e8400 cores vary in temp as much as 12C!

[Homer Simpson]

since i got my e8400 a couple weeks ago, the cores have always been anywhere from 5-12 degrees different. on average its 9-10, idle or load. based on searching, typically its only gonna vary a few degrees. mine seems excessive. it was this way with the stock hsf and with the zalman thats on there now. i used AS5 with both using the strip down the middle method per AS5 website. could my heatspreader be really uneven or could it not be making full contact to the cores? i dont know that i can return it since it "works". thoughts?

 

[TC91]

Does the temp difference occur when say you are running dual prime 95, or is it just during normal usage? If you are only using one core, then there will be a temp difference, if you are stressing both cores, then the difference shouldnt be more than 1-3c between your cores. If they are like 12c different while under full load, then there may be a problem with the temp sensor.

 

[Cutthroat]

More pressure on one side of the IHS than the other. You might need to lap it to fix it.

 

[Homer Simpson]

the difference shows at idle, load, prime95, all the time.

lapping? argh, was afraid that might be a reply i got. never done that to a cpu but i guess i can try and if i hose it up, im only out $200

 

[Homer Simpson]

i guess it could also be flaky sensors since these e8400's have a known sensor issues. it idles at ~40/30 according to realtemp (supposedly the most accurate for 45nm chips). under prime its typically 60/53. i think the highest i saw was flickers up to 62/55. just always seemed kinda hot compared to my e6600 but if they are really as inaccurate as ppl say, then i guess i shouldnt worry too much.

 

[rge]

lapping is not going to help, nor is reseating. One of your DTS diodes on core is off, or both are off 5C in opposite directions. You can put your heat sink (and I have) on just half of cpu, and you still can not even create a 2C gradient between two cores. The thermal conductance on die is too high, and E8400 are too small to create a real gradient of much more than 1-2C between the two dies at steady state.

I have taken IHS off and measured cores against real temp, and have stacked, with TIM, 2 IHS on my cpu to measure the gradient across tim and IHS. http://www.xtremesystems.org/f...3085792&postcount=1525 Tjmax is 95C, realtemp is telling you what DTS is reporting. And one or both of your sensors is misreporting. Intel does not consider this a problem, as they may both become accurate around 95C when throttling would occur, and even if both 5C off, your cpu will still be ok for throttling. Intel does not care how inaccurate they are until then, hence they do not calibrate in the low end. BTW...most people's e8400 read 5-20C too high at idle end, and most read very accurate past 60-65C, it just your are one that has one DTS reading high and one lower, while most have matched discrepancies.

nevertheless, I would write to intel, more people that complain (I certainly have), the more likely intel is to correct this "nonissue" for them but an "issue" for us. Just dont expect anything more than a generic, and often inaccurate response.