[SOLVED] I9 9900K Prime 95 reaching 100C

Selynelar

Prominent
Feb 19, 2019
51
1
545
1
Hi all,

I tested today my cpu regarding that yesterday started the pc freeze while i was gaming. When I started the prime 95 everything was fine the max temp was 80-85 C degrees. After few minutes I left the computer and when I came back i saw that some of the cores were running on 100c degrees. I stoped it immedietaly because I was afraid that i kill my cpu. The system is not OC-d. Everything is on auto.

Here is my system spec:

Gigabyte z390 Aorus Master
I9 9900K (not oc-d)
Noctua NH-D15
Kingstony HyperX 2666mhz 32GB
MSI RTX 2080 TI Gaming X Trio
Phanteks Enthoo Pro M SE case additional 3 x 14cm Pwm fans
BeQuiet Straight Power PSU 800W

Is it possible that it damaged my cpu that it was running for few minutes on 100c degrees? The system still is stable.
And I was running after the aida64 because on many forums i have read that it is more trustable than Prime 95.
These are the results with Aida64 during stability test.

http://www.kepfeltoltes.eu/view.php?filename=956CPUID_HWMonitor_2020._.png

Thanks for your replies.
 

CompuTronix

Intel Master
Moderator
Selynelar,

I think you've reached a premature conclusion. Although some elements of the answer that ChumP provided for you are correct, respectfully, neither of you are very well informed concerning Prime95 and AIDA64.

For the record, there's nothing wrong with Prime95.

The issue is most users don't know that to perform a proper thermal test which conforms to Intel's Datasheets, Prime95's latest version 29.8 must be run using only Small FFTs with all AVX test selections disabled. As per Intel’s Datasheets, TDP and Thermal Specifications are validated “without AVX.



When Prime95 is run in this mode, it's a steady-state 100% TDP workload. However, if the AVX test selections remain enabled, then just as ChumP correctly pointed out, "Prime95 puts an unrealistic load on your CPU", which is a brutal 130% workload. This is a known problem of which, unfortunately, many users remain unaware. It's a critical detail that makes a huge difference.

Concerning AIDA64, the problem is that it has 4 CPU related Stress Test selections (CPU, FPU, Cache, Memory) which have 15 possible combinations that yield 15 different workloads and Core temperatures. As with Prime95, most users again are unaware of the implications, and don't mention which test(s) to run, or explain how they should be configured and why.

As you can see from the lower left side of the scale below, AIDA64's individual CPU test is a very weak workload at only 70%. The CPU/FPU test combination is about 90% workload, while the individual FPU test is about 115% workload. Only Prime95 Small FFTs (No AVX) provides a steady 100% workload. As an alternative, OCCT - Small Dataset - No AVX is nearly identical at about 97% workload.

“Stress” tests vary widely and can be characterized into two categories; stability tests which are fluctuating workloads, and thermal tests which are steady workloads. Games, apps, streaming, rendering, transcoding and most utilities have partial, fluctuating workloads with fluctuating Core temperatures that are not well suited for testing thermal performance.

Utilities that don't overload or underload your processor will give you a valid thermal baseline. Here’s a comparison of utilities grouped as thermal and stability tests according to % of TDP, averaged across six processor Generations at stock settings rounded to the nearest 5%:



Although these tests range from 70% to 130% TDP workload, Windows Task Manager interprets every test as 100% CPU Utilization, which is processor resource activity, not actual workload. Core temperatures respond directly to Power consumption (Watts), which is driven by workload.

We have a Guide that covers this topic. It's a "Sticky" which is located at the top of the CPUs Forum where you posted this Thread. If you look up there at the Stickies you can't miss it. All the answers you seek are already in the Intel Temperature Guide.

We encourage our Members to check the Stickies before posting Threads, as they frequently contain the information you need. This can save you time needlessly searching for answers, or waiting for others to responds with perhaps less than "well informed" answers.

Give the Guide a read. Pay especially close attention to Section 10 - Thermal Test Basics and Section 11 - Thermal Test 100% Workload.

CT:sol:
 
No... you didn't damage the CPU. Even though that's Tjmax for the 9900k, when it reaches 100C it will instantly downclock itself to prevent overheating.

Also Prime95 puts an unrealistic load on your CPU, so I wouldn't be too concerned about it. Aida is enough for stress testing.

