GPU failing because of power supply?

[russcatt]

[FONT=&quot]Hey guys first time posting here after I stumbled upon the How To Test Your PSU Rails sticky.

Anyway getting to the point in hand since building my new rig around august I am having trouble with the GPUI have had 3 of them since then:

1x all of the display ports were dead on arrival (tested in 2 systems)
1x lasted a while and a single fan died (tested in 1 system)
1x after only having the replacement a week all of the display ports stopped working (tested in 1 system)


After arranging a RMA with scan for the current card and finding they have only had a 4% failure rate with the exact card it got me thinking either im really unlucky or something else is wrong with the system, the guy on the phone said it could something to do with the motherboard or a faulty PSU potential a rail. I have done no overclocking except changing the RAM to the manufactures speeds. [/FONT]
[FONT=&quot]I have already checked the voltages in the BIOS and speedfan though I don’t really know what I am looking for(all taken without the GPU installed ) :[/FONT]
[FONT=&quot]BIOS voltages:[/FONT]
[FONT=&quot]Vcore – 1.104 jumps around 1.112, 1.120[/FONT]
[FONT=&quot]12.00v – 12.038 jump up and down to 12.091[/FONT]
[FONT=&quot]5.00v – 5.64[/FONT]
[FONT=&quot]3.30v – 3.312[/FONT]
[FONT=&quot]Speed fan voltages:
Vcore – 1.06v, 1.08v, 1.10v
+12v - 12.04v jumps to 12.09v AVCC - 3.31V

[/FONT]
[FONT=&quot]Specs:
Intel Core i5 3570K
Cooler Master Hyper 212 Evo
Asrock z77 extreme 4
2GB MSI GTX 670 Power Edition OC
Corsair 8GB 2X4GB DDR3 1866Mhz
Crucial 128GB M4 SSD
1TB Western Digital Caviar Black
Corsair TX550M 550W
CM Storm Trooper Case
Trooper windowed sidepannel
NZXT 24x Red LED Sleeve - 2m[/FONT]

 

[alpenwasser]

Hi there, and welcome to the forums!

From what I've been able to find out, the Corsair TX550M has a single 12V rail, and is a pretty good PSU (as is the habit with Corsair's PSU's in general). Also, the 3.3 V and the 5 V rails are powered off the 12 V rail, and there is one of each.

Power gets to the GPU in two ways:

• The PCI socket and
• additional PCI-Express power plugs (6 pin/8 pin).

The additional power plugs are 12 V only. The PCI-Express slot uses 12 V for main power supply and 3.3 V for debugging ports and power on/off information.

I think that if your PSU were truly defective, that defect should have affected other components by now since everything is powered off the same rail (the 3.3 V and 5 V being defective should have affected the M/B and/or HDD/SSD, same goes for the 12 V).

Of course that could be the case, and it could very well be that that defect which the PSU did to the M/B is now cascading from the M/B to the GPU card, but I consider it a bit peculiar that the M/B is still working fine if that is indeed the case.

I think that more likely than having a defective PSU either you really have been very unlucky or there is something wrong on the M/B. Have you tried running the GPUs in different PCI-E sockets?

You could still get a PSU tester for a few bucks and check your PSU. Of course that would not detect every possible defect a PSU could have, for that you would need some serious lab equipment, but it would be a start.

 

[airdeano]

you'd need to verify the output of the GPU harness. the specs you've quoted
are from a bus and not the actual GPU point. since the 550 is semi-modular
your can back probe the yellow leads in the GPU plug (should be 12.??v+) and
measure the black ground wires for continuity, voltage drop, and actual 12.??-.
the only common is the PCIe slot and GPU power.

 

[russcatt]

Thanks for the response guys, I dont really have the means of testing the power at the moment and the PCI cables are coated black so there is no way of seeing what colour the wire is and by looking at examples online tthe position of the wires seen to vary,would something along the lines of this be worth getting ?
http://www.amazon.co.uk/Power-Suppl...365095589&sr=1-1&keywords=Power+Supply+Tester

 

[airdeano]

the top row of a 6-pin connector with the lock-clip are the +12v cables and the
lower group is the -12v cables. you can back-probe the connector with the
voltmeter probes.

 

[andrewmchugh]

the top row of a 6-pin connector with the lock-clip are the +12v cables and the
lower group is the -12v cables. you can back-probe the connector with the
voltmeter probes.

