My CPU temp has rather suddenly increased by about 15-20% and this does not have too much to do w/ the summertime we're in here.  The temperature is not considered high, there is no throttling and well shouldn't be, but still, about 15-20% higher than previous.  I've checked fan speeds, case and component cleanliness etc, and all looks great.  I 'm down to wondering if, despite 3y of predicable stable temps, this sudden change might have to do w/ the thermal paste I used back when I assembled the parts.  It would be especially useful if you did this for similar reasons, replaced the thermal paste, and how it turned out.

Thanks in advance!

TM can dry out, but I think it would be odd to suddenly get a big jump in temps vs a more gradual rise.  I suppose if it dries enough to crack, it's possible.  I'd be more suspicious of a loose heatsink.

Are you watching temps on all the cores?  Looking for a hotspot might lend a clue.  Also, I'd want to rule out a new background process running that's putting a new load on the CPU.

To replace, remove the heat sink, use an old credit card or similar with a nice plastic straight edge to scrape the old TM off both the CPU and the heatsink, then clean both with a section of paper coffee filter soaked with 90% isopropyl alcohol (filters don't leave fibers behind like a paper towel).  Reapply fresh TIM and reassemble.

 

My default surveillance is via RTSS OSD looking at its entry for the whole package temp I think.  When I look at individual core temps it's generally in line with the package temp though they fluctuate a little more per core.  Not seeing any other background processes in play using TM.  I think I will go ahead and R&R the heat sink and TM and see if it helps.  Thanks for the tips I'll follow them.

Cheers

16 hours ago, Noel said:

My CPU temp has rather suddenly increased by about 15-20% and this does not have too much to do w/ the summertime we're in here. 

 

Have you made any changes to settings? XMP off drops my temps by 10 degrees. If you had XMP off and now it's on, that would do it. Or HT off and now you have it on. 

1 hour ago, martin-w said:

 

Have you made any changes to settings? XMP off drops my temps by 10 degrees. If you had XMP off and now it's on, that would do it. Or HT off and now you have it on. 

Thanks Martin, nope, no changes save Windows 10 updates.  I didn't mention this in the OP but just prior to this I had an overclocking failure crash, first time pretty much ever since I assembled the box 3y ago. Okay, I concluded the CPU at 5gHz/1.215v has accumulated some wear and tear, so moved vCore up from 1.215 to....1.23v, and left it at 5Ghz.  Used Intel Ext Tuning Utility and ran that for 20 minutes and all appeared well.   Last night after 1.5h of flight, another overclock failure manifested.   That's more voltage I know, but not much.  It's weird because I saw brief temps up to 78C, which I've never seen before.  Highest was around 64C during summer time temps in past summers.  So I went back to [email protected] and so far all's well about an hour into my first flight at that setting.  Looking at individual core temps Core 2 and 3 show the highest peak temps since taking off at 67C, which again is higher than previously, and yet utilization on those cores is much lower than the main thread Core 7 which has a high temp so far of 63C.  I think i will go ahead the reset the IHS and see if it helps.  Maybe there is a section that isn't interfaced well.

1 hour ago, Noel said:

Thanks Martin, nope, no changes save Windows 10 updates.  I didn't mention this in the OP but just prior to this I had an overclocking failure crash, first time pretty much ever since I assembled the box 3y ago. Okay, I concluded the CPU at 5gHz/1.215v has accumulated some wear and tear, so moved vCore up from 1.215 to....1.23v, and left it at 5Ghz.  Used Intel Ext Tuning Utility and ran that for 20 minutes and all appeared well.   Last night after 1.5h of flight, another overclock failure manifested.   That's more voltage I know, but not much.  It's weird because I saw brief temps up to 78C, which I've never seen before.  Highest was around 64C during summer time temps in past summers.  So I went back to [email protected] and so far all's well about an hour into my first flight at that setting.  Looking at individual core temps Core 2 and 3 show the highest peak temps since taking off at 67C, which again is higher than previously, and yet utilization on those cores is much lower than the main thread Core 7 which has a high temp so far of 63C.  I think i will go ahead the reset the IHS and see if it helps.  Maybe there is a section that isn't interfaced well.

 

What make of paste are you using? Some pastes do dry out over time, wouldn't expect it to be a sudden increase in temp though, unless you just didn't notice it creeping up.

None of your temps are excessive, 78 degrees is no big deal at all, your TJ Max is 100C I recall for your chip, so you have plenty of margin.

Have a bash with a quality paste, like Thermal Grizzly Kryonaut. No cure time and long lasting.

In  any case, if its just brief spikes to 80 or so it's no big deal, and in fact quite normal.  

I'm not overlocking at all with my 12900K, just no point anymore with the 12900K to be honest, CPU's are pushed pretty much to the limit these days.

Are you using any kind of stress test? If you use Cinebench R23, then expect 90 degrees, its hammers CPU's. 

 

1 hour ago, Noel said:

Thanks Martin, nope, no changes save Windows 10 updates.  I didn't mention this in the OP but just prior to this I had an overclocking failure crash, first time pretty much ever since I assembled the box 3y ago. Okay, I concluded the CPU at 5gHz/1.215v has accumulated some wear and tear, so moved vCore up from 1.215 to....1.23v, and left it at 5Ghz.  Used Intel Ext Tuning Utility and ran that for 20 minutes and all appeared well.   Last night after 1.5h of flight, another overclock failure manifested.   That's more voltage I know, but not much.  It's weird because I saw brief temps up to 78C, which I've never seen before.  Highest was around 64C during summer time temps in past summers.  So I went back to [email protected] and so far all's well about an hour into my first flight at that setting.  Looking at individual core temps Core 2 and 3 show the highest peak temps since taking off at 67C, which again is higher than previously, and yet utilization on those cores is much lower than the main thread Core 7 which has a high temp so far of 63C.  I think i will go ahead the reset the IHS and see if it helps.  Maybe there is a section that isn't interfaced well.