Temps in aida64 seem ok for the most part, max temps on some cores are a bit higher than normal. You could try manually overriding VCore in the BIOS and set it to 1.2V from auto(you'll need to test the stability of the system through aida64 after manually setting the VCore, 30 mins should be enough) ... the majority of 9900ks will run at this voltage with no trouble on stock clock speeds(you'll most likely see a reduction in temps). If it's stable at 1.2V you can try shaving even more... you can do it in steps - 0.05V each time until you're noticing the first signs of instability in aida64. When it becomes unstable you just revert to the VCore of the previous successful run.
 
Last edited:

Selynelar

Prominent
Feb 19, 2019
51
1
545
1
No... you didn't damage the CPU. Even though that's Tjmax for the 9900k, when it reaches 100C it will instantly downclock itself to prevent overheating.

Also Prime95 puts an unrealistic load on your CPU, so I wouldn't be too concerned about it. Aida is enough for stress testing.

Temps in aida64 seem ok for the most part, max temps on some cores are a bit higher than normal. You could try manually overriding VCore in the BIOS and set it to 1.2V from auto(you'll need to test the stability of the system through aida64 after manually setting the VCore, 30 mins should be enough) ... the majority of 9900ks will run at this voltage with no trouble on stock clock speeds(you'll most likely see a reduction in temps). If it's stable at 1.2V you can try shaving even more... you can do it in steps - 0.05V each time until you're noticing the first signs of instability in aida64. When it becomes unstable you just revert to the VCore of the previous successful run.
Thank You so much for your help. Now I can be sure that there is nothing to worry about it.
Thanks again.
 

CompuTronix

Intel Master
Moderator
Selynelar,

I think you've reached a premature conclusion. Although some elements of the answer that ChumP provided for you are correct, respectfully, neither of you are very well informed concerning Prime95 and AIDA64.

For the record, there's nothing wrong with Prime95.

The issue is most users don't know that to perform a proper thermal test which conforms to Intel's Datasheets, Prime95's latest version 29.8 must be run using only Small FFTs with all AVX test selections disabled. As per Intel’s Datasheets, TDP and Thermal Specifications are validated “without AVX.



When Prime95 is run in this mode, it's a steady-state 100% TDP workload. However, if the AVX test selections remain enabled, then just as ChumP correctly pointed out, "Prime95 puts an unrealistic load on your CPU", which is a brutal 130% workload. This is a known problem of which, unfortunately, many users remain unaware. It's a critical detail that makes a huge difference.

Concerning AIDA64, the problem is that it has 4 CPU related Stress Test selections (CPU, FPU, Cache, Memory) which have 15 possible combinations that yield 15 different workloads and Core temperatures. As with Prime95, most users again are unaware of the implications, and don't mention which test(s) to run, or explain how they should be configured and why.

As you can see from the lower left side of the scale below, AIDA64's individual CPU test is a very weak workload at only 70%. The CPU/FPU test combination is about 90% workload, while the individual FPU test is about 115% workload. Only Prime95 Small FFTs (No AVX) provides a steady 100% workload. As an alternative, OCCT - Small Dataset - No AVX is nearly identical at about 97% workload.

“Stress” tests vary widely and can be characterized into two categories; stability tests which are fluctuating workloads, and thermal tests which are steady workloads. Games, apps, streaming, rendering, transcoding and most utilities have partial, fluctuating workloads with fluctuating Core temperatures that are not well suited for testing thermal performance.

Utilities that don't overload or underload your processor will give you a valid thermal baseline. Here’s a comparison of utilities grouped as thermal and stability tests according to % of TDP, averaged across six processor Generations at stock settings rounded to the nearest 5%:



Although these tests range from 70% to 130% TDP workload, Windows Task Manager interprets every test as 100% CPU Utilization, which is processor resource activity, not actual workload. Core temperatures respond directly to Power consumption (Watts), which is driven by workload.

We have a Guide that covers this topic. It's a "Sticky" which is located at the top of the CPUs Forum where you posted this Thread. If you look up there at the Stickies you can't miss it. All the answers you seek are already in the Intel Temperature Guide.

We encourage our Members to check the Stickies before posting Threads, as they frequently contain the information you need. This can save you time needlessly searching for answers, or waiting for others to responds with perhaps less than "well informed" answers.

Give the Guide a read. Pay especially close attention to Section 10 - Thermal Test Basics and Section 11 - Thermal Test 100% Workload.

CT:sol:
 

ASK THE COMMUNITY

TRENDING THREADS