Q6600 B3 temps seem off

Hey guys,

I recently checked out my temps for my Q6600, and found them to be high - B3 stepping, not OCed, running with a Thermaltake V1 heatsink. Coretemp reports my temps at:

Core 0: 52C
Core 1: 49C
Core 2: 44C
Core 3: 44C


This is with multitasking, but not stressing the CPU, in a room that isn't hot. These temps seemed a bit off, so I felt the heatpipes on my V1, and the entire heatsink feels warm - not hot. Could this be a bad IHS...bad sensors? It seems hot, and I know the B3s run hotter, but this just seems abnormal.

Oh, and a eVGA 680i SLI Motherboard, Antec 900 case ( have the side off now, no real change. Hell, I have the AC on blowing in the general direction of the intake fans on my Antec 900, and STILL no real change.)

Thanks,
RA

run prime95 and see the load temps, but possibly in the realm of normal.

Running two instances of Orthos:

Core 0: 71C
Core 1: 67C
Core 2: 61C
Core 3: 59C

Isn't that...MORE than the max the B3 stepping can handle!? My CPU hasn't died...and the mobo has not shut down my system automatically. Could these readings just be wrong?

What the hell? I'm wondering if I need to pull my CPU and re-apply some AS5?

Try using Realtemp, that should detect the proper Tjunction for the B3 (85ºC). So your temps would be 15º less which should be about right.

Reseated the CPU, reapplied AS5, temps dropped 5C according to coretemp (without burn in, of course...)

Realtemp says my cores are:

50
47
44
44

Under full load with Orthos. Is this normal (its certainly better than I thought....)

Both Coretemp and Realtemp agree on the distance to Tjunction max, but the temp differences are amazing

Readings taken at the same time (and with each core loaded 100%)
Coretemp:
63
62
58
57

Realtemp
48
47
43
43

Has anyone done a real comparison and looked into which is "right"?

OK (here we go again)

The only "right" or "correct" value is the "distance to Tjunction max". This value is read directly from the CPU registers and is the ONLY hard data the CPU provides. Everything else, in this case a reported "real temperature" is calculated and how that calculation is done is using guessed at values because Intel does not publicly publish some needed data and guesses have been made.

At the factory each CPU is calibrated during testing and a two temperature limits are established. Exactly how this is done is not public. There is:

1) Tjmax which is not really a max, see item 2. This is the temp that when the internal Digital Temp Sensors reach some unknown value the processor will put out a PROCHOT signal which is intended by Intel for motherboard manufacturers etc to use to spin the fans up to 100% speed. I think of this as a signal from the CPU screaming " I am HOT, spin up the fans because if I get much hotter I am outta here ! "

2) A Catastrophic Thermal Trip point. Not clearly documented but some Intel literature suggests an internal temp 10-15C ABOVE/hotter than the Tjmax value. If spinning up the fans does not reduce the CPU temp and this temp trip point is reached the CPU will shut itself down to save itself.

Now the ONLY data the CPU provides to the outside world is the value of the "distance to Tjmax. This is like a countdown timer. At distance to Tjmax = 0 PROCHOT is send out. So higher is better, and tells you in the number of Deg C how much headroom you have before the CPU thinks it is hot. In theory you could/should be able to run at Delta to Tjmax = 1 24hours x 7days a week x 3 years (warranty period) without issues and if it failed Intel would replace it. Of course we don't do that and I recommend at least 15C margin under heavy (stress testing) load. And the more the better.

OK so if the ONLY thing the CPU tells us is the distance to TJmax, how do we get a "real" temperature. Well, if we knew what Tjmax was it would be easy:

Tjmax (some temp in Deg C) - Delta to Tjmax (from the CPU) = actual CPU core temp.

If we knew that Tjmax (and we don't) was equal to 105C and we had a delta to Tjmax of 35 then the actual core temp would be 105C - 35C = 80C.

Its the "if we knew" bit that is the issue. No one in the public knows, and since it varies per CPU it would not matter a whole lot anyway but I guess its better than guessing, what Tjmax is. So the programmers have to guess and plug in some value to the math above. As more experimentation by enthusiasts has been done we have a better guess at what Tjmax is for a family of CPUs. Thats why it is important to use the newest version of the program, which ever one you like, if you swap CPUs or get a newer CPU. The programmers have to guess, and in general they are good guesses but sometimes they guess wrong.

Note: There is some early documentation from Intel on the mobile CPUs that started the 105C Tjmax value. And it appeared correct for a lot of early C2Ds. Things have changed a lot in the last 2 years and Intel has not been forthcoming in publishing any internal core temp values. Tcase is NOT A CORE TEMP and the Intel web page with processor specifications used to have a typo in the temp column that lead some people to wrong conclusions/misunderstandings.


All of the above can be found in the data sheet for your specific CPU and the Thermal and Mechanical Guidelines document for your processor family from the Intel web site.

/start mini-rant not really directed at OP.
So there is no "right" there is only Delta to Tjmax. Use Realtemp and go in and put in whatever adjustment to Tjmax results in the program spitting out whatever "real temp" makes you happy or feel good. If you are concerned about your actual cpu temp then look at the actual delta to Tjmax and decide how much thermal margin you want under load. Who the fhell cares what idle temps are. By definition at idle your computer is just a big paperweight anyway. Idle temp bragging is for noobs.
/end rant

I had done some reading on Realtemp before I posted asking what was right - and I do understand how the system works, and that distance to Tjmax is the only absolute. I was hoping someone had gone through the pains on a Q6600 or similar to see what the actual Tjmax value was. So, my own fault for not posting a better question.

I didn't post idle temps, because as you said, they are meaningless to me. I'll just use distance to Tjmax from now on.

Thanks!