Uneven Core Temp(s)?

[Domain]

I'm not sure if this is the right place (or maybe CPU cooling, or elsewhere ), but i've encountered a few questions hopefully someone can answer. For reference, I'm referring to the system posted here (Core i7 920/Mugen 2):

viewtopic.php?t=53925

The biggest question is ... should I be seeing uneven core temperatures? Example:

@idle: 48, 47, 46, 44
@prime95 load: 73, 72, 72, 69

Essentially there is almost always a 4 degree gap... which leaves me questioning if this type of variation is acceptable... or if perhaps there is an issue with the mounting of the cooler/thermal interface? I really hate to have to tear down the entire system... at least if it isn't going to make a difference.

Then along those same lines... do these temps look somewhat high for this setup? Idle temps really don't fall much farther below the above... maximum load temps are fairly close to documented maximum for the i7, but the system controlled CPU fan isn't being activated until the highest temp core reaches 71/72.

Any insight that could be provided is appreciated

[smilingcrow]

A 4 degrees difference is negligible I would say.

[new2spcr]

I'm not sure if this is the right place (or maybe CPU cooling, or elsewhere ), but i've encountered a few questions hopefully someone can answer. For reference, I'm referring to the system posted here (Core i7 920/Mugen 2):

viewtopic.php?t=53925

The biggest question is ... should I be seeing uneven core temperatures? Example:

@idle: 48, 47, 46, 44
@prime95 load: 73, 72, 72, 69

Essentially there is almost always a 4 degree gap... which leaves me questioning if this type of variation is acceptable... or if perhaps there is an issue with the mounting of the cooler/thermal interface? I really hate to have to tear down the entire system... at least if it isn't going to make a difference.

Then along those same lines... do these temps look somewhat high for this setup? Idle temps really don't fall much farther below the above... maximum load temps are fairly close to documented maximum for the i7, but the system controlled CPU fan isn't being activated until the highest temp core reaches 71/72.

Any insight that could be provided is appreciated

I've read somewhere that a 4-5 C difference between the cores is expected.

[burebista]

Don't bother with those 4°C it's normal.
Nehalem is hot like hell but frankly I'd expect more from Mugen 2. Default frequency? What Vcore?

[Domain]

Hmmm, well...

Due to all the power saving features the motherboard enables, the core voltage hops around all over the place... under load it is approximately ~= 1.17.

I haven't explicitly attempted to do any overclocking (this thing is plenty fast enough as is), however the XMP profile on the memory (DD3 1600) raises the QPI/Uncore voltage slightly (1.35 vs 1.15), but leaves the memory voltage at 1.5. Additionally, the board enables "Turbo Boost" which seems to push the processor from 2.66 to 2.93 depending on... well what it feels like I suppose

As for the Mugen 2... the system keeps the included fan around the 150rpm range in general... once any of the cores reach around 71/72, it spins the fan up to 550-600 rpm range, and maintains the temperatures around this level. If I explicitly override the fan control, and allow it to run at maximum speed (1300ish), full load never exceeds around 63.

I've tried turning off all said features/lowering memory speed, but it really has no significant impact on the temperatures. For what it is worth however, the motherboard Ambient temperature is reported to be around 41 at idle, primarily due to the GTX275 space heater that I call a graphics card

[baconandeggs]

dont worry mate. mine had like 2-3deg difference but they have now equalised. i think it might be the TIM settling in.

[Vibrator]

4c difference? If I were you, I'd just leave it, but try a TIM reseat if you really care

[Domain]

Well after some experimentation:

I went ahead an unmounted everything again, to validate there wasn't a problem with the seating... only to find that it was pretty much perfect to begin with. The chip surface itself is somewhat uneven, but honestly I don't want to go through the effort of lapping...

Having explored various fan controlling options, I have come to the conclusion that 40-50C idle temperatures are probably not all the uncommon for a silent build, and that apparently Intel (at least in their motherboard BIOS) appears to believe that 70-75C is range when the chip starts to need additional cooling.

Auto RPM control @ idle (50-150rpm): 45,43,43,42
Manual RPM control @ idle (800-1200rpm): 34, 32, 32, 31

Auto RPM control @ peak (600-700rpm): 73, 72, 72, 69
Manual RPM control @ peak (800-1200rpm): 63, 62, 62, 59

Having spent hours reading/searching/asking, the general consensus seems to be that people are horribly confused. While a great deal seemed to indicate their processors idled in the mid 30's, and peaked in the upper 50's, they rarely made mention of their fan speeds, and the ones that did all ran in ranges way outside of my acceptable audible range (1900-3000rpm). This all of course is also neglecting that their main system fans are running way higher then the 800rpm range as well.

So i've come to believe that the Mugen 2 is doing a fairly good job, and really it is a matter of airflow... I may swap the stock fan PWM fan for a constant RPM fan that perhaps runs a little slower... or I may just let the system idle a little higher then the "norm".

Thanks everyone for the input

[austinbike]

Most likely, the processor is loading the cores from 1-4. That means that 1 always gets the load first. 2 gets the next thread if 1 is busy. 3 if 2 is busy and finally, core 4 fires up only when the first 3 are engaged. That would mean that, in most circumstances, no matter how you adjust things, 1 would be the hottest and 4 would be coolest.

[mlc]

I've got an i7/Mugen 2 setup as well, and I noticed over the several times I've reseated my CPUs (and cleaned/reapplied TIM) that the variance between individual cores went from 2 or 3 degrees C to the more recent 4C like yours.

what I did this time was to use less TIM than last time, hoping for temp drops. that didn't really happen - until I realized that I didn't have to stick to using "AUTO" voltage settings (undervolting works wonders, I just found).

the previous installation, when I had a 2C variance, I applied more TIM so that it would squish and fill the gaps on the side (have you seen a lot of pics of peoples TIM imprint after they remove their coolers? and noticed how the contact tends to be in the middle in the shape of an "I"). maybe that's what kept the variance to a 2 degree difference.

Anyways, I found that the temps drop when you undervolt and now that I'm on 0.95V cpu voltage, the temps don't exceed 56C, and the variance has dropped to within 2C as well.

[ascl]

The guy that wrote RealTemp (unclewebb) who has done a bunch of testing, theorised that all quads (intel anyway) had different TJMax on each pair of cores (actually, this was for Core 2, not sure about i7). Hence, he suggested that a Q6600 had a 100 TJmax for Core 0 + 1, and a 105 TJMax for Core 2 + 3.

His reasoning was that this would stop all 4 cores throttling at the same time.

Using this suggestion, my cores are very very consistent when under load. The intel thermal measurements are more accurate at higher temps (ie when they need to be), and so there is more variance at lower temps.