Core #3 is hotter than the rest

[joemiller 269]

Hello gents, 

Thought I would ask this question here as this has been bothering me for a few weeks. I have an Intel 8086K overclocked on all cores @ 5.3Ghz.

The problem is that under load  (P3D), the third core gets around 18 degrees hotter than the other 5- always. Why is this, and what can I do to resolve or fix this issue?

 

[Bob Scott 7,716]

1 hour ago, joemiller said:

Hello gents, 

Thought I would ask this question here as this has been bothering me for a few weeks. I have an Intel 8086K overclocked on all cores @ 5.3Ghz.

The problem is that under load  (P3D), the third core gets around 18 degrees hotter than the other 5- always. Why is this, and what can I do to resolve or fix this issue?

 

18 deg?  Yeah, that's a lot.  You might have a bubble/void in your TIM over that part of the CPU package. 

I'd remove the heatsink, scrape the old TIM off both the CPU and the heatsink with an old credit card, wipe both sides clean with a section of coffee filter soaked in 90% isopropyl alcohol, then re-apply the TIM and try again. 

Also, make sure the heatsink holddown screws are against the stops on all four corners when you're done.

Regards

[joemiller 269]

1 hour ago, w6kd said:

18 deg?  Yeah, that's a lot.  You might have a bubble/void in your TIM over that part of the CPU package. 

I'd remove the heatsink, scrape the old TIM off both the CPU and the heatsink with an old credit card, wipe both sides clean with a section of coffee filter soaked in 90% isopropyl alcohol, then re-apply the TIM and try again. 

Also, make sure the heatsink holddown screws are against the stops on all four corners when you're done.

Regards

Thanks Bob. I will try that. Something's not right somewhere. The CPU would hold 5.3Ghz with 141 Volts. Now it's asking for more voltage but, it gets too hot and goes BSOD. Now, it will only hold 5.1Ghz with 136 volts. So, I'll try what you say, or maybe the CPU is losing it's "life". But, it's barely a 2 month old CPU.

Also, what type of thermal  paste would you recommend? 

[ckyliu 1,749]

@joemiller 1.52V and 105'c is the Intel recommended maximum but that's just the max they think you can use without very quickly destroying the chip, it would die within months at that! I wouldn't exceed 1.4Vcore and 85'c if you want the CPU to last a few years; if you need more voltage than that's just the nature of the silicon lottery and you'll have to settle for a lower clock speed to keep it stable. Sounds like could have pushed it too hard before and prematurely aged it (you didn't mention absolute temperatures, only relative?), are you using automatic overclock utilities?

Edited December 12, 2018 by ckyliu

[martin-w 3,234]

11 hours ago, joemiller said:

Hello gents, 

Thought I would ask this question here as this has been bothering me for a few weeks. I have an Intel 8086K overclocked on all cores @ 5.3Ghz.

The problem is that under load  (P3D), the third core gets around 18 degrees hotter than the other 5- always. Why is this, and what can I do to resolve or fix this issue?

 

 

 

 

 

Perfectly normal to have one core hotter than the rest. Also depends on the software you are running. It's usually around a 10 degree delta between coolest and hottest core. The question to ask though is whether this core is being pushed harder by P3D? What are the temps like with your other software? Try it with a stress test. 

18 degrees is higher than I would have expected. Other than software, a dodgy IHS/die interface courtesy of Intel can do it, not a very proficient attempt at delidding, uneven cooler mounting etc. 

Try stressing it with something else and check temps, try remounting cooler. Unlikely that a void in the TIM would be the cause, as the pressure from mounting usually does a good job of evening out the TIM. Remounting cooler might help if it's not tightened down evenly.

If all else fails, and you perceive it as an issue, then it could be Intel at fault re their IHS mounting in which case a delidd would probably help.

What you are looking for is around 10 degrees delta while running a stress test. That would be within the norm. 

 

 

 

 

Edited December 12, 2018 by martin-w

[joemiller 269]

4 hours ago, ckyliu said:

@joemiller 1.52V and 105'c is the Intel recommended maximum but that's just the max they think you can use without very quickly destroying the chip, it would die within months at that! I wouldn't exceed 1.4Vcore and 85'c if you want the CPU to last a few years; if you need more voltage than that's just the nature of the silicon lottery and you'll have to settle for a lower clock speed to keep it stable. Sounds like could have pushed it too hard before and prematurely aged it (you didn't mention absolute temperatures, only relative?), are you using automatic overclock utilities?

Well, the average temperature is around 98 degrees. And, I manually overclock it.

