Is my psu failing, or something else?

[konet]

I recently upgraded my gpu from a 9800gx2 to a gtx 580, and since then, my computer has been crashing to a black screen a few minutes after starting a game, or a bit longer when watching videos. I had the same problem with a hd6970 I bought, but returned for a refund because I thought the card was bad. My 9800 still works fine, and I'm pretty sure neither my CPU nor gpu is overheating (this has happened with an open case). My power supply, an ultra x3 1000w that's about 2.5-3 years old was in operation when my motherboard fried a few months ago. I don't own a multimeter, but according to my bios, the rail outputs are within normal ranges, except that my 12v rail is outputting at about 12.6v (this is in the bios, so I'm guessing that's idle?). Under clocking the card seems to delay the problem quite a bit, to the point where I can play games for much longer before the crash occurs. The crash dumps point to a file called nvlddmkm.sys, which, after researching online, can be caused by many things. So, do you think the issue is with my psu, or something else?

 

[ko888]

Was the motherboard also replaced? What motherboard did you have before and what motherboard do you have now?

 

[konet]

I had a asus striker2 extreme, I now have an intel dx58so

 

[JackNaylorPE]

That Ultra PSI was the likely cause of the MoBo failure.

That nvlddmkm.sys is an nvidia file..... your failures are graphics related, my guess either from voltage or temperature issues.

and I'm pretty sure neither my CPU nor gpu is overheating (this has happened with an open case).

An open case doesn't mean anything. An open case would allow you to diagnose inadequate case cooling **if** you had a nice big house fan blowing on all the components. Do this:

Download OCCT from here and run the GPU and PSU tests. OCCT will monitor the results on screen and if you manage to pass the 60 minute tests will graph the results.

http://www.ocbase.com/perestroika_en/index.php?Download

DO NOT even think about running the OCCT PSU test on the Ultra .... that test has been known to toast better units than the Ultra. The maximum allowable voltage variation according to the atx spec is 5%. I look for 1%. Your 12.6 is right at the spec limit. However, since you are looking in the BIOS, that's a relatively no load condition, I hazard to guess that it's the under load voltage that might be the issue.

 

[ko888]

Did you do a fresh install of the operating system and did you download and install the latest graphics card driver from NVIDIA's website?

 

[konet]

Ok, I'll run that test now. By the way, since the upgrade, the screen has started flickering for a few seconds right after the "press f2 to enter bios" screen, and takes a long time to get past the "starting windows" screen during boot. Could this be related? I tried formatting, as well as all recent drivers.

 

[konet]

Hmm, for some reason OCCT is not detecting votages.

 

[ko888]

Download and run CPUID's Hardware Monitor freeware utility and see if the system voltages appear in there.

 

[konet]

the only voltage listed in hwmonitor is one labeled VIN0 under my gpu (it's at .96v if that matters). The only software I've found that can display my voltages is the intel desktop control center. As of right now, the voltages are as follows: CPU 1.26v, Uncore 1.16v, IOH 1.11v, +12v rail 12.67v, +5v rail 5.04v, +3v rail 3.28v

 

[ko888]

None of these third-party utilities seem to be reporting the voltages on the Intel DX58SO because Intel has used their own proprietary ASIC (Application-Specific Integrated Circuit) chip to control fans and monitor temperatures and voltages.

Have you tried placing a heavy processing load on your system and monitoring the voltages?

 

[konet]

I ran the occt processor test for 35 mins with no errors and no unreasonable temps/voltages. Same for the gpu test in windowed mode. however, the black screen crash occured when I tried to run the gpu test in fullscreen. (there is no way for me to monitor voltages in the fullscreen test because I have to use the intel software, which can't be on top of the fullscreen test, so I can't see the numbers before the crash)

 

[konet]

Hmm, I went out and bought a multimeter, and my 12v and 5v rails actually seem pretty stable, never straying far from 12.5v and 5.04v respectively. I can't test the 3.3v without jumping some wires, which I don't know if I feel confident doing. Could I have just gotten 2 bad cards in a row, however unlikely that may be?

 

[ko888]

The only problem with using a multimeter is that you can't measure ripple and noise.

The GeForce 9800 GX2 may be more immune to high ripple and noise than the GeForce GTX 580.

 

[minitron815]

You only tried the one HD 6970? It was probably the card if you underclocked it and that delayed the problem....

Everyone is quick to blame PSUs and recommend wasting money on upgrades on these forums.