Overclock.net banner

1 - 15 of 15 Posts

·
Registered
Joined
·
7 Posts
Discussion Starter #1
I've stumbled across a PSU stress test combo that can be very useful in some cases. Finding it let me fully resolve an intermittent problem I was having.

I was exercising a new RTX 2080ti by running Furmark's "GPU Stress Test" in a 1366x768 window with 8X MSAA. The PSU was a high quality but very old BFG 800W unit, which had been powering a GTX 1080ti for 6+ months with no problems.

All was good until after a week or so the PC did an instant and total power-off, requiring wall power to be cycled. These shutdowns became more frequent, always when I ran some other app along with Furmark. It was likely that a PSU weakness was being exposed, but I couldn't rule out an issue with the new GPU. Unfortunately these crashes remained unpredictable and non-repeatable.

Then I happened to try O&O Software's free "ShutUp10" utility (along with Furmark). Nothing happened during normal use of this util until I dragged the right-side scrollbar down. Then the PC instantly shut down - and this was exactly repeatable over three or four more tries.

That convinced me that the problem was with the PSU, not the graphics card. I didn't think to try the 1080ti for comparison (...) but went ahead and installed a new Corsair HX1000 PSU, set for 'single rail' 12V. That fixed things - no more shutdowns with ShutUp10 nor with several taxing games running along with the Furmark window.

And note this: with the new PSU, when I drag the ShutUp10 scrollbar, I hear ticking noises from the machine. This only happens if Furmark is also running. I'm guessing that these sounds are the result of current spikes impacting the PSU's reactive components, and that the old BFG's filtering was no longer sufficient to prevent the transients from triggering a protection circuit.

I've now also tested Furmark and ShutUp10 together on two other PCs having a GTX 1070 and Corsair 500W and 600W power supplies. I get the same ticking sounds with these, but no shutdowns.

So it looks like these two apps together somehow create significant PSU transients. They would be useful if you suspect a PSU is failing but want proof, given that replacing one is a chore at best.
 

·
Senior OCN Member
Joined
·
39,423 Posts
Hope it was the new HX1000 and not the old HX1000
1000 watts is massive overkill by the way
 

·
9 Cans of Ravioli
Joined
·
19,207 Posts
A power virus plus that puts an extremely unrealistic load on your components plus another program causes issues? Whodathunk.

Stop using Furmark.
 

·
Registered
Joined
·
892 Posts
A power virus plus that puts an extremely unrealistic load on your components plus another program causes issues? Whodathunk.

Stop using Furmark.
Sweeping the problem under the rug doesn't make the problem go away.

@OP, overcurrent protection was kicking in with your BFG power supply. Thats why you had to completely power cycle the PSU before it would let you turn it on again. If you tick a paper click into a 12v pin and another in a ground pin and turn on the PC, it'll do the same behavior.
 

·
Facepalm
Joined
·
8,210 Posts
A power virus plus that puts an extremely unrealistic load on your components plus another program causes issues? Whodathunk.

Stop using Furmark.
While Furmark isn't usually recommended to be used, it DID help the OP prove that his old PSU was dying. So it did its job for its intended purpose--stress testing.
While furmark is the prime95 for GPU's, both prime95 and furmark, when used properly, can help expose borderline or flaky hardware. That's different than using prime95 (in AVX virus mode) or furmark to determine if your high overclock is stable or not.

I'm with the OP on this.
 

·
professional curmudgeon
Joined
·
10,391 Posts
personally i find kombuster fuzzy dougnut combined w/prime95 maxes out my components more than anything.

thats my PSU stress test.

nvidia drivers get in the way w/furmark.

sorry but, it is hard to believe a simple GUI such as 0&0 causes any power stress - its just a GUI. correlation doesn't equal causation.
 

·
Registered
Joined
·
7 Posts
Discussion Starter #7 (Edited)
@Pook and looniam: It's not clear whether you're incapable of understanding what I explained, or just don't bother to read things before kicking out an arrogant, dismissive reply. Either way, while that does boost your post count it isn't doing the forum any good. Just FYI. I'll check back in another ten years and see if things are back to where they were.
 

·
9 Cans of Ravioli
Joined
·
19,207 Posts
thanks for your valued contribution, I'm sure it'll reform stress tests everywhere.

that the kind of response you were looking for?

sorry, I guess :rolleyes:
 

·
professional curmudgeon
Joined
·
10,391 Posts
@Pook and looniam: It's not clear whether you're incapable of understanding what I explained, or just don't bother to read things before kicking out an arrogant, dismissive reply. Either way, while that does boost your post count it isn't doing the forum any good. Just FYI. I'll check back in another ten years and see if things are back to where they were.
maybe check your own reading comprehension. then your "testing" method. you have at best a theory. what have you done to confirm it? used a killawatt meter? mulit meter? oscillator? anything?

nope. sorry if your ignorance about a program being pointed out caused you so much butt hurt as a simple inquiry causes you to be so defensive.

if that is really what you believe and can't handle criticism of a flawed test - then you sir are the one who isn't doing this forum any favors.

good luck with that.
 
  • Rep+
Reactions: The Pook

·
professional curmudgeon
Joined
·
10,391 Posts
^EXACTLY

but i guess it doesn't matter when testing an 8-10 year old psu anyways.

or how long has BFG been out of business?
 

·
Registered
Joined
·
5,374 Posts
O&O shouldn't add any real load to the CPU, GPU or PSU, O&O is just a UI for editing your registry and a few other things.
However if you're editing running Windows functions while the GPU is banging off the power limit, down clocking/undervolting and repeating until the GPU finds a clock speed it can settle on, I can see you having any number of issues.
Furmark used to be a thermal test so that you could see if your voltage and power settings were able to be cooled by the cooler and powered by the PSU...to test overclocks usually. It's not a stress test and it never really was, if you use it to test thermals and power, you're doing it right. However now that GPUs clock speed and voltage fluctuate based on temp, you can't thermally test an OC this way because it simply won't hold the clocks. If we had full BIOS control we could force it to hold boost and voltage at a much higher temperature, I think they should hardware voltage lock the cards to an extent, maybe +10%, and give us full control over the rest. The voltage and power are going to stop scaling long before you risk damaging anything with voltage.
 

·
professional curmudgeon
Joined
·
10,391 Posts
DEBUNKED

running O&O adds no additional load than running just furmark.
Untitled.png
the scrolling was laggy as all hell, but i'm not surprised as i said, its just a GUI.
 

·
Vermin Supreme 2020
Joined
·
25,774 Posts
you had me until furmark.
 

·
professional curmudgeon
Joined
·
10,391 Posts
you had me until furmark.
well since i have a minute. furmark isn't good at stressing the PSU w/nvidia card. since the 580/590 vrms going all popcorn, they have put the kibosh via drivers for furmark.

980ti w/350-400 watt limit in vbios - i have confirmed, long time ago in the 980ti owners club, that gpu-z reading is accurate comparing to a kill-o-watt and accounting for psu efficiency.

running furmark:
Untitled1.png

230 watts in gpu-z

running kombuster fuzzy doughnut:
Untitled2.png

396 watts in gpu-z

if looking to really stress the PSU , kombuster looks to be better. run it w/prime, settings depending on the cpu gen and be prepared for the fireworks.
 
1 - 15 of 15 Posts
Top