• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

2500k core temperature variation is strange

D

DeadFalling

n00b
Joined
Jul 29, 2008
Messages
48
So I'm trying to OC my 2500k to 4.5Ghz and I'm getting some funny readings off of the individual cores.

Aa8i1.png


As you can see, Core #0 is consistently cooler than the rest. The temp difference ranges from 9 degrees all the way to 15 degrees. I tried removing the heatsink (Hyper 212+) and reapplying thermal paste but it's the same deal... Hell I think it got worse when I reapplied it! :confused:
 
My 3770 on a Hyper 212+ is the same way, if core 0 is baseline, then core 1 and 2 are usually 1-2c above that, and core 3 is +8-10c (current temps: 30/32/32/39c). However, they even out under load so I haven't really worried about it. After a session of BF3, realtemp will report a max of around 64ish for all cores.
 
eloj said:
My 3770 on a Hyper 212+ is the same way, if core 0 is baseline, then core 1 and 2 are usually 1-2c above that, and core 3 is +8-10c (current temps: 30/32/32/39c). However, they even out under load so I haven't really worried about it. After a session of BF3, realtemp will report a max of around 64ish for all cores.
Click to expand...

Looking at the graph mine don't even out :S


Took the heatsink back off and applied a bit more thermal paste where it looked like it could possibly be lacking. Here's a new graph:


X4ZEo.png


Seems pretty much the same deal?
 
This is normal... Unless one core was way off charts, i.e. 20deg higher than others, there is nothing to worry about...
 
You are assuming that Intel uses core temperature sensors that are 100% accurate but they are not. A range of +/- 5C when all 4 cores are equally loaded is considered normal. What this means is that if one core was reporting 70C and another one was reporting 65C and another one was reporting 75C, you are likely only seeing the error in the sensors. Run Prime 95 - Small FFTs and during the first test when your core temps have stabilized, see if they fall within the +/- 5C range. 99.9% of the time they do.

These sensors are only calibrated at full load temperatures and are only designed to control thermal throttling and thermal shutdown. 100% accurate core temperature reporting was never Intel's goal for these sensors.
 
My 2600K is like that too. The best thing to do is use a program that reads the "CPU Package" like AIDA. That reading will be the hottest core reading and that's a good sensor to go by. What is your 100% load temps across cores? Mine is normally 5c difference during idle but 1c-2c when stressing
 
The peaks on the graphs I posted above are all 4 cores at 100% running Intel Burn Test at High.
Alwcc.png


For example here, the max temps there is a discrepancy of 10 degrees between core #0 and #1, 12 degrees between #0 and #2, 11 degrees between #0 and #3, and 14 degrees between #0 and Package.

Maybe I did something wrong on my OC?


pontiac4life said:
Yeah, that's a bit more of a difference then I would like too. Did you check for any BIOS updates for your motherboard? Did you try to use other programs to check the temps to see if CPUID is wrong?
Click to expand...

Yes it's got the newest BIOS. Yes it shows the same in CPUID HM, Open HM, and SpeedFan.


pontiac4life said:
For me using AIDA the CPU Package is always the hottest core but yours isn't reading what it should. I'd try another program
Click to expand...

Package is the hottest core on mine as well?
 
Last edited:
Yeah, that's a bit more of a difference then I would like too. Did you check for any BIOS updates for your motherboard? Did you try to use other programs to check the temps to see if CPUID is wrong? For me using AIDA the CPU Package is always the hottest core but yours isn't reading what it should. I'd try another program

Here is mine, as you can see the CPU Package is the same as the hottest core, your isn't reading correctly. My Package always reads the hottest core regardless on idle or load

Temps.png
 
Last edited:
pontiac4life said:
Here is mine, as you can see the CPU Package is the same as the hottest core, your isn't reading correctly. My Package always reads the hottest core regardless on idle or load

Click to expand...

The Package being 1 degree off in CPUID HM is most likely irrelevant. It's probably just a slightly different calculation or rounding. Installed AIDA to prove this:

Nou0b.png



Package and hottest core are the same, but there is still the large gap between Core #0 and the others.
 
AIDA is reading the Package correctly from what you posted. Package should always be the same as the hottest core and is really the only temp you need to be reading because if you know what the package temp is then all the other cores have to be equal or lower. Stress test for a few minutes and see if AIDA package reads correctly and post back

The gap between core 0 and the others is ok. The hottest core is what is important
 
That's not far off of the temp difference I get on realtemp. Core0 is always the coolest, core2 always the hottest and core1 and core3 are within 5c of both 0 and 2.
 
Stress test while watching AIDA Package temp and see if it changes and if the same as the hottest core, if yes just use that to call it a day because there is no reason to watch all cores when the package is reading the hottest and that's what counts. It's perfectly normal for there to be a difference between cores. AIDA shows 8c difference between your cores and the other program shows like 14c, so maybe AIDA is better for you. Was the AIDA pic from idle? See what the core difference is and the package is when stressing
 
pontiac4life said:
Stress test while watching AIDA Package temp and see if it changes and if the same as the hottest core, if yes just use that to call it a day because there is no reason to watch all cores when the package is reading the hottest and that's what counts. It's perfectly normal for there to be a difference between cores. AIDA shows 8c difference between your cores and the other program shows like 14c, so maybe AIDA is better for you. Was the AIDA pic from idle? See what the core difference is and the package is when stressing
Click to expand...

The Aida test isn't that important. The IBT is much more taxing. See this:

H1pAU.png


Since IBT is a harsher test, here's Aida stats after an IBT test:

VFO5b.png


Still reports the first core temp as 11 degrees off from Package temp. The graph above also shows how the first core is always far under the rest, regardless of CPU usage... :confused: :S
 
The one core being cooler is ok, we have told you that before, just make sure the hottest core is within safe temps and call it a day.
 
Everyone with a SB will tell you the same thing - there is always one core that is about 10C different from the rest. It is completely normal with that chip. One theory is that that core is located next to the GPU block on the die, and so has all that dark silicon to act as a heatsink, but whatever the reason, it is no big deal.
 
Forceman said:
Everyone with a SB will tell you the same thing - there is always one core that is about 10C different from the rest. It is completely normal with that chip. One theory is that that core is located next to the GPU block on the die, and so has all that dark silicon to act as a heatsink, but whatever the reason, it is no big deal.
Click to expand...

Heck, 1366 did the same thing. The only chips I've seen where all cores read the same temps are AMD CPUs, and they only have one thermostat actually logging data for all cores.
 
Alright well thanks for the confirmation that my cpu is ok. During this whole debacle I cleaned + reapplied thermal paste once and reseated/touched up paste 3 times. Ended up also moving around some wiring while I had it open and with the heatsink + wiring changes I shaved off a couple degrees from the max core temps :p
 
The results I just got from my 2600K:
0: 56C
1: 65C
2: 66C
3: 60C

I've tried applying different TIM, reseating the block, etc, but I can't get those middles cores to be closer to the outer cores. I figure it must be error in the temps, or probably more likely, those cores just run warmer due to position or internal TIM. It's been ~18 months, but I think I've finally accepted that I can't do anything about it... unless I wanted to delid my cpu, but I think I'll pass on that front. Though having 56C all around would be ideal....
 
My 2700k 4.8Ghz on Noctua NH-D14 is the same way. Core 0 runs much cooler.
 
You must log in or register to reply here.
Back
Top