Whats it really mean to have a high delta between cores?

[LOUISSSSS]

does it merely mean that your heatsink is tighter/better fitted on one core than the rest?

 

[aigomorla]

correct.

It means a ton of things... uneven pressure on mounts. Unflat IHS or Sink.

less work being done by the other core. A ton of things.

now if your running prime and its uneven then my top line applies.

 

[Thor86]

Or, it could mean poorer contact from Core to IHS on one core.

Originally posted by: aigomorla
correct.

It means a ton of things... uneven pressure on mounts. Unflat IHS or Sink.

less work being done by the other core. A ton of things.

now if your running prime and its uneven then my top line applies.

 

[SanDiegoPC]

I'm running two instances of Prime right now. One with large FFT and one with small FFT. In addition of course to my normal web browsing.

I see 65* on cores 2 and 3, yet 57* on the other two. Is that the uneven pressure mentioned? I'm running stock HSF and thermal paste as from Intel.

 

[initialised]

Slightly off topic:

During Orthos runs on my main rig (E4500 @ 3.36, 420x8, Swiftech NCW6000, Maximus Formula) SpeedFan reports CPU 10-15C lower than the cores but on a system I recently built (E2140 @ 2.72GHz, 340x8, Stock HSF, GA-G31MX S2) CPU reads 5C higher than the cores any idea why?

 

[LOUISSSSS]

Originally posted by: initialised
Slightly off topic:

During Orthos runs on my main rig (E4500 @ 3.36, 420x8, Swiftech NCW6000, Maximus Formula) SpeedFan reports CPU 10-15C lower than the cores but on a system I recently built (E2140 @ 2.72GHz, 340x8, Stock HSF, GA-G31MX S2) CPU reads 5C higher than the cores any idea why?

u need to set and offset for all of your CPU core temps of +15 on both of ur chip's cores...

 

[LOUISSSSS]

but in the long run, assuming the highest core is still within the CPU's spec limits, there will be no real problem correct?

 

[Idontcare]

Louisssss there are two reasons you will observe the temperature delta.

The first is the one highlighted by posters already...namely inefficiencies in thermal transfer to get the heat out of the chips under the IHS.

As this involves inefficiencies in thermal transfer, anything affecting the uniformity of thermal transfer will also effect temperature deltas at the cores. These uniformity issues can be rooted in non-uniform thermal paste thickness, IHS coverage, IHS flatness, HSF flatness, etc, etc.

The second reason though has to do with the heat generation itself. The most uniform thermal transfer system in the world can't change the temperature delta if one core (or chip for a quad) pumps out 2X the heat as the neighboring core or chip under the same IHS.

In this regard a temperature delta can actually be an indication of the efficiency of the chip/core in question. If a chip creates more heat output than a neighboring chip then that chip will operate at a higher temperature than the neighboring chip.

For example I have a G0 Q6600 that ran at a ridiculous 8C temp delta between the two conroe chips when fully loaded with small FFT. Even at idle there was a 5C temp delta. This was with a Tuniq120.

I tried AS5, TX-1, TX-2 and ceramique. Multiple times each. All to no effect on the temp delta. I tried another Tuniq, another motherboard. No effect on the delta. I even lapped the IHS and the HSF...still same 8C delta at full load.

My conclusion for this chip - thermal transfer inefficiencies are not the cause of the temperature delta, one chip in my quad simply converts more power into waste heat than the other.

Other quads I have with temp deltas were clearly thermal transfer limited as lapping or merely reseating the HSF eliminated all but 1 or 2C of the delta.

 

[LOUISSSSS]

I think I have a similar but opposite case. At idle there is up to 8c delta. At prime95 load there is a 0-2c delta

 

[Idontcare]

Originally posted by: LOUISSSSS
I think I have a similar but opposite case. At idle there is up to 8c delta. At prime95 load there is a 0-2c delta

To me this suggests that your temp delta is due to thermal transfer inefficiency that could very well be due to coefficient of thermal expansion mismatch.

As the IHS and HSF heat up under CPU load they will expand and make an already tight fit even tighter.

After expanding and squeezing that thermal paste ever more so and you then let the chip idle and the IHS and HSF cool down and contract ever so slightly as well, shrinking in different way for each. If there was going to be an air gap form, it would be then. And that air gap could be temporary too, disappearing once the HSF and IHS heat back up again.

Just my suspicion, being an engineer I am of course biased to theorize about thermal mismatch and the such, could be complete BS in reality as far as explaining your observation.

 

[tshannon92]

I had a 1c delta with Stock Intel HSF, after Lapping CPU it rose to 3c. I then lapped AF7 and CPU and now I usually have no delta at all, sometimes one is faster to riase by less than a second or so but thats it.

I also get 45c under load now which makes all that sanding worthwhile.

 

[MadScientist]

Originally posted by: Idontcare
My conclusion for this chip - thermal transfer inefficiencies are not the cause of the temperature delta, one chip in my quad simply converts more power into waste heat than the other.

I've come to the same conclusion, or it may be that my Q6600's IHS sits unevenly on the cores.

After checking both my HS and IHS for flatness and re-installing the HS a number of times with different TIMS and application methods, and replacing the plastic push pins with a Thermalright Bolt-thru kit, at idle there was always a 7 C difference between the cores, which dropped to a 2-3 C difference at load.