One core very hot no matter what. Thoughts?

[ozzy702]

I have a 6700k that I was using with a previous motherboard. I popped the 6700k in a brand new ASRock Z270m pro4 and when I went to check temps like I normally do with a new build I noticed that one of the cores was significantly hotter than the other three under load. That core (same one every time I've swapped coolers) under Prime95 will hit 100C and lock up the system. With handbrake it almost immediately hits 96C and then starts to climb. The other three cores were 73C, 57C and 61C with the smallest cooler I tried.

I've used the following cpu coolers, all with the same results. Hyper TX3, Cryorig H7, Cryorig M9. I had all three on hand.

After removal, thermal paste appeared to be spread evenly across the cooler/cpu.

Vcore is only ~ 1.12v.

Thoughts?

Given that I've tried three different coolers with five different applications of paste all with the same results I'm inclined to think that either the mobo has an issue (cpu or cooler not sitting right and contact is poor) or the CPU itself is warped or broken.

CPU seems to work fine other than stress testing.

 

[Burpo]

What version Prime are you using? Newer versions will cause your voltage to shoot up to 1.4+

Try an older version that doesn't use AVX..

 

[ozzy702]

What version Prime are you using? Newer versions will cause your voltage to shoot up to 1.4+

Try an older version that doesn't use AVX..

Newest version but I watched voltage and that wasn't an issue. I used it on my 7700k box and it wasn't a problem with temps. I'll try another stress tester but I don't think that's the problem given that only one core is going thermonuclear.

The core that's running hot runs hot when any load at all is present. Only time when all four cores are close is when it's idling.

 

[Bouowmx]

Latching on to this thread with Intel Core i7-870: core 3 (0-based indexing) is 5-7 C higher than the remaining cores, observed with CRYORIG H7 and ARCTIC Liquid Freezer 120. I ignored the difference and attributed it to variance, although more knowledgeable users can chime in.

 

[ozzy702]

Latching on to this thread with Intel Core i7-870: core 3 (0-based indexing) is 5-7 C higher than the remaining cores, observed with CRYORIG H7 and ARCTIC Liquid Freezer 120. I ignored the difference and attributed it to variance, although more knowledgeable users can chime in.

I wouldn't worry about 5-7C as long as it's not hitting crazy high temps. That's not that huge of a delta. I'm seeing a 27C delta between this core and the next hottest core no matter what cooler and paste I use.

 

[ozzy702]

What version Prime are you using? Newer versions will cause your voltage to shoot up to 1.4+

Try an older version that doesn't use AVX..

Just tried handbrake. Immediately hits 96C and then starts to climb from there.

 

[Rayniac]

Does that CPU model have an integrated GPU? If it's enabled that could be the culprit.

 

[beginner99]

Are you sure you are running multiple threads? if not it would explain why only 1 core gets so hot.

Else could be an issue with thermals paste, mounting of the cooler or even the IHS or whole chip package being warped (due to a defect or having been under too much and uneven pressure from a cooler.

 

[TheELF]

Are you sure you are running multiple threads? if not it would explain why only 1 core gets so hot.

No it wouldn't, windows uses thread hopping,on every time slice all threads are reallocated onto different cores to prevent something this.
Unless the mobo has some weird setting that would stop that from happening (never heard about this being possible) it could very well be that the TIM wasn't correctly applied at the fab.

 

[Abwx]

Thermal contact between the die and the metallic coverage could be not homogeneous for some reason, irregularly deffused thermal paste, very slightly bended metal surface (0.05mm is enough.)...

 

[maddie]

Is nobody considering the possibility that the CPU has a defective core? Some unlucky few do get hit by lightning.

 

[Hail The Brain Slug]

I would RMA to intel, that is all kinds of not right.

Unless you're more adventurous, then I would delid it.

 

[mtcn77]

I'd be careful. As a long time fan of Arctic Silver 5, I always keep in mind should it cure, the die can go bust. Runny pastes that don't cure aren't subject to this, but the point is the temperature elevation due to non-conductance which you seem to ascribe. *Or it could be the Skylake spikes*.

 

[LTC8K6]

I concur with contacting Intel about an RMA.

 

[sm625]

You're not going to get that type of variance from anything happening outside the heatspreader. You need to delid it. There is probably a gap right above that core.

 

[scannall]

I concur with contacting Intel about an RMA.

Another vote for RMA. Don't delid something that's defective.

 

[ozzy702]

Are you sure you are running multiple threads? if not it would explain why only 1 core gets so hot.

Else could be an issue with thermals paste, mounting of the cooler or even the IHS or whole chip package being warped (due to a defect or having been under too much and uneven pressure from a cooler.

All cores are enabled and all cores are being stress tested.

I'm going to swap out the mobo and see if the same behavior is present. If it is it's a problem with the chip, which is weird because it worked fine in my system for almost a year. If not then it's the mobo and I'll RMA it.

 

[ozzy702]

Update: I popped the 6700k into a Z170 mobo. Initially I thought everything looked good because the mobo was reading temps in the 30s instead of temps in the high 50s like the first mobo did. Boot into windows, idle temps all look great. Start stress testing...

Core# 0 59C
Core# 1 71C <- problem core
Core# 2 49C
Core# 3 51C


First thing I notice is there is no way this little cooler is keeping any of my cores at 49C during stress testing. Second thing I notice, Core #1 the problem core is still significantly hotter than the other three cores, even though it's not hitting 100C like it was in the prior mobo.

So now I'm thinking it's actually the 6700k that's bad. Thoughts?

What's the warranty process look like with Intel? This CPU is about a year old.

 

[LTC8K6]

Update: I popped the 6700k into a Z170 mobo. Initially I thought everything looked good because the mobo was reading temps in the 30s instead of temps in the high 50s like the first mobo did. Boot into windows, idle temps all look great. Start stress testing...

Core# 0 59C
Core# 1 71C <- problem core
Core# 2 49C
Core# 3 51C


First thing I notice is there is no way this little cooler is keeping any of my cores at 49C during stress testing. Second thing I notice, Core #1 the problem core is still significantly hotter than the other three cores, even though it's not hitting 100C like it was in the prior mobo.

So now I'm thinking it's actually the 6700k that's bad. Thoughts?

What's the warranty process look like with Intel? This CPU is about a year old.

http://www.intel.com/content/www/us/en/support/processors/000005494.html

My RMA of a 4790K went very smoothly.

 

[ozzy702]

http://www.intel.com/content/www/us/en/support/processors/000005494.html

My RMA of a 4790K went very smoothly.

I've gone through 50+ Intel CPUs over the years and this is the first time that I have had a problem. They seem to have phenomenal quality and it's really good to know that the RMA process is solid.

 

[LTC8K6]

I've gone through 50+ Intel CPUs over the years and this is the first time that I have had a problem. They seem to have phenomenal quality and it's really good to know that the RMA process is solid.

The 4790K I got back was brand new in the box just as if I had ordered it from Amazon. It's still going strong.
I have heard that Intel doesn't really question you about the return if it's within the warranty period and you haven't made a habit of returning CPUs.
That was true of my RMA. I was only asked to give them the numbers off of the CPU. They never asked me what happened or anything.