~10°C difference in temp between cores under load?

TJ Hooker

Titan
Ambassador
I just installed my new i5 6600k, and decided to run a few benchmarks/stress tests (haven't overclocked anything yet). I have a 212 EVO, using MX-4 paste.

I noticed there seems to be a consistent temperature difference between cores, partcularly between cores 1 and 3. Core 1 is the hottest, 0 and 2 a little cooler, and then 3 is the coolest. Running P95 v26.6 small FFTs, core 3 is consistently 8-10°C cooler than core 1.

Now, I know some difference is expected, due to varying core loading and sensor error, but I figured something like P95 would be a pretty consistent load across all cores. Is ~10 C temp difference normal? Is it worth repasting? I admit I ended up fiddling with the cooler to line up the screw holes after I placed it, think I lifted it up slightly once or twice, maybe some bubbles got in.

Edit: max temps for each core running P95 after about 10-15 min are 53/57/52/48.
 
Solution


From the Intel Temperature Guide - http://www.tomshardware.com/forum/id-1800828/intel-temperature-guide.html

Section 13 - Thermal Testing @ 100% Workload:

Under "Results", second paragraph ...

"Intel’s specification for Digital Thermal Sensor (DTS) accuracy is +/- 5C. This means deviations between the highest and lowest...

CompuTronix

Intel Master
Moderator


From the Intel Temperature Guide - http://www.tomshardware.com/forum/id-1800828/intel-temperature-guide.html

Section 13 - Thermal Testing @ 100% Workload:

Under "Results", second paragraph ...

"Intel’s specification for Digital Thermal Sensor (DTS) accuracy is +/- 5C. This means deviations between the highest and lowest Cores may be 10C ... "

Given that your 8 to 10C Core temperature deviations are on the edge of spec, and you're uncertain as to whether you've compromised the quality of your TIM application, I recommend that you re-TIM and reseat you CM Hyper 212 EVO, then re-test. If you still have the same deviations, then you may want to consider RMA'ing your i5 6600K.

However, before doing so, since you haven't yet overclocked, you might want to know if the sample you have in hand is a good overclocker before you exchange it for a potential dud. If it turns out to be a good overclocker, then perhaps you might be willing to overlook marginal thermal deviations. Nevertheless, keep in mind that when overclocking, thermal deviations may change at higher Core temperatures.

Also, 53/57/52/48 is rather low for Prime95 V26.6 Small FFT's with that cooler on a 6600K running at stock.

Q: What is your ambient temperature?

Q: Do any of your Cores idle below ambient?

Q: What monitoring utility are you using to read Core temperatures?

CT :sol:
 
Solution

TJ Hooker

Titan
Ambassador
1) Unfortunately there's nothing in my apartment to measure temp. I'd guess 22-24 C.
2) No, I think they were all at least high 20s, or in the 30s.
3) I tried HWInfo and RealTemp. I think both were reporting more or less the same values.

Regarding the OC potential of my chip, I think there's something seriously wrong with it, or/and my mobo. Tried OCing to a measly 4.0 GHz (all cores). Went all the way up to 1.35V and would still crash within minutes. Went back to stock, and things continued to crash, sometimes couldn't even make it to the desktop without hanging/crashing. I noticed that, if I looked at clock speed with HWInfo right when I got to desktop (things still starting up), the CPU was getting up to 3.9 GHz on all cores, which I don't think should even be possible at stock? (Turbo should only go to 3.9 on 1 core). Disabling turbo seems to solve the crashing issues, but it was getting late so I didn't get much time to confirm.

The reason I'm worried that it might be something other than my new CPU being a dud is that I recently moved somewhere with super low humidity, and I've been shocking the hell out of everything I touch, including my PC. A couple times I've touched it with a shock and it immediately powered off/rebooted, and at least one time it blue-screened. Worried I may have partially damaged something. Might try sticking my previous i3-6100 back in and see if I get any crashes. On a possibly-related note, I remember when I tried non-k BCLK OC with my 6100 it overclocked pretty terribly as well (this was some time ago, before any of the shocking).

I'll probably RMA my 6600k, although I might start a new thread specifically on the crashing issue first.

Thanks for your help.