Prime95 - stress tests

harry.wesol161e02f8

harry.wesoly
Joined
Nov 2, 2023
Messages
55
Is using Prime95 to test the stability of a computer system a good solution? In my own example, I've noticed that this test can be very, very computer-intensive. My previous settings in tests like cinebench r23 and r24 showed a temperature of 78 degrees. I also used the Aida64 and RealBench test and the temperatures there were also much lower than in Prime95. After turning on Prime95, the processor was able to reach 100+ after just a few seconds. Of course, there was also a blue screen with an error code IRQL_NOT_LESS_OR_EQUAL. Until I did the Prime95 test (time 10-20 minutes because there was an error) I wouldn't even think that the system could be unstable. To sum up, it seems to me that using Prime95 doesn't make any sense. In my opinion, the test conditions it offers to the computer are significantly different from everyday use. Strangely enough, other tests don't see conflicts. What do you think? Do you use this tool?
 
Prime95 is made to literally stress it to the absolute max, beyond what any normal processes would ever cause it to do.
If you're using the PC for just gaming, etc.....I wouldn't worry about it. If you're using it for other intensive tasks like video editing and/or decoding/encoding, then maybe it's worthwhile.
Cinebench is a good test of what gaming will generally do to your PC, on the upper end of torturing it. But even that is still rather limited, since most games still don't use the full CPU at this point.

Personally, I've always felt Prime95 was pointless. Some like to state it verifies stability, but I tend to think it just verifies that under abnormal/special conditions, it's stable. Which most user won't ever get to with their use cases.
 
For me, the PC should be able to be Prime95-stable for a reasonable amount of time, let's say 20 minutes. Then you know that even in the most averse conditions, you will not get any miscalculations. Some people let Prime95 run overnight, that is overkill. The most i will do is probably an hour or so. That means it's pretty much rock stable. Because yes, this is an artificial test. But nonetheless, i wouldn't want it to fail there almost immediately, or even throw a BSOD. That means something isn't dialed in properly yet.

If your cooling cannot handle full theoretical load like Prime95, you should first of all set power limits to limit the power draw and temperatures to what your cooling can handle, and you should make sure it's actually at its best capabilities (going to full fan speed at 85°C for example). Then you don't have to worry about overburdening your cooling and running into thermal throttling.

So, yes, i use Prime95, because i undervolt my CPUs using CPU Lite Load, and this is a good tool (especially Small FFTs test) to see when it becomes unstable from it, so i know i have to back off with my tuning and go one or two steps higher again.
 
Thank you for your valuable information. In my case, on the current settings, i.e. offset - by cpu 0.100 caused the computer to give BSoD after about 3 minutes in the Prime95 test. Inspired by the advice of my colleague CITAY, I started playing around with the settings in the bios related to CPU Lite Load. I've given up on the manual vcore voltage offset. Instead, I set the Cpu Lite Load to Mode 6. As a result, the computer was stable for 21 minutes of the Prime95 test, there was no reset and no BSoD. Maximum CPU Temperature in Prime95 was 96 degrees and the maximum current consumption was 311 W :-D Lower down in mode 5 there were already problems. In addition, I also disabled the IA CEP support option. In cinebench r23 max temperature 83. I guess I'll stop there, unless I still have to set up a TDP muzzle? Thank you for your advice.
 

Attachments

  • Zrzut ekranu (98).png
    Zrzut ekranu (98).png
    1.2 MB · Views: 41
  • Zrzut ekranu (103).png
    Zrzut ekranu (103).png
    1.2 MB · Views: 42
Yes, you could set power limits, because 311W is way too much to get rid of fast enough from that tiny surface for most coolers. It would help if HWinfo was set to English, but i can see enough. If you set the power limits to 250W, it will probably be spot on, because the actual CPU package power will have a slight overshoot anyway. So you will be limited to mid/high 80°C temps, that is really the maximum you want to allow, to also have a little bit of headroom for higher ambient temperatures in the summer.