[joemiller 269]

23 minutes ago, martin-w said:

 

 

 

 

Perfectly normal to have one core hotter than the rest. Also depends on the software you are running. It's usually around a 10 degree delta. The question to ask though is whether this core is being pushed harder by P3D? What are the temps like with your other software? Try it with a stress test. 

18 degrees is higher than I would have expected. Other than software, a dodgy IHS/die interface courtesy of Intel can do it, not a very proficient attempt at delidding, uneven cooler mounting etc. 

Try stressing it with something else and check temps, try remounting cooler. Unlikely that a void in the TIM would be the cause, as the pressure from mounting usually does a good job of evening out the TIM. Remounting cooler might help if it's not tightened down evenly.

If all else fails, and you perceive it as an issue, then it could be Intel at fault re their IHS mounting in which case a delidd would probably help. Worth also remembering that the silicone lottery doesn't just apply to the die, it can also apply per core. With some cores overclocking more or less, or cooler or hotter. 

What you are looking for is around 10 degrees delta while running a stress test. That would be within the norm. 

 

 

 

 

Stressed with Cinebench... Thinking about replacing the CPU, and the Corsair liquid cooler. 

[martin-w 3,234]

6 minutes ago, joemiller said:

Well, the average temperature is around 98 degrees. And, I manually overclock it.

 

Way too hot!!!!

It will be throttling at that temp!

Cores can throttle independently. 

Aim for max of 85 degrees. 

Edited December 12, 2018 by martin-w

[ckyliu 1,749]

@joemiller The chip will throttle if it hits 100'c, so you have potentially reduced performance running at 98'c. I feel you're cooking the processor, consensus seems to be that Coffee Lake should be kept below 95'c in all circumstances and ideally below 85'c although I don't know what they base that figure on. But we can be sure that cooler is better and running close to the throttle threshold is not good. There maybe settings you can disable in BIOS to get temperatures down a bit too, in addition to reduce voltage and improving cooling (I trust you're not overclocking with a standard Intel cooler!?)

Edited December 12, 2018 by ckyliu

[joemiller 269]

Just now, martin-w said:

 

Way too hot!!!!

 

It will be throttling at that temp!

For sure, that's why it crashes- BSOD. I'm going to replace the CPU and the Liquid cooler (this one is almost 2 years old) . 

-- What thermal paste would you recommemd? Oh, and the way I delid my processor.

[martin-w 3,234]

The CPU is probably not damaged just run it at the proper temp. Aim for an absolute max of 85 degrees. 

5.1Ghz with 136 volts is a good result. Don't expect miracles. The extra 200 MHz amounts to hardly anything, just a frame per second or thereabouts. 

Edited December 12, 2018 by martin-w

[joemiller 269]

2 minutes ago, ckyliu said:

@joemiller The chip will throttle if it hits 100'c, so you have potentially reduced performance running that hot. Personally I feel you're cooking the chip, consensus seems to be that Coffee Lake should be kept below 95'c in all circumstances and ideally below 85'c although I don't know what they base that figure on. But we can be sure that cooler is better and running close to the throttle 100'c threshold is not good.

Yeah, and I had delid the processor, but I'm surprised it has lost holding 5.3 with only 141V this quick. When you delid the CPU, you notice how small the little space where you apply thermal paste is. Wouldn't't it be better if Intel would make this space bigger to cover almost the entire size of the processor, thus transferring more heat to the cooler ?

[martin-w 3,234]

Quote

Wouldn't't it be better if Intel would make this space bigger to cover almost the entire size of the processor, thus transferring more heat to the cooler ?

 

Huh! The liquid metal is applied to the surface of the die. And the underneath of the IHS. 

[joemiller 269]

4 minutes ago, martin-w said:

The CPU is probably not damaged just run it at the proper temp. 

5.1Ghz with 136 volts is a good result. Don't expect miracles. the extra 200 MHz amounts to hardly anything, just a frame per second or thereabouts. 

Yep, for sure. I agree. 

--- Also worth mentioning, during taking off or landing at a high-end airport with a high-end a/c all cores max usage at 100%. Once it finish rendering  the terrain, traffic,  and airport, it lowers to around 40% usage.  

[Bob Scott 7,716]

In my experience, when running a stress test, it's pretty rare to see more than 5 deg difference between cores under full load.

I shoot for 80 deg C or below--during the stress test.  With P3D running, temps stay in the 50s.

Also, the fan curve on my Corsair H100i cooler (on a secondary PC running a 7700K) was not anywhere near aggressive enough...could be the pump/fans are not ramping up fast enough if you're using the default "quiet" fan curve.  I use the coolant temp rather than CPU temp, and go from minimum fan (500rpm) at 25 deg C to full speed at 30 deg C coolant temp.

Regards