core temperature differences?

[pulverizor]

ok maybe I'm being paranoid but I guess I just want peace of mind

so I got this old 1156 proc an i5-760 quad-core from 2010 (Lynnfield Nehalem etc.)


so the problem (if it is a problem) is simple: core temps ain't quite the same:

core0 & core2 are always a bit hotter

- at idle, core0 & core2 are 3-5°C hotter than core1 & core3
(and task manager shows no higher activity for any core over the others)

- at very heavy load (Prime95), core0 & core2 are 1-2°C hotter than core1 & core3


so is this normal? especially the differences @ idle
especially since all 4 cores are covered by same metal shield so it should be evenly spread

 

[Mr.Scott]

ok maybe I'm being paranoid but I guess I just want peace of mind

so I got this old 1156 proc an i5-760 quad-core from 2010 (Lynnfield Nehalem etc.)


so the problem (if it is a problem) is simple: core temps ain't quite the same:

core0 & core2 are always a bit hotter

- at idle, core0 & core2 are 3-5°C hotter than core1 & core3
(and task manager shows no higher activity for any core over the others)

- at very heavy load (Prime95), core0 & core2 are 1-2°C hotter than core1 & core3


so is this normal? especially the differences @ idle
especially since all 4 cores are covered by same metal shield so it should be evenly spread

Totally normal.
I've seen core temps differing up to 12-15c.

 

[pulverizor]

Totally normal.
I've seen core temps differing up to 12-15c.

so whats the explenation for that?


btw when I applied the mx4 I just put a blob in the middle of the cpu (volume equivalent to 2 rice grains maybe) then pressed the heatsink & moved it around a bit before setting it, however despite this I noticed that the grease only leaked out slightly from 1 edge which means it didn't reach the other 3 edges so I reckon the chip ain't fully covered in grease. no idea if this will result in sizeable loss of cooling power (installing this huge HSF ain't exactly a cakewalk so I hesitate to dismantle everything again) but even so this shouldn't result in temp discrepancies between the cores should it?

 

[Mr.Scott]

so whats the explenation for that?


btw when I applied the mx4 I just put a blob in the middle of the cpu (volume equivalent to 2 rice grains maybe) then pressed the heatsink & moved it around a bit before setting it, however despite this I noticed that the grease only leaked out slightly from 1 edge which means it didn't reach the other 3 edges so I reckon the chip ain't fully covered in grease. no idea if this will result in sizeable loss of cooling power (installing this huge HSF ain't exactly a cakewalk so I hesitate to dismantle everything again) but even so this shouldn't result in temp discrepancies between the cores should it?

It could. Un-even tightening.
Most of the time the differences are due to poor contact or worn out TIM under the lid.

 

[pulverizor]

ok so I made sure all 4 screws were tightened same (btw heatsink if heavy plus it's high so center of gravity is far from the mobo, so I tried tilting the case at different angles ^^ didn't change anything for the temps)
and TIM is fresh (just applied today, new seringe)


btw are the cores in a "line" or in a "square" layout on a Nehalem 4 core chip? (do core0 & core2 touch or not?)

are there other possible explanations for the sort of temp differences I'm getting?

 

[trents]

Yes, there are other explanations. For one, the core temp sensors may not be exactly placed, meaning they are not the same distance from each core. Or some other component on the die closer to some of the temp sensors is generating a little extra heat. Or, small differentials on calibration of the sensors themselves. Actually, your core temps are pretty close to one another compared to what we see on a lot of CPUs. Sometimes we see 10+ C differences. You will also find that the spread will increase with an increase in overall CPU temp. For instance, at 60c there might be a 3c spread between the coolest and the hottest but at 80c there might be a 5c spread.

 

[pulverizor]

You will also find that the spread will increase with an increase in overall CPU temp. For instance, at 60c there might be a 3c spread between the coolest and the hottest but at 80c there might be a 5c spread.

I didn't know that

it's weird cause in my case it's the opposite, temp differences are greatest @ idle & decrease with load (which is less scary come to think of it)

so is that a different phenomenon at play?

 

[trents]

Well, idle doesn't really count because the temp sensors are not calibrated for accurately reading the temps on the low end. Also, not all the cores are loaded all the time unless you have Turbo and Speed Step disabled. If you will heat the thing up with IntelBurnTest or Prime95 and have HardwareMonitor open during the stress test you can see the individual core temps jump around quite a bit as one, then the other are loaded. What you want to check is the "Max" column of HWMonitor after 5-10 minutes of loading to see what the average differential looks like over time.

 

[pulverizor]

I know but in my case all 4 cores are @ ~0% when idle yet temperature differences can reach as high as 5°C

and the difference decreases when activity goes up (in Prime95 all cores are @ 100% and temp differences are reduced to 2°C at most) whereas you said it's suppose to increase with activity, hence why I'm even more intrigued now

 

[WhitehawkEQ]

Check the multiplier for each core, I built a Z97 1150 LGA i5-4670K for a friend and not all cores had the same multiplier, so I manualy set them all to same for 3.1GHZ.
Here is screen shot of HWMonitor and Prime 95:

 

[pulverizor]

Check the multiplier for each core, I built a Z97 1150 LGA i5-4670K for a friend and not all cores had the same multiplier, so I manualy set them all to same for 3.1GHZ.
Here is screen shot of HWMonitor and Prime 95:

wait one, individual cores each have a multiplier? how can I check mine? (AFAIK only 1 multiplier for the whole cpu)

 

[EarthDog]

Some boards do not have that option... and I am not sure your generation of CPU had those options anyway. I don't recall with certainty though.

 

[pulverizor]

not mine so I guess same multiplier for all cores

btw for a split second all temps were equal on all cores
anyway I'm tempted to remount my HSF cause I've absolutely no idea if the entire surface of the cpu heat spreader should be covered in thermal paste for max efficiency or not

 

[trents]

Should be covered or is covered? Yes, the entire surface of the heatspreader should be covered with TIM. The center part is the most important, however.