[Paradigm Shifter]

You disabled thermal throttle, a feature designed to protect the CPU from overheating.

What are temperatures at stock?

What cooler are you using? Is the cooler mounted correctly?

Motherboard? PSU? RAM? Linux distro, kernel version?

The freeze was likely an unstable overclock hard locking the system.

What settings were you using? Over 1.4V and that may have damaged the CPU anyway, although I'm perhaps overly cautious regarding voltages and others will be able to chime in with more knowledge.

If the CPU was actually running at 110 for a prolonged length of time, rather than the sensor reading incorrectly by +27*C (which is a known bug with Ryzen CPUs on Linux still) I'd be pretty worried.

 

[MothTLOU2]

Temps are normal. Max 65-75c full load.

k10temp ---> 110c-117c, linux 6.1... just on 4.6ghz 1.45v prime95 which ofc couldnt pass but it kept running for a while 30min or less

Thermal throttle doesnt work with performance governor on kernel and also manual OC. So it would not matter

I went on windows11 too and tctl and tdie were same for r7 5700g.
How can cpu go 117c when tjmax is 95c.
I didnt check the ryzen master temp tho, just aida64

Yes the freeze was a sing of unstable OC but where is hardware shutdown?

Also what does a damaged cpu from heat looks like?

Thank you

 

[Paradigm Shifter]

Temps are normal. Max 65-75c full load.

And the cooler? What is the cooler? Is it a heatsink? An all-in-one watercooler? Is it warm to the touch? If an AIO, does it sound like the pump is running? Can you feel it vibrating?

75*C load on a 65W CPU sounds a bit high for any good cooler (and certainly means I would not have been overclocking it). Check paste/cooler contact.

k10temp ---> 110c-117c, linux 6.1... just on 4.6ghz 1.45v prime95 which ofc couldnt pass but it kept running for a while 30min or less

Not tried a kernel newer than 6.0; on my 3900X box is reports temperatures correctly, on a box at work (Threadripper 2950X) it reports +27*C.

Thermal throttle doesnt work with performance governor on kernel and also manual OC. So it would not matter

Very well. I've never had a system report getting that hot, so I don't know. I don't overclock my Linux boxes, because there I care about bulletproof stability and reliability.

I went on windows11 too and tctl and tdie were same for r7 5700g.
How can cpu go 117c when tjmax is 95c.
I didnt check the ryzen master temp tho, just aida64

If you're running Linux, Windows temperatures are not of interest. However, it definitely sounds like it is overreporting the temperaure and it's actually 90*C (which is still dangerously high for Ryzen).

Yes the freeze was a sing of unstable OC but where is hardware shutdown?

I had an old system which if it locked up in the "wrong" way, none of the board protection features would kick in. A different one (its replacement Z97) defaulted to overcurrent protection which was way too low, and the system would spontaneously reboot with no logs or warnings if I compiled gcc with all threads.

Also what does a damaged cpu from heat looks like?

Can manifest in different ways. Instability, refusal to boot, memory problems, data issues. If it's extreme you might be able to see physical damage, but I wouldn't expect that.

 

[MothTLOU2]

My current OC which is 4425 all cores draws like 150w in prime95 which i ran for 10h to ensure OC had no errors.

I just read that shutdowm temp is 115~125c so i think i didnt get any thermal damage as it didnt even reach the crit temp, let alone pass it

My temp during gcc compile are below 80c
Temps during kernel compile are 65c

I just had those crazy temps at 4.6ghz locked prime which was like 180w+ ... cooler has no matter there lol, prime is crazy

 

[Paradigm Shifter]

Prime is pretty close to a worst case scenario, sure, and it's a load which you will never see 99.99% of the time, but that nevertheless has utility.

Of course the cooler matters. If you've strapped a potato to it, it's not going to cool as well as a big dual-tower with 140mm fans from Thermalright/Noctua/etc.

 

[MothTLOU2]

Yes but i have kraken x63 and temps still go insane on prime thats why i said it

OC must pass prime for at least 6h to be stable else cache errors, other issues happen inside the cpu even if it passes daily stuff
U must find the correct V of that OC and prime is the only way

I was running cpu at 4.6ghz for 5months no issues..tried prime first time, couldnt pass 10sec.

Current 4.425ghz passed 10h of prime so its all good, i can maybe get another 50-100mhz but that would mean leaving prime for another 6h at 100c+ temps as i have to increase the V to 1.425, so i dont want to bother

 

[Blameless]

TJMax is when PROHOT is triggered. PROCHOT normally forces throttling, not a shutdown. Shutdown is triggered by reaching Thermtrip.

Thermtrip is in the 115-125C range, depending on part, and is a last resort to save the CPU if something catastrophic happens.

Don't disable thermal throttle.

 

[Nighthog]

There is a setting to set thermal throttle to a custom value, no need to disable the setting entirely.

 

[Desolutional]

There is a setting to set thermal throttle to a custom value, no need to disable the setting entirely.

And unless you are using some sort of extreme high end cooling where you want to set a world record no matter what the cost to the CPU, this setting is best left at default like the manufacturer intended.