Over the past month my PC has been doing some weird stuff, such as random blue screens, not logging into windows, display driver not responding, and the on board NIC disappears and then reappears. All of this is sporadic and I cannot reproduce the issues with any confidence. Sometimes the PC runs great for days no issues.
By swapping parts with my other computer I believe I narrowed it down to the Motherboard or the PSU. Both are about 3.5 years old, and I am a little rough on them with constant folding@home or FPS games.
I tested with OCCT in an attempt to figure something out and noticed the 12 volt rail is hanging out at 11.84 during the OCCT CPU test. As soon as the test ends it jumps back up to 12.04 That test is just CPU, it is not even using my video cards. I never noticed this before, but honestly the computer has been running great for years so I never looked. The funny thing is according to my Kill-A-Watt wall plug the computer is only pulling like 400 watts during the test.
Do you folks think this dip in the 12v could explain all the weird stuff? Should I look into a new PSU?Right now everything is running at stock speeds and I have removed my second video card to try and narrow down the list of possible causes.
Based on that graph your 12V rail is still in spec but software isnt the best way to check voltages, the better option would be to use a digital multimeter to check the voltage at idle and at full load, but based on your description of the issues i would suspect the motherboard rather than the PSU, especially with the randomly disappearing NIC.
I do normally always measure the Voltage with a DVM at least to verify that the compure software is correct. Once verified I RELY on the software.
If I was a quessing man, I'd quess that the +12 V rail is OK; However I'd run furmark on the video cards to verify. The Spec is Min > 11.4 V, But I pitch a PSU if it drops to 11.6V.
The Hard part is evaluating the neg spikes, on graph they look OK, but that is not conclusive (A DVM is normally no help on spikes either). Best check, unless you have a expensive O'scope, is to try with a know good PSU.
Just a couple of comments.
How do you know to trust the DVM. There are cases where the computer software is correct and the DVM is inaccurate Example Computer soft ware = 11.8 V, DVM = 11.7 Which is correct????. At work we use relatively High end DVMs approx $200+ andTthey are calibrated once a year. For Home use I have a Fluke 85 III. I verify the aaccuracy periodically.
Ref: " software measures from the main board," this is really where he wants the measurement. This measurement factors in the losses from PSU to measuring point.
The graph he is displaying (once verified against a DVM) is better than the readings on many low end DVMs. Low end DVMs filter the input DC, while the graph may show spikes that the DVM clips and or can not respond to. BUT if this is a problem then use a O'scope
when i run furmark and measure voltage with HWMonitor, the voltage drops also, it did so with my old thermaltake 470w and my new Antec Trupower 650w, that puts the drop down to the motherboard. A multimetre measures the 12v pretty stable regardless what the software is telling me.
As you indicated the DVM is stable when connected directly to the PSU. If readings are close (ie with 1 %) you DO NOT know witch is more accurate unless the DVM has been calibrated recently.
In his case, the software readings looks valid. While I would verify with a DVM just to be sure, the DVM as you put it provides a buffered reading and if values are within a 1/10 volt agreement I would go by the software readings - ie you do not know which reading is more accurate unless you have calibrated the meter.
As stated, it may not be the "stead state level", it may be a problem with spikes and the DVM does not show these spikes. In this case there is only two ways to verify (1) replace PSU with a known good PSU and (2) the use of a O'Scope which will not only show the staty state voltage but also the ripple and any spikes.
Generally you can tell if the "software" is not providing a valid reading as the value will be out in left field - ie 5 V reading 3 volts, or 12 V reading +16V which is normally the programmer is looking at an incorrect location and returning a software "rail" value.
Case inpoint: HWmonitor does not even show my +12 V, but shows a reading on -12V low and behold that reading was exactly the same as a DVM reading on my +12V (and varied identically under load). So I just mentally change -12V rail to +12V rail. BUT it is accurate.
I have a cheap generic multi-meter that I used to test the PSU with. Under load any of the 12 volt pins on the pci-e and the 24 pin connector fluctuate wildly. They bounce from 11.7 to 12.4 so fast I can't even read the screen of the multi-meter. The OCCT software does not record these crazy fluctuations. Could these fluctuations be causing the issues.
As I am getting desperate and I just got a "video driver has stopped responding and has been restarted" I decide to act on a hunch and wire up an old 350 watt power supply to run just the video card. I simply plugged in the pci-e connecter, then used an adapter to use two molex for the other power connector on the video card. Used a paper clip to jump start the PSU and then powered on the CPU and board (which are running off the old PSU).
The 350watt PSU running the video card is stable at like 11.9 volts all the time. It almost never moves. The bigger PSU that runs the CPU and board is still moving a bit but it is much more subtle and only like .3 of a volt.
My conclusion is that the PSU is wacked.
Any way it has been a week and not one problem.... So I ordered a new SeaSonic 80+Gold 750 Watt PSU.
Hopefully installing the new PSU will resolve the issue for good.
/The new PSU has been installed and running for 3 days now. I had both cards and all 4 cores folding@home at 100% usage for the last 8 hours. The kill-a-watt is recording around 600watts being pulled from the wall.
I also played a bunch of blackops and no issues yet. It seems the PSU replacement resolved the issue. I will report back after a few more days. Thanks again for the multi meter suggestion and the thoughts about measuring the voltage from the PSU vs using software from the board. Lots of info to think on.
Hey guys its been over a week now and not 1 crash. The new PSU really seems to have solved the problem. Thanks for the suggestion of the DVM. I had a feeling the PSU was the issue but I was not sure until the multimeter showed the 12v rail jumping so much.