You could also play with different values for Short/Long power limit, but at such high limit, it doesn't make a big difference anymore. Above 250W it mostly gets into the "junk power draw", meaning the power draw still keeps increasing a lot, but for minimal performance gains. Plus, a lot of day-to-day load will not nearly tax the CPU as much as Prime95. So these 250W are only set to keep your cooling on top of things at all times, even with extreme stress tests.
 
During this test, where it indicated 96 degrees CPU, the CPU was already a little tired from the previous tests. I made another attempt of 20 min. in Prime95 after a long break and the CPU reached a maximum of 95 degrees. Power consumption surprisingly dropped from 311W to 303W. The computer will actually be used mainly for gaming and VR. I guess I could skip the settings TDP but I tried it anyway. I set the long and short power limit to 250W. In the Prime95, the CPU reached a maximum of 90 degrees in a 20 minute test. According to HWinfo, it doesn't fall into throttling. I'll leave it at that. Best regards.
 

Attachments

  • Zrzut ekranu (106).png
    Zrzut ekranu (106).png
    1.1 MB · Views: 27
I have one more question. I was forced to mess with the bios again today because I had problems with the disappearing temperature of both RAM memories (dimmA1 and dimmA2 at the same time) Previously they disappeared one at a time and I thought that the basis of the problem was too aggressive lowering of the CPU Vcore (-100mV) After changing the lite CPU load from 9 to 6 and disabling the IA CEP SUPPORT option, the temperatures after saving the changes reappeared. Today, as soon as I turned on the computer, HWiNFO showed no temperature of both RAM chips and that's why I started again bury. Previously, restoring the default settings in UEFI fixed the problem, but I didn't want to do it again. I was looking for the cause. Experimenting again today with the IA CEP SUPPORT option from off to AUTO and after the reset, the temperatures returned. I did the 20 min test again PRIME95 and surprisingly the maximum CPU temperature was 85 degrees. As a reminder, with the IA CEP SUPPORT option turned off, it was 95-96 degrees at its peak. What does this option actually do? I can leave it on AUTO because of the Better thermal culture of the processor? Vanishing temperature case reported to MSI along with HWiNFO, AIDA64, Msi Center, Msinfo32 reports because I still suspect H9 bios.
 
IA CEP is the "Current Excursion Protection" mechanism for the IA cores (normal CPU cores), and it wants to prevent any undercurrent over overcurrent from a narrow window that is expected for a CPU. Once it sees a break from the norm, it will work against it by also lowering performance a lot. With an active IA CEP, when using a lower CPU Lite Load mode, the performance can massively decrease, similar to here. Of course the temperatures will be much lower as well, but without the performance, that is no good. So run Cinebench R23 for example, and if you get a lot lower scores, you know that this is not a solution.
 
In fact, the performance dropped in cinebench from around 35k to around 33.5k. Thanks for the valuable advice. I also went back to the previous H8 BIOS and surprisingly in the prime95 test the maximum temperature was 86 degrees, where on the H9 BIOS it was 95-96 degrees on the same settings. The cinebench score itself is also higher than on H9. However, this bios seems to be underdeveloped. I'll also be watching to see if the problem with RAM temperatures disappearing will happen again. Thanks
 
Last edited:
I was forced to mess with the bios again today because I had problems with the disappearing temperature of both RAM memories (dimmA1 and dimmA2 at the same time)
This is a bios bug with HWinf64 update to the latest BIOS, it should have the patch or Disable the Asmedia SATA controller in the Bios if not using all 6 SATA ports


I generally just run the AIDA 64 Stress test which will find most instability problems.
That I ignore as long as I can run Cinabench and Furmark at same time for 15 to 20 Minutes I'm petty set for how I use the PC
1702044995689.png

Screenshot 2023-12-09 003320.jpg
 
Last edited:
Back
Top