Voltages jumping around

Dominic87

Honorable
Apr 14, 2013
6
0
10,510
Hi there

I've been measuring GPU voltages on HWInfo, MSI Afterburner and GPU-Z. At idle the +12V reading is steady.

However, when I stress test on FurMark or in gaming the voltage randomly drops or rises (lowest reading 7.5V, highest reading 14.5V).

I have been having problems with freezes and crashes. Do you think the PSU is the problem? I have a Lepa W-Series 500W '80 Plus' PSU.
 


Well, I think you definitely have problem with PSU. Almost all voltages are x±10%, where 'x' represents the voltage. If your warranty expired, open PSU and check your capacitor. If they are ‘bubbled’ replace them. If they are ‘flat’, then something else is bad. Consider replacing your PSU.
Cheers.
 


Thank you mironso - I haven't actually checked my PSU warranty; guessing I may be able to RMA
 
NOTE: Specs for the +5V and +12V is 5%
For +5V = 4.75->5.25 V This is based on TTL logic chips which do not work below 4.75V (except for some Mil spec chips)
For +12V = 11.4V ->12.6V (Personally I use 11.6V as Min). Low voltage on the +12V results in a increase in current, to compensate for low voltage which could result in overheating.

Not normally found is the deviation when switching from Idle to full load. For this I use a max delta of 0.3V for the +12V rail. ie if at Idle my voltage was 12.2 Volts, I ditch the PSU if voltage drops to 11.8V when switching from Idle to running furmark. Even though the spec is down to 11.4V, as a drop of 0.4 V indicates the PSU is losing regulation.

My Bottom line here is to DISCONTINUE usage untill you replace the PSU, Your +12V is WAY out of spec.
 


Thanks Chief. I am going to. I have ordered a new Corsair CX600, will see if that fixes things.
 


OK I've just tried the replacement Corsair CX600 PSU. I'm getting the same readings on the +12V using GPU-Z/MSI Afterburner/HWInfo (but only for the GPU). Does this mean it's the GPU which is the problem? It's a 6870 - clean install of the newest drivers.
 
So your +12V still shows a drop to +7.5V when running Furmark, BUT NOT with prime 95??
Are you still getting the Lock-ups/crashes/freezes??

First I don’t think that the 7.5 V on your +12V rail is valid unless it is a voltage spike that is faster than your PSU can react to. 7.5V > xx microseconds should shut down the system. More likely the 7.5V value is a Rail Limit imposed by the software, that is any reading below a set point are all displayed as a set value (your case 7.5) example the software expects the +12V to be between say 10.5 v and 13.5 V. If it detected a value of say 10.4V it would not display 10.4, but display 7.5

Your initial PSU is listed as a very good PSU, better than the corsair CX series.
Since you are getting the same results with both PSU, I would tend to rule out the PSU.

This tends to point more toward the GPU (Not normally the culprit).
This is ONLY a guess, but what could be happing is that your GPU is momentarily causing a very low impedance to the +12V (ie a couple of microseconds), The PSU voltage is dropped, but before the system say SHUT down the Low impedance has returned to normal and the PSU rather than return to normal first overshoots the normal value ie jumps up to say 13.5 V then back to +12V. In extreme cases this can cause what is referred to as damping oscillations.

And herein lays the problem, you nor most local computer repair shops are equipped to measure this. Requires a dual trace O'scope, one trace for Voltage and a 2nd on for measuring current. Problem compounded it that a current measurement would require some special cables as compute PSU cables are not designed to allow this.

You, or the local shop are kind-of stuck with the Trial an error method of replacing components - Can be expensive as you already paid for a 2nd PSU to replace a quality PSU.

Then there is also a possibility that it is sensor related (Motherboard reporting) or software (reading what the MB is reporting. I tend to rule these two out as they in themselves should not cause the "freezes/crashes"

Sorry I'm not more helpful, Hard to say, but for grins and LOLs
1) Place O'scope voltage probe on Pin X
2) Short the +12V wires to one wire then expand back out to x number of wires to Pci-e pwr connector (this is for a current probe.
3) Set trigger to above nominal value, and slope to Positive
4) ect
O, My Dual trace scope cost list price is around 3 Grand.
 
OK I bought a new GPU. No problems since running it, and the voltages are reading as stable. Looks like it was the 6870.

Just thought I'd post in case someone stumbles across this thread with similar problems.