[SOLVED] CPU temps quickly reaching 100 degrees celc. when stress testing

[greengear]

Just got my PC built yesterday. Tried stress testing today using Prime 95 using these settings, and the CPU temps quickly reach 100 deg C even when the pump/cooler fan is at full speed. When I cancel the test, it reduces back to 30 degrees within a couple of seconds. Weirdly, CPU temps remain sub 35 deg at idle even at 25% of cooler max speed. What may be the issue?

Getting 85 deg temp when stressing 4 cores with no hyperthreading.

Relevant info:
CPU: Intel i7 12700KF
Overclock: No
Cooler: silverstone pf240 (AIO)
MotherBoard: MSI PRO Z690-A DDR4

 

[DimkaTsv]

When I cancel the test, it reduces back to 30 degrees within a couple of seconds. Wierdly, CPU temps remain sub 35 deg at idle even at 25% of cooler max speed.

This part seems normal, as effective clock of modern CPU raise up instantly as they get load. Otherwise they stay quiet and chill.

Tried stress testing today using Prime 95, and the CPU temps quickly reach 100 deg C even when the pump/cooler fan is at full speed.

And for that point... well your cooler system may not be able to handle full load of this processor...
Just for fonzies... try to put load only on 4 cores. i wonder if it will instantly thermal throttle. Or don't, as better be safe than sorry.

 

[Phaaze88]

1)Prime 95 settings matter. You can't just say you ran Prime 95.
2)Do not run auto Vcore settings.
3)Probably better off just leaving Prime 95 alone.

 

[greengear]

1)Prime 95 settings matter. You can't just say you ran Prime 95.
2)Do not run auto Vcore settings.

Fair enough. These are the settings I used. I'm not sure what auto Vcore is but I'll look into that.

3)Probably better off just leaving Prime 95 alone.

What would you say is the most reliable way to stress test my CPU?

 

[Phaaze88]

Well you did the settings right.
Auto Vcore is leaving the voltage going to the cpu on auto. That's a no-no when doing heavy benchmarks on a cpu like that. The motherboard supplies more Vcore than the cpu needs by default, as vendors can't predict silicon lottery, so the curves are designed around the worst bins to guarantee advertised clocks and performance.

What would you say is the most reliable way to stress test my CPU?

Depends on what you use it for. There is no best otherwise.
In games: the games themselves, since they don't all put the cpu through the same paces.
Everything else: Cinebench R23 or Blender.

 

[geofelt]

Opinions vary on stress testing.
Some use instructions explicitly designed to generate heat.
Some use AVX instructions which are not common in gaming.

I use the simple CPU-Z bench stress test.
I have the same motherboard and a 12900K.
My cooler is a noctua NH-D15s.
It has a similar cooling capability as your 240 aio.

When I start the stress test, in two seconds, all threads jump to 95c. Shortly after, some 4 threads will show 100c. using hwmonitor. Clearly, there is some throttling going on.
The good news is that the processor keeps on trucking.
I am not alarmed because my workload never loads all possible threads.
I think the motherboard has implemented, by default, some agressive cpu management options. One day, I might look more into those options. No matter, these processors are so stupidly fast.
Originally, there was a 115 beta bios which I installed.
It was subsequently withdrawn.
I suspect that there might have been some windows 11 issues that needed fixing.

 

[DimkaTsv]

What would you say is the most reliable way to stress test my CPU?

Depends on what are you going to test ofc.
Personally use single thread prime95, 4 thread prime95, 12 thread 95.
But i don't have any thermal problems, and test my undervolting stability usually. And 4 thread Prime95 is omega fast to crash a thread on even 0.012V misset. Actually fastest way to crash.
To test thermals - 4 thread Prime95, Linpack and y-cruncher iirc
To test game related stuff (FPS, etc.... Well ofc games.
Memory controller... I may be wrong there, but y-cruncher or SuperPi, because memory fail also will cause these tests to fail.
But that memory fail must be REALLY obvious.

 

[greengear]

Just for fonzies... try to put load only on 4 cores. i wonder if it will instantly thermal throttle. Or don't, as better be safe than sorry.

Ok so I tried this. temps go up to 85. It takes my cooler like a 25 second delay to start ramping up for some reason. But even after that, temperature does not reduce in the slightest. Might be the cooler.

 

[greengear]

My cooler is a noctua NH-D15s.
It has a similar cooling capability as your 240 aio.

The only reason I didn't get a Noctua NH-D15s is because the store (or any place online) didn't have mounting support yet for the LGA1700. I'm thinking of returning my 240 and upgrade to the Noctua if the cooler really is the issue.

Originally, there was a 115 beta bios which I installed.
It was subsequently withdrawn.
I suspect that there might have been some windows 11 issues that needed fixing.

Do you think this problem is arising due to a BIOS issue?

 

[geofelt]

The only reason I didn't get a Noctua NH-D15s is because the store (or any place online) didn't have mounting support yet for the LGA1700. I'm thinking of returning my 240 and upgrade to the Noctua if the cooler really is the issue.


Do you think this problem is arising due to a BIOS issue?

No, I do not think bios is the issue.
Possibly, the default bios settings are boosting the performance excessively. Check the msi forums for the motherboard, of course, you should expect some issues.
To date, I see no new bios updates, and so I will stick with 115 since all is going well.

New noctua stock will in time, include lga1700 mounting.
The chromax black units apparently do.
If you have a nh-d15 cooler, noctua will mail you a lga 1700 mounting kit for free. I did that and was much impressed with noctua service.
You fill out a form and show documentation of the cooler you bought and the lga1700 motherboard. There are actually two different kits.
mp83 for nh-d15 types, and mp 78 for NH-U12s types.
I ordered both, the 78 to test with a older cooler and the 83 for the D15s. The difference is the height needed to get the proper hold down pressure.

What is the make/model of your case, and what graphics card is inside?
Any cooler needs a good source of fresh air to let it do it's job.
See if you do better with the case side cover off and directing a house fan at the innards.

 

[DimkaTsv]

Ok so I tried this. temps go up to 85.

Oh wow, maybe Intel have something really different in boosting algorithm that set up frequencies depending on type of load.
I never saw my 5600x have less temperature on 4 thread load compared to max thread one.

It takes my cooler like a 25 second delay to start ramping up for some reason.

Also, how is that? Temp instantly go to 85 degrees, but cooler ramps up for 25 seconds? That definitely should not work like that. But maybe AIO have something different compared to air coolers, though, i don't know.

 

[Arbër1041]

Fair enough. These are the settings I used. I'm not sure what auto Vcore is but I'll look into that.


What would you say is the most reliable way to stress test my CPU?

Intel burn test