had the impression that the method on techpowerup stuck sensors into account and offered a SpeedFan workaround. Then again, reading something that complex on that small of a screen is not exactly the best way to read with comprehension.

I am not going to give up just yet, even though I fear you are probably right.

[Chike]

Originally Posted by Psycho101

If your sensors are stuck at 39c then idle temperature calibration (as detailed in the Toms Hardware link) will be impossible.

The idle calibration is done against the diode(mistakenly refered as Tcase) stuck DTSs are not a problem.

[Psycho101]

The tech powerup guide involves setting your idle temps via offsets to be what is worked out from your ambient temp (room) plus x degrees c for the method of cooling. Therefore if you can not read true idle temps (due to them being stuck) the method is useless, sadly.

You can calibrate them at load using the thermal diode that is also on chip though. That method isn't detailed in the article. Here is some useful info for load calibration amongst other things: Sticky: Core i7 and Core 2 Temperature Guide . I strongly recommend that if it is difficult to comprehend (it's not exactly light reading) then you probably don't have a crucial need to calibrate sensors. It will still only give you more accurate load temps. Your idle readings are permanently screwed with stuck sensors.

ChiKe, Idle temps have zero to do with the diode. Maybe a re-read of the article is needed?

COOLING...............................IDLE DEGREES ABOVE AMBIENT
High end water...............................6C above ambient
High end air (true push/pull) ........6-7C
High end air (1fan).........................7C
Mid air (zalman 9500)....................8-9C depending on fan rpm
Intel stock cooler...........................10-11C

If your reported temperature is too low then you will need to go into the RealTemp Settings window and use a positive calibration factor to increase your reported temperature. If your CPU is reading too high, you will need to use a negative calibration factor. On a multi-core processor during this test, you should have equal temperatures for all of your cores. Though this calibration test is done when your processor is idle, the calibration settings entered will improve the accuracy of your reported tempertures from idle to TJMax. Return your computer to your normal MHz and core voltage settings and if you need to, you can make some further minor adjustments to equalize your core temperatures at idle. You shouldn't need to make any major adjustments at this point.

This clearly states you use the offsets in realtemp (RT doesn't even read the diode) to adjust DTS readings using ambient and the above table. Notice the wording of temperatures, and the fact that the only two variables used are ambient and the figure in the table. In no way, shape or form does the diode reading come in to this guide.

This means that if your ambient is 20c and you're on high end water, it's assumed the current idle temp at the FSB and voltage and multi detailed IS 26. No diode mentioned any where there.... You then procede to adjust the offset with a negative value if cores are above 26c and a positive if below, until you see, for eg a quad "26 - 26 - 26 - 26". If the sensors are stuck, this is pointless.

Again, for clarity, at no point in the article or the detailed calibration method do you even need to know that the diode exists, it's not an issue.