Core Temps over 10C Difference?

[Zerohm]

I just built a new Core 2 Quad Q9400 rig and I think I'm seeing some pretty strange stuff.

Idle:
CPU 29C
Core 0: 39C
Core 1: 27C
Core 2: 38C
Core 3: 38C

Load:
CPU 41C
Core 0: 40C
Core 1: 34C
Core 2: 40C
Core 3: 40C

http://img410.imageshack.us/img410/7203/q9400burnin.png
The dark green is the CPU. (only 2 out of 5 sensors take a normal shape) Realtemp confirmed results.

So it appears that I have 3 cores that won't read below 38C or 39C, and one that is reading below the CPU temp, which I didn't think should happen.

Any thoughts? Is this worth an RMA?

BTW: (All stock speeds, mobo drivers and bios out of box)
Intel C2Q Q9400
Asus P5Q SE/R
4 GB Gskill 1066MHz DDR2
EVGA Geforce 9800 GT

 

[elmonen]

I don't think there's anything wrong..

I have two cores about 10C abowe the other two on my Q9400.. some people seem to have this..

Asus P5K
Q9400 @ 3,2 GHz

 

[BonzaiDuck]

We're not miffed that you posted your concerns about this -- it's just been reported over and over with sufficient discussion before.

The Wolfdale and Yorkfield cores are (frequently) known to have defective core sensors -- all, or in part. For instance, my Wolfie E8600 won't register idle temperatures below 51C, and the core sensors only begin register when temperatures rise above that level.

Your best bet is to limit your over-clocking and focus your thermal solution choices based on readings from the legacy TCase sensor. The thermal spec for these cores is defined in reference to that sensor, anyway.

Even if the core sensors register temperatures above some "stuck" value, I don't trust them. On my Wolfie, they actually show 1C-degree BELOW the TCase value, when they should be 5 to 10C higher.

Just off the top of my head, you probably shouldn't worry too much about the thermal issues if your over-clock settings report a CPU voltage in CPU-Z with an idle voltage below the upper limit of the Intel-spec safe-range, which should be 0.85V to 1.3625V.

Of course, enabling the LLC or similar feature may result in voltage behaviors that might suggest a slightly lower setting. Somebody else -- familiar with your chipset, would know better. Just a minute: . . . . mmm . . . yeah . . . . P45 chipset. I've been using nVidia for the last couple years.

EDIT: It's possible things have changed, but nobody here recommends using SpeedFan to monitor temperatures. You're better off using CoreTemp or RealTemp -- or a recent edition of Everest Ultimate (trial or licensed).

 

[Gillbot]

you can also check your layout and verify you don't have a weak mount somewhere:
http://www.itwissen.info/bilde...-qx6700-foto-intel.png
http://img165.imageshack.us/im.../core2quadro425ky3.jpg

Though it is usually more likely with a hard mount waterblock if you have one stud not tightened properly.

 

[sdsdv10]

As noted by a previous poster, check temps using RealTemp.
http://www.techpowerup.com/realtemp/

It has a "stuck" senor test. I don't know if you will be able to get a RMA for a single stuck senor, but I kind of doubt it.

 

[zagood]

Originally posted by: sdsdv10
I don't know if you will be able to get a RMA for a single stuck senor, but I kind of doubt it.

Intel freely admits that the temp sensors on 45nm chips aren't accurate below 50c but don't identify it as a defect. So no, they won't RMA.

-z