Quad core temps... how much variance should there be between cores? i have 7C difference

[PremiumG]

At full load my Quad core temps are around:
63C
68C
67C
60C

Should there be that much variance?? My heatsink, Arctic Cooler 7 Pro, is on correctly. I double checked the pins in the back of the motherboard. Thermal compound was applied as well.....

7C seems like a lot to me....

 

[aigomorla]

Originally posted by: PremiumG
At full load my Quad core temps are around:
63C
68C
67C
60C

Should there be that much variance?? My heatsink, Arctic Cooler 7 Pro, is on correctly. I double checked the pins in the back of the motherboard. Thermal compound was applied as well.....

7C seems like a lot to me....

im guessing your IHS either dips in the middle, or your heat sink has a dip in the middle. Another guess is that your push pins on the right side arent fastened securly. Another reason why i hate push pins. [If everything checks out, then yeah, your sink is fashioned properly]

But 7C is around normal for the average sink tho. You can only get even to each other if you can tune your mount like on a waterblock. [well i havent seen one for air yet]

 

[BonzaiDuck]

Aigo is probably right about this. He can correct me on my own observations.

7C range between high and low cores is probably to be expected.

Per his remarks about the IHS, many people report a narrowing of the range once they've lapped the CPU heat-spreader (IHS). Hardly worth mentioning that this will void your retail-box warranty, but you accept that anyway if you over-clock.

Several months ago, I posted a series of bar-charts -- bars in four colors for four cores -- showing a distribution of load temperatures sampled every 8 seconds by CoreTemp with ORTHOS or PRIME95 running over several hours, at various over-clock settings and measured room ambients.

After doing enough of that, I concluded that lower voltage and over-clock settings would narrow the range of core temperatures, and that higher voltage and higher over-clock settings would widen it.

For instance, at 3.0 Ghz, my Q6600 would show idle values with an average range of 3C degrees, and load values with an average range of 4 to 5C. Pushing the voltage higher and over-clocking to 3.2Ghz, the load range would widen to between 6 and 7C degrees.

I have yet to lap my heatspreader cap. Again, others say that this will reduce the range, and they mostly emphasize that it reduces processor temperatures -- TCase and the average core temperatures -- by several degrees (or at least a few.)

 

[BonzaiDuck]

PS and footnote:

there seems to be two explanations for the range of core temperature variations:

1) A slight unevenness in the fabrication of the cores and material beneath the heatspreader.

2) Core #0 -- possibly others -- are used more intensively. Without delving into the queueing of threads and data, I'd intuitively think that there are slight (yet) quantum differences in how the cores are utilized, and therefore an unevenness in utilization.

 

[PremiumG]

BonzaiDuck,

well, the cores seem to be all at 100%. I'm using Prime95 to test all the cores.

 

[Mucker]

I'll agree with the un-even base theory, but the pressure of the clamping matters too. I had a Ultra 120 Extreme on a Q6600 that sucked, as high as a 10C delta on all cores. I switched back to my Ultra 120 with the better clamping mechanism and now it's only a 4C spread. They really botched the clamp on the Extreme, what a pity and a high price mistake for me. Thermalright support was non-existent. The tech was the type of guy that let's say, enjoys talking down to people. A-hole.............

 

[Gerbil333]

I was about to ask the same question, though it makes sense for each core to operate at a different temp. The four cores in a Core2Quad are arranged in a line, right? Although the heat spreader will distribute heat between all four cores, I bet the outer two cores will run slightly cooler since they have an extra free side, whereas the inner two cores are surrounded on both their left and right sides.

I'm using a Thermalright XP-90 (LGA775 RM adapter) and Arctic Ceramique:

At near idle:

Core0: 37C
Core1: 38C
Core2: 33C
Core3: 37C


At full load:

Core0: 54
Core1: 55
Core2: 49
Core3: 51

 

[LOUISSSSS]

currently my idle temps are:

37, 32, 31, 34.

when i load them up for a few mins they all pretty much even out to within 1-2C of teach other. never more than 3C variance

 

[Duvie]

Originally posted by: Gerbil333
I was about to ask the same question, though it makes sense for each core to operate at a different temp. The four cores in a Core2Quad are arranged in a line, right? Although the heat spreader will distribute heat between all four cores, I bet the outer two cores will run slightly cooler since they have an extra free side, whereas the inner two cores are surrounded on both their left and right sides.

I'm using a Thermalright XP-90 (LGA775 RM adapter) and Arctic Ceramique:

At near idle:

Core0: 37C
Core1: 38C
Core2: 33C
Core3: 37C


At full load:

Core0: 54
Core1: 55
Core2: 49
Core3: 51

Nope....

They are 2 dual cores lined up side to side as seen in this....all cores have equal 2 free sides....

AS5 instructions

(open up pdf and go down 3-4 pages)

the only thing I have never figured out is core 0-1 one of the dual cores and core 2-3 the other...or some other combination of the 4....

right now my 4 cores under 2 instances of SMP folding at home is

core 0=58
core 1=55
core 2=54
core 3=58

seems like core 0-3 may be on the same core...

I noticed this same pattern in orthos prime as well...middle 2 always the lowest by 3-4c.

That has been seen with 2 times application of AS5 with my TR-120extrem e and 3 times with my coolit freezone coolers.

 

[Idontcare]

Originally posted by: PremiumG
At full load my Quad core temps are around:
63C
68C
67C
60C

Should there be that much variance?? My heatsink, Arctic Cooler 7 Pro, is on correctly. I double checked the pins in the back of the motherboard. Thermal compound was applied as well.....

7C seems like a lot to me....

I recently purchased the hardware necessary to build 5 identical Q6600 G0 systems, so I had the luxury of testing out a few prevaling theories during the builds over holidays.

Like a lot of the advice you read in this thread, I walked into my builds expecting to accept anything upwards to 6-8°C temperature delta between my cores when fully loaded.

Sure enough my fourth system had a CPU which when clocked to 3.3GHz (9x367) there was a temp delta of 8°C between the cores. Sigh.

All I did was pull the HSF (Tuniq120), cleaned off the thermal grease (AS5) both the HSF and the IHS, reapplied the AS5 and reattached my HSF.

Powered-up and back into prime95 small FFT. Temp delta dropped to a mere 2°C.

Now obviously my temp delta was created by either a poor application of my thermal grease (too much, too little, entrained air pockets, etc) or the HSF wasn't tightened down uniformily (got to torque it down evenly).

Not everyone is limited by this one culprit. Temp delta can be poor contact for a number of reasons, and if it is poor contact between the IHS and the chips themselves then there is very little you can do.

But the message here is do not simply "accept" it because enough people drill it into your head that it is something you must live with. (unless it really doesn't bother you)

Go ahead and play around with the HSF and the thermal grease, and don't be too surprised if seemingly nothing more than just "redoing it" makes a world of difference as it did in my case.