quick ? on cpu thermals

[JZSmith]

so I was kind of bored and decided to over clock my i9-10850k, I used the ASUS AI overclock and just adjusting the cooling score to get better MHz on the clock. normally at load my CPU runs at 66C but after running the overclock I was hitting 90C max, no matter how high I went in upping the speed I always hits 90C, MOBO must be locking it at 90C but then I decided to turn off and I accidently turned the AI mode to all cores sync thinking this was default and went and did cinebench and was hitting 100C max ran that for 3 min till I stopped realizing this is not stock. went in to bios and set everything to default and back to normal operations. my question is how much damage might I have done to the CPU, I ran cinebench probably 9 times adjusting the speed hitting 90C each test is 15mins, I just got this 4 days ago and dont want this to go bad, also when boosting the clock speed I went all max 5.4ghz on all cores but I believe this CPU only goes up to 5.2, when I ran the benchmark on 5.4ghz I made it though the first pass then computer froze no BSOD I had to manually reset on the pc case is this something that could have caused damage, going over the rated clock speed.  and if anyone is wondering i am using a corsair h100i AIO in a push and pull config

[Airbus 123]

ammm i think it will probably have quite a major damage....as you have ran it to up to 100C.

 

but i’m not an expert

[Airbus 123]

also having 100C on a AIO water cooling seems abnormal........

[Mister Woof]

Automatic overclocking is a bad way to do it - you probably have enough voltage and current going through that chip to raise Frankenstein's monster.

 

As far as damage, probably nothing serious as these chips will automatically throttle when too hot, although if the voltage and current was that high it may have taken a few days off its lifespan.

 

Check with Hwinfo64 how much VCORE (not VID) you are running while in use at your present settings. Anything over 1.3v or so DURING STRESS (vdrooped) will be very hard to keep cool on most coolers.

 

My recommendation is in BIOS, to set it at 1.3v fixed voltage, a medium LLC setting, 50x core and 47x cache multiplier, and turn off all EIST/P states. Then set your power limits to maximum and disable TVB throttling.

 

This sounds worse than stock, but it isn't. The chip won't ever really see 5.2ghz, and likely will be at around 4.7 (or 4.8ghz with TVB under 70c) at stock, so a fixed all-core 5ghz will still be an improvement 99% of the time.

 

Run a few stress tests to see if that works - I recommend Cinebench for a quick pretest and if it passes, run 30 minutes of ASUS Realbench. If it fails Cinebench or throws WHEA errors, increase voltage by about .005mv until it passes both tests or hit about 1.35v. if it doesn't, drop down frequency, apply a -1 or 2 AVX offset, or raise LLC (or some combination of these) u til you find your preliminary stability.

 

Then when you're feeling good let Realbench run for an hour or so.

 

Alternatively, you could play with a fixed or offset voltage + a modified active core turbo table so as not to lose that single core boost, but that takes more time to tune and get right.

 

 

[tikker]

7 minutes ago, JZSmith said:

when I ran the benchmark on 5.4ghz I made it though the first pass then computer froze no BSOD I had to manually reset on the pc case is this something that could have caused damage, going over the rated clock speed.

This just means your overclock wasn't stable. Lower the clocks and test again.

 

90C seems okayish, a little on the high side, but 100C is hot and it will likely have started thermal throttling. It's not a good idea to run that 24/7. Check your cooler mount if you mounted it yourself to see if that is solid and as said try OCing yourself instead of through automatic settings.