Frage Hat das fehlgeschlagene CPU-/GPU-Stresstest etwas mit dem Netzteil zu tun? / Does the failed CPU/GPU stress test have anything to do with the PSU?

[psuforalllife]

A friend of mine tested my PSU by using Prime95 to stress the CPU (13900KS) and 3DMark to stress the GPU. During his tests, 3DMark was experiencing lagging issues, and Prime95 even crashed at times.

1st test video
In the first video, while 3DMark was being opened, he started stressing Prime95, which caused 3DMark to open slowly. At 00:30, Prime95 shut down unexpectedly.

2nd test video
In the second video, he opened 3DMark first and waited until Prime95 started stressing before running 3DMark. Although 3DMark took some time to load, it eventually ran smoothly.

3rd test video
In the third video, Prime95 closed at 00:16 while 3DMark was still loading. He reopened both Prime95 and 3DMark, but an error occurred in 3DMark.

4th test video
In the fourth video, he used an Asus Loki 1000W Platinum PSU to test the same system and stress applications. The process was similar to the second video, where 3DMark took some time to open but was able to stress the CPU and GPU simultaneously.

From my perspective, I don't believe the issue lies with the PSU. It seems more likely that there is a conflict between the test process and software, which is causing high CPU resource usage. What are your thoughts on this?

-------translated by ChatGPT-------

Ein Freund hat mein Netzteil getestet. Er hat Prime95 verwendet, um die CPU (13900KS) zu belasten, und 3DMark, um die GPU zu belasten. Während seines Tests hatte 3DMark Verzögerungen und manchmal ist Prime95 abgestürzt.

1st test video
In diesem Video konnte 3DMark nicht reibungslos geöffnet werden, während Prime95 lief. Bei 00:30 hat sich Prime95 heruntergefahren.

2nd test video
In diesem Video hat er 3DMark im Voraus geöffnet und gewartet, bis Prime95 in Betrieb war, bevor er 3DMark belastet hat. 3DMark hat einige Zeit zum Laden gebraucht, konnte letztendlich aber ausgeführt werden.

3rd test video
In diesem Video wurde Prime95 bei 00:16 geschlossen, während 3DMark geladen wurde. Er hat Prime95 und 3DMark erneut geöffnet, doch es ist ein Fehler bei 3DMark aufgetreten.

4th test video
In diesem Video hat er die gleiche System- und Belastungsanwendungen mit einem Asus Loki 1000W Platinum getestet. Der Ablauf war derselbe wie im zweiten Video. 3DMark hat einige Zeit zum Öffnen gebraucht, konnte aber schließlich sowohl die CPU als auch die GPU belasten.

Aus meiner Sicht liegt das Problem nicht beim Netzteil. Es scheint eher, dass es einen Konflikt zwischen dem Testablauf und der Software gibt, der zu einer hohen Auslastung der CPU führt. Was denken Sie dazu?

 

[Derfnam]

Which PSU was used before the switch and which GPU was used throughout?

 

[psuforalllife]

GPU is 4090, but I am sorry I couldn't say the former PSU name because of reputation issue.
(I have uploaded videos to YT because I found here is problem to see video in google drive)

 

[Derfnam]

You're not really helping here...
4090 @ stock? Can you at least tell the wattage of the unnamed PSU?

 

[8j0ern]

Hello, if you use Prime95 which Version of the Programm do you have ?

Because the newer Versions are Calculating with FMA3, which can stress the CPU more then AVX2 it self.

So try both seperate is valid enough, as you can push the GPU more if the CPU ist not "overloaded"

 

[Moeppel]

I am more confused about the fact anyone would try to run anything else when Priming it up.

Its very essence is to torture your CPU and things getting slow under 100% CPU load is hardly a surprise. In fact it is to be expected.

Given the noise once Prime95 kicks in: What CPU cooler is being used? This sounded nothing but utmost unhealthy, if not outright insufficient.

As for the PSU: 1000W is sufficient and lest I am mistaken it is Seasonic creating these devices, so quality is usually to be expected. An insufficient PSU would usually shut the system down or start smoking/exploding but not have individual applications crash.

Check out windows event viewer whenever a program taps out. I would suspect that a calc/core failure causes the crashes. These issues can be power as well as heat related.

The solution would be to re-seat/paste the CPU and cooler, if the cooler was at all sufficient to begin with and only if the system fails Prime95 on its own.

Finally, when using these tools use HWINFO or something to that effect on the side, to keep an eye on temps and voltages and you do NOT test GPU and CPU simultaneously unless you want to test your PSU. When you think about for a second you'd immediately notice that a 100% CPU couldn't provide notable drawcalls anyway, as it is busy crunching numbers and calculations (i.e gimped FPS/performance).

Further, if that is what you really wanted you should ditch 3DMark for Furmark instead