Does a 16c delta between max core temps suggest poor surface contact between IHS and cpu block?

[Kalm_Traveler]

hi guys, I was finally able to get my 10980xe yesterday and swapped out the old 7960x, but noticed that the temp delta between cores is much worse with the new chip.

 

On the delidded + LM 7960x (stock IHS, not lapped), max core-to-core delta was 10c, but with the same setup on this 10980xe the delta is 16c. It will do 5ghz all core just fine but temps are ridiculous (6 cores thermal throttling at 110c) so I had to dial it back to 4.9ghz all core, but even with this the hottest few cores are hitting 90c, while the coolest cores top out at 74c.

 

I didn't think to check the IHS for flatness before installing but given the pretty bad delta my first thought was bad surface mating between the IHS and cpu block cold plate. Optimus (I'm using an Optimus Signature V2 block) recommends not lapping the IHS unless you order a special super-flat cold plate, but I'm not sure if there is a better route to address this.

What do you guys think?

[Fasauceome]

lapping the IHS may help with this delta, but it's also possible that the discrepancy is between the die and the IHS, especially with the amount of cores. (which would really suck because that puppy is soldered)

[LIGISTX]

19 minutes ago, Kalm_Traveler said:

hi guys, I was finally able to get my 10980xe yesterday and swapped out the old 7960x, but noticed that the temp delta between cores is much worse with the new chip.

 

On the delidded + LM 7960x (stock IHS, not lapped), max core-to-core delta was 10c, but with the same setup on this 10980xe the delta is 16c. It will do 5ghz all core just fine but temps are ridiculous (6 cores thermal throttling at 110c) so I had to dial it back to 4.9ghz all core, but even with this the hottest few cores are hitting 90c, while the coolest cores top out at 74c.

 

I didn't think to check the IHS for flatness before installing but given the pretty bad delta my first thought was bad surface mating between the IHS and cpu block cold plate. Optimus (I'm using an Optimus Signature V2 block) recommends not lapping the IHS unless you order a special super-flat cold plate, but I'm not sure if there is a better route to address this.

What do you guys think?

Lapping could help, but personally I wouldn’t do that. I don’t know any voltage or expected speed settings for that chip, but is 5ghz universally accepted as doable? And at what volts? What volts are you running? And have you checked how your thermal paste spread to see if your cold plate is making good contact?

 

But you also could have just got a crappy chip, remember, overclocking is a crap shoot. I have had a beyond amazing i7 920, and I have a very very good 8700k, but I have had moderate chips, and have tried to help a buddy overclock a core 2 quad back in the day and it wouldn’t do a anything past it’s stated speed, that was a dud of a q6600. But point is, could be bad luck, could be pushing more volts than it needs... Not sure if you have, but make sure your stable at the absolute least volts possible, then I usually bump it up by .005v. 

[Kalm_Traveler]

3 minutes ago, LIGISTX said:

Lapping could help, but personally I wouldn’t do that. I don’t know any voltage or expected speed settings for that chip, but is 5ghz universally accepted as doable? And at what volts? What volts are you running? And have you checked how your thermal paste spread to see if your cold plate is making good contact?

 

But you also could have just got a crappy chip, remember, overclocking is a crap shoot. I have had a beyond amazing i7 920, and I have a very very good 8700k, but I have had moderate chips, and have tried to help a buddy overclock a core 2 quad back in the day and it wouldn’t do a anything past it’s stated speed, that was a dud of a q6600. But point is, could be bad luck, could be pushing more volts than it needs... Not sure if you have, but make sure your stable at the absolute least volts possible, then I usually bump it up by .005v. 

5ghz on this one is stable with ~ 1.325v but I run into thermal limitations.  No idea what is 'universally accepted as doable' as hardly anyone has these, and the few reviews I've seen were using worse cooling and way too much voltage. 

 

Haven't removed the block to check spread yet but I've never had a problem with a thick line down the middle and 4 blobs on the corners for LGA2066. Could be the first, but I'm skeptical.

 

Crappy chip would be fine, but I'm not sure that would explain the temp delta unless as the other guy mentioned maybe there's a solder thickness variation under the IHS.

 

If I have some time tomorrow I may pull the block off to confirm TIM spread - if that looks fine I may need to lap the IHS.

[LIGISTX]

7 minutes ago, Kalm_Traveler said:

5ghz on this one is stable with ~ 1.325v but I run into thermal limitations.  No idea what is 'universally accepted as doable' as hardly anyone has these, and the few reviews I've seen were using worse cooling and way too much voltage. 

 

Haven't removed the block to check spread yet but I've never had a problem with a thick line down the middle and 4 blobs on the corners for LGA2066. Could be the first, but I'm skeptical.

 

Crappy chip would be fine, but I'm not sure that would explain the temp delta unless as the other guy mentioned maybe there's a solder thickness variation under the IHS.

 

If I have some time tomorrow I may pull the block off to confirm TIM spread - if that looks fine I may need to lap the IHS.

All of that’s true/good points. Hard to know for sure. Can you delid a 10980XE? That could possibly help assuming it is a bad solder job. But, if it IS soldered, delid is super risky. Checking TIM spread may tell you if you have to lap it, but still not sure I would personally. I delid my 4770k and dropped load temps 22c, and brought the temp deltas to 2-3c coldest to hottest, but that also isn’t something I would recommend to people unless they know what they are getting into. 

[Kalm_Traveler]

38 minutes ago, LIGISTX said:

All of that’s true/good points. Hard to know for sure. Can you delid a 10980XE? That could possibly help assuming it is a bad solder job. But, if it IS soldered, delid is super risky. Checking TIM spread may tell you if you have to lap it, but still not sure I would personally. I delid my 4770k and dropped load temps 22c, and brought the temp deltas to 2-3c coldest to hottest, but that also isn’t something I would recommend to people unless they know what they are getting into. 

Oh that's a risky question - sure it 'can' be delidded but as you said very risky. I didn't mind on the 7960x since it was not soldered, and also not worried about LGA1151 chips like the 8086k/9900k/9900ks since their cpu die is much smaller and there are no SMDs nearby to the IHS mounting so the only real chance of damage is if the solder rips a chunk of the die off somehow.

With the 99x0 and 109x0 chips being soldered and larger die, as well as having SMDs extremely close to where the IHS touches the PCB I'm leery, although end-game version of this rig would be direct die I'll have to think about that.

 

Yeah I saw about 20-25c drop delidding the 7960x so I know it helps a ton on those kinds of chips. That one was easier because it was stability limited, not thermals. Its sweet spot for ambient cooling was 4.7ghz all core on 1.245v and on that no core ever got above 76c; to do 4.8ghz stable required over 1.300v and would thermal throttle. This 10980xe though... it's just all thermals. Given how easily it does 5.0 with relatively lowish voltage I bet with just outdoor winter ambient I could get it doing 5.4-5.5 no problem (last winter I got the 7960x passing 3dmark benches at 5.2 but that required 1.425v and in -12.2c  ambient it was still hitting 80c under load).

 

Thanks for the input - I will try to check the paste spread tomorrow and report back. Don't mind lapping the IHS if it will help but just want to make sure I'm gathering as much info as possible before making that irreversible change.

[Bitter]

Could you return/exchange it and try again with the silicon lottery?

[Kalm_Traveler]

1 hour ago, Bitter said:

Could you return/exchange it and try again with the silicon lottery?

oh man... it took me 2 months just to find one in stock somewhere. Central Computers in California listed 4 of them on Wednesday last week, I found out about it Wednesday night and ordered one, they were sold out the next morning.

If an exchange was possible I might try that but these things are like unicorns.