[SOLVED] 12700K at 100C after few minutes of Prime95

BlackAzrael

Distinguished
Jun 21, 2011
108
0
18,690
12700K at 100C after few minutes of Prime95

Hello everyone,
Recently just got a CPU + Motherboard upgrade and got a Noctua NH-D15 CPU cooler to go with it.

Case Specs: http://nordic.thermaltake.com/products-model.aspx?id=C_00001719
Ran Prime95 and noticed the CPU goes to 100C within 3-5 minutes. I stopped the test immediately as it didn't seem normal. Now I'm kind of wondering if that's normal and checking if I should revise my fan setup for my case.
My current fans:
Noctua NF-P14s redux 1500rpm = 78.69 CFM (BACK EXHAUST)
Noctua NF-A15 x 2 (CPU Cooler) = 67.98 CFM (EXHAUST directed towards back 140mm exhaust fan)
Noctua NF-A20 = 86.46 CFM (SIDE INTAKE)
Noctua NF-A20 = 86.46 CFM (TOP EXHAUST)
Noctua NF-F12 = 54.97 CFM (BOTTOM INTAKE just under the GPU and besides PSU)
Thermaltake Pure 20 TT-2030 = 129 CFM (FRONT INTAKE, kind of doubtful if that's the true CFM)

Anyone have any suggestions on what to change or swap around?


Current system specs:
CPU: Intel 12700k
MOBO: MSI PRO-Z690A DDR4 Wifi
RAM: Corsair Vengeance LPX DDR4 2x8GB
GPU: Gigabyte 1080ti Aorus Xtreme
PSU: Thermaltake M850W
Case: Thermaltake Chaser MK-1
 
Solution
1)Prime 95 settings matter. The load can be moderate to ridiculously stupid high, depending on what was set... or wasn't set.
The go-to settings are:
-Small FFT.
-AVX, AVX2, and AVX-512 disabled. Ignore the 'not recommended' bit next to Run a Weaker Torture Test.
-15mins, should be long enough for any cooler to soak.

2)Don't benchmark Prime 95 on auto Vcore and high/very high load line calibration settings. That's asking for a bad time with it.
Play around with Cinebench R23 in your free time to aid you in finding a lower voltage offset and maybe go back to Prime 95 later.

3)Disable overclocks.
Dragon Center likes to run an overclock in the name of 'performance optimization'... without even telling you. Uninstall it, if present... or...

Phaaze88

Titan
Ambassador
1)Prime 95 settings matter. The load can be moderate to ridiculously stupid high, depending on what was set... or wasn't set.
The go-to settings are:
-Small FFT.
-AVX, AVX2, and AVX-512 disabled. Ignore the 'not recommended' bit next to Run a Weaker Torture Test.
-15mins, should be long enough for any cooler to soak.

2)Don't benchmark Prime 95 on auto Vcore and high/very high load line calibration settings. That's asking for a bad time with it.
Play around with Cinebench R23 in your free time to aid you in finding a lower voltage offset and maybe go back to Prime 95 later.

3)Disable overclocks.
Dragon Center likes to run an overclock in the name of 'performance optimization'... without even telling you. Uninstall it, if present... or don't; maybe you like DC for some reason.
Enhanced Turbo - a general term, since they don't all call it the same thing - is an E-Z OC, but uses too much Vcore by default. Some Z series boards will have it enabled by default. Disable it in bios if it is. I do not know what Msi's version of enhanced turbo is called.




Thermaltake Pure 20 TT-2030 = 129 CFM (FRONT INTAKE, kind of doubtful if that's the true CFM)
That's likely the fan's true rated CFM... at 100% speed, with no obstacles present at all. The catch with 200mm fans is that they have low static pressure, so their performance falls harder than 120/140mm the greater the air resistance is.
With mesh panels and not much else in the way, they can rival the smaller fans. Not many options available in the market though.
 
  • Like
Reactions: CompuTronix
Solution
12700K at 100C after few minutes of Prime95

Anyone have any suggestions on what to change or swap around?

Use a more realistic stress tester.

Prime95 is an extrema algorithm when using any form of AVX and will thrash your CPU unrealistically. Workers will stop without you knowing why.

AIDA64 is a not so intense an Algorithm for thermal testing. AIDA64 has decent graphs to view for rail voltages and temperature so you can monitor and stop the test if temps get excessive. You can also test your RAM better, test your GPU and system storage drives. All in all a more modern method and updated regularly.

There is also Realbench a real world tester for ultimate stability.
 

TRENDING THREADS