Well the addition of crashes and voltage increases to the scenerio changes the discussion a lot!

I'm curious about what led you conclude that it's an overclocking failure.  Did you run a check on your RAM?  In my experience, memory issues are a far more common cause of system instability than CPU degradation.  I'd also want to watch the voltages on the RAM and the CPU to see if they're sagging excessively or otherwise abnormal, which could indicate a developing problem with your PSU or motherboard's voltage regulation.

That 15 mv voltage increase, though not large, very likely does explain the moderate temp increases you're seeing, which aren't really extreme or troublesome.  When you get out to the far end of the voltage curve, heat produced as a function of voltage increases in a decidedly non-linear fashion, and that's complicated by the fact that dissipation of any additional heat is limited by the small surface area of the die and the chip's IHS.

 

4 hours ago, martin-w said:

 

What make of paste are you using? Some pastes do dry out over time, wouldn't expect it to be a sudden increase in temp though, unless you just didn't notice it creeping up.

None of your temps are excessive, 78 degrees is no big deal at all, your TJ Max is 100C I recall for your chip, so you have plenty of margin.

Have a bash with a quality paste, like Thermal Grizzly Kryonaut. No cure time and long lasting.

In  any case, if its just brief spikes to 80 or so it's no big deal, and in fact quite normal.  

I'm not overlocking at all with my 12900K, just no point anymore with the 12900K to be honest, CPU's are pushed pretty much to the limit these days.

Are you using any kind of stress test? If you use Cinebench R23, then expect 90 degrees, its hammers CPU's. 

 

As mentioned in the OP, I know the temps are well within range for no thermal throttling.  It's the abrupt change that caught my attention.

I went ahead and R&R the paste this time using Corsair XTM70 which is rated the same in terms of thermal conductivity at 11.2W/m2.  It's definitely better, and it is also core-core quite consistent and stable.  Whereas package temp of 54C would jump to 78C and now the highest any core has been is 67C, and 5 of 8 cores are exactly that.  I had much more variability between cores previously.  So I think it's helped and restored the prior expectation.  I noticed when I removed the heat sink from the CPU the center area was rather uncovered w/ paste, so maybe the initial install wasn't the best.  But thermal performance has always been so who knows. 

I think it's also possible as Bob brought up physical connection of the heat sink on the CPU might have become compromised, or slightly loose or detached intermittently, which maybe comports with the weird random spikes in temperature that I'm not getting now.  It's been 47-53C during this cruise at FL240 so all good again.

Cheers

4 hours ago, Bob Scott said:

Well the addition of crashes and voltage increases to the scenerio changes the discussion a lot!

I'm curious about what led you conclude that it's an overclocking failure.  Did you run a check on your RAM?  In my experience, memory issues are a far more common cause of system instability than CPU degradation.  I'd also want to watch the voltages on the RAM and the CPU to see if they're sagging excessively or otherwise abnormal, which could indicate a developing problem with your PSU or motherboard's voltage regulation.

That 15 mv voltage increase, though not large, very likely does explain the moderate temp increases you're seeing, which aren't really extreme or troublesome.  When you get out to the far end of the voltage curve, heat produced as a function of voltage increases in a decidedly non-linear fashion, and that's complicated by the fact that dissipation of any additional heat is limited by the small surface area of the die and the chip's IHS.

 

It was the big light blue Windows 10 indicating a problem occurred or something I forget.  As it goes thru its initialization, it then reboots to the DOS type text stating overclocking had failed, and do you want to reboot to the last successful yada yada.  This happed several times until I went down to 4.9Ghz and that was at 1.22v, up from 1.215v which ran the CPU at 5.0Ghz all cores for 3 years now.  And that kind of voltage on this CPU is low, that is for sustained 5.0Ghz.  I'm hoping the quality of the connection between heat sink and CPU was slightly compromised as you hinted and new paste application is hopefully better.  It's definitely improved now so was worth the work.  I will run it back at 1.23 and see how it performs heat-wise.  That's still a very low voltage for this CPU from what I read, and my thermal performance on air cooling has been great.

I will check RAM thanks for the reminder.  

I checked RAM using the built-in pre-login to Win 10 memory checker and it passed. I repeated it in Intel Extreme Tuning Utility and it passed there as wel.

I had a couple of uneventful flights at first 5.0Ghz, then a hard crash again.  Went to 4.9Ghz at 1.21v, two flights went well, then another hard crash.  So right now am at 4.9Ghz at 1.22v about halfway thru a 2 hour flight.  We'll see.

I guess this could be the PSU maybe?  Or something in the motherboard?  Or of course Occam's the CPU has degraded slowly with 3y at 1.195v at 5Ghz.  This is to be expected at some level, at least according to the Data Sheets on a couple of Intel CPUs I've used over the years.  I have a PSU tester so I can try that next.

If this is a harbinger of more imminent failures I need to start planning for parts for a CPU upgrade.  I can see water cooling just to get rid of the space occupying fact that are the very effective Noctua coolers.  I understand they still compare well to AIO solutions.

What is a decent AIO brand  should look at that will support 13900K?  Or the top AMD offering?  Also, can the radiators be mounted outside the case? 

Edited July 8, 20232 yr by Noel