Weird core0 and core3 temp differences

jrhilsch

Distinguished
Mar 16, 2009
53
0
18,640
I'm running a core i7 920 with a xigmatek dark knight. At idle, my core0 is about 37C while my core3 is about 33-34C. When Prime95 kicks in, my temp difference on my core0 and core3 is about 8-10C. My temps I think are pretty good at load, core0 66C and core3 60. I would like to get all my cores a little closer to my core3 which seems to always be the coolest. Any advice on getting these temps a little closer together, or is it not something I should worry about.

If you guys tell me to reseat my heatsink, what should I be looking for to cause this big differential?

Thanks.
 

freezed1

Distinguished
Jan 17, 2007
283
1
18,790
Hi, this is a common problem. Either your cooler is not seated properly / evenly, or the thermalpaste is not destibuted evenly. Also remember that diff cores will sometimes show different temps even with 100% load and you cant fix that.

 

jrhilsch

Distinguished
Mar 16, 2009
53
0
18,640
Ugh, I just reseated this hs last week because of high temps. I don't think I'm very good at seating this xigmatek dark knight. I might just buy a hs with a completely flat bottom and save myself the trouble to this hdt crap.

I've tried the double line method, I've spread paste in between the heat pipes, and I've spread paste directly on the cpu (not all at once). It seems that spreading on the cpu and in between the heat pipes gave the best results. Any tips that anyone can give in addition to these would be great.

Thanks for the reply btw.
 

overshocked

Distinguished
Feb 14, 2009
1,999
0
19,960



Freized this is kind of incorrect.

You do NOT need to reseat it. This is normal. i usually have my core 0 6-10c different than my other cores during a load.

Totally normal dude. [:thegreatgrapeape:5]
 

jrhilsch

Distinguished
Mar 16, 2009
53
0
18,640
Uh oh, two different opinions. I think I might go with overshocked here only because reseating this hs is a pain. Getting the fan off the heatsink with those weird rubber things is next to impossible even with needle nose pliers. I think I will stick with what I have. Thanks guys.
 

overshocked

Distinguished
Feb 14, 2009
1,999
0
19,960



Yeah, and BTW thos kinds of push pins suck.

I cant tell you how many re-builds that i have done only to realize that my cpu coolers pins cant be put through again.

They only have so much life in them.
 

overshocked

Distinguished
Feb 14, 2009
1,999
0
19,960



Yeah, np.

The i7 is especially bad when it comes to core temp differences because turbo mode makes it very even more than just a normal c2q
 

freezed1

Distinguished
Jan 17, 2007
283
1
18,790
Hi, yes overshocked you are right, i have noticed that pattern also, but i have noticed if you mark an X (for quad cores) with the thermalpaste on your cpu, the delta temp between the cpu cores becomes closer and closer ( they are still off by a few degrees).
 
Overshocked is correct. It is very rare if not impossible to have a multicore CPU have all the same or even very close to the same temperature on all cores. Not all cores on a particular die are going to be the same. When dies are speed binned, in most cases one or more cores limits it to a particular speed bin. Either the core works at a give VID and frequency, but generates excessive heat, or requires a higher than nominal VID to work at a particular frequency (which would also generate excessive heat). So the weakest link sets how the die is sold.

For example lets take a Q6600. When it's die was tested by Intel, tested as a QX6800, the required VID was out of spec for the QX6800 and core 0 generated too much heat. The tester dropped back to Q6700 and now the VID falls in the high end of the tolerance for a Q6700, but core 0 still generates too much heat. So the tester drops back to the nominal specs for a Q6600 and it works within the thermal and voltage limits set by Intel. Intel probably could have sold it as a Q6700, but given it was outside of their own thermal limits for a Q6700, they sell it as a Q6600. Intel laser cuts (or doesn't cut, not sure which) a few traces on the substrate so that it's identified by the motherboard as a Q6600. Now these tests are performed under the assumption that stock cooling is used. Now in the example I used, core 0 was the limiting factor, in fact it could have been any of the four. Also, even though core 0 was our hot core, the other 3 could have had variances in temperature too, just not out of tolerance for a particular speed bin.

I have simplified the testing procedure in the above example somewhat, because Intel also checks the entire range of VID for their dies as well, because this also needs to be identified. Not all Q6600's will have the same VID, but they're guaranteed to fall within a certain range. There are people that swear up and down the lower VID's (more attractive to overclockers) are sold OEM, and the higher end of the spectrum is sold as retail. It seems counter intuitive since OEM is sold cheaper and thus you would think that they would sell them the higher VID's. Another thing to factor in though is the volume sold to OEM's. I don't have any official numbers, but I would guess that less than 10% is sold to retail. Given those numbers it sort of makes sense that OEM's get the lower VID's since this is where Intel makes it's most money even though their margin is thinner.

Hope that helps.

 
Aug 25, 2015
206
0
4,690
I have an i5 4440 cpu.my core 3 temp is always higher than the other cores. Even when my core 0 is at full load and my core 3 at 10% usage ,still core 3 will be 1-2°C higher than core 0. Is that a major problem?? :(
 


You've dug up a pretty old thread. However the logic still stands for your particular processor. A few degrees difference in temperatures is normal. Although your Core 0 is being used and essentially Core 3 is sitting close to idle, all the cores are on the same die, so that heat spreads through the die. I would expect that initially when the load is placed on Core 0, it would be the highest until that heat has a chance to heat the entire die as the heat would have to spread out from the portion of the die that Core 0 occupies.

Do the other cores rise with Core 0 when they are idle? Like Core 3?

If your temperature delta were larger between dies (say 10 degrees), I would think that you might have a seating problem with your heatsink.
 
Aug 25, 2015
206
0
4,690
No no...at idle my temps are like core 0 39C , core 1 36 C , core 2 38 C, core 3 40C. The temp difference at 100 % load on all cores is like core 0 60C, core 1 59C, core 2 57C, core 3 60C. When I play CS 1.6 the core 0 is used but core 3 temp is around 1-2 C higher or same as core 0.