I don't think there is a PD of 24v over a graphics card. The lower is ground 0v. [1]

1. http://www.playtool.com/pages/psuconnectors/connectors.html#pciexpress

 

[alpenwasser]

Thanks for the response guys, I dont really have the means of testing the power at the moment and the PCI cables are coated black so there is no way of seeing what colour the wire is and by looking at examples online tthe position of the wires seen to vary,would something along the lines of this be worth getting ?
http://www.amazon.co.uk/Power-Suppl...365095589&sr=1-1&keywords=Power+Supply+Tester

For a few bucks it's not a bad investment; I have a similar one I use to test every PSU before I hook it up to any components, just to check if everything is wired correctly (especially important if you've done some sleeving or had someone else do it for you).

However, be aware that these things cannot find every fault possible. But it's a good start.

I don't think there is a PD of 24v over a graphics card. The lower is ground 0v. [1]

1. http://www.playtool.com/pages/psuconnectors/connectors.html#pciexpress

According to ATX specs the only -12V line is in the 24 pin to the M/B, and the wire's usually blue (not sure if the blue part is official spec or just most common, couldn't find it in the pdf).

Official Specs (*.pdf)
Wikipedia: ATX Power Supply


As for the original problem, you can indeed measure out the voltages over the cables, pins etc. but you'd need a (decent) voltmeter for that, naturally. As you mentioned you don't seem to have access to one nor intend to buy one. And even if the voltages are all good, you might still have spikes or ripple that's breaking your components and could only be picked up by an oscilloscope. And those suckers are not cheap.

One possibility I'd try in your situation is to contact Corsair, explain your situation to them and ask if they can have a look at your PSU since you don't have the required tools to do so. If you still have warranty, that should also get you a replacement.

 

[russcatt]

I went ahead and bought the PSU tester thought it could be useful in the future and I may be able to get my hands on a voltmeter, think I will test it with both before potentially sending the PSU back to corsair.



As for the voltmeter I have no idea what I’m doing with it, any advice on what setting to use, which probes(black and red) to use where and for how long I should take the readings for?

 

[alpenwasser]

As for the voltmeter I have no idea what I’m doing with it, any advice on what setting to use, which probes(black and red) to use where and for how long I should take the readings for?

The settings should be such that it is set to measure a difference in potential (many voltmeters are multimeters that can also measure direct current), and the scope should be so that 0 to 12 V falls within it.

Red/black probes: doesn't matter afaik, they're just color coded so that you know which is which when you're working with them.

Duration of measurement: Voltages work pretty fast, so as soon as you get a value that doesn't change anymore you're good.

Location of measurement: I would measure over the length of the PCI-Express auxiliary power cable. Measure if the voltages at the connector which plugs into the GPU are ok, if they're not, measure if the voltages coming out of the PSU into the cable are ok. If those are ok, you're cable is shot (probably bad crimp connection or defective pin), if they are not ok either your PSU has gone bad.

Obviously, you will have to do this for each pin (so, one probe goes into the pin, the other probe is connected to ground).

You could also measure the voltages at the 24 pin connector and the voltages in the power part of the PCI-Express socket. The last part might be tricky though; getting a probe into that socket without damaging it.

I'm not an expert on this, so someone more knowledgeable feel free to correct any stupidities I might have uttered in this post or amend any important stuff I've forgotten.

 

[russcatt]

Thanks for the reply’s on this guys you have really been helpful, I managed to borrow a voltmeter and I got the following stable voltages that seem to be ok I also tested the PSU rails.
PSU rail test:
12v – 12.4v
5v – 5.05v

For the PCI I checked voltages on the cables and PSU checking each pin with each earth and got the following results:

Cables and sockets were the same either -12.04 or 12.05 depending on the prong orientation
Also I was able to get readings from the PCI slot either -12.03 or 12.04 depending on the prong orientation.

 

[alpenwasser]

Huh. Well, the good news is your voltages seem ok from the looks of it. Certainly nothing that would fry your components in such quick succession imho. Bad news is, still no idea what's wrong :lol:

It could still be extreme ripple and/or bad voltage spikes, but I wouldn't know how to measure that without an oscilloscope. Also, I think it's very peculiar that nothing else has broken, just your GPUs. If you really did have ripple/spikes bad enough to break your GPUs, I would expect that to have broken something else by now as well.

Then again, it really might have been bad luck.

:headscratch:

 

[russcatt]

A big thank you to any body who posted in this thread this really is a helpful forum glad I signed up :lol:

Too bad I didn’t solve the problem hopefully I was really unlucky with the GPUs, think I will go for another brand this time round, maybe a different card I’m between a new 670 or a 7970, its been a while since Iv hade any ATI/AMD in my rig think it was a and athlon 3200 and a radeon 9200 many years ago.

 

[russcatt]

Just a little update, I sent the card back yesterday and today they said there was nothing wrong with it found in the test they have done (tested all of the display ports with 3Dmark firestrike and they all worked).

The day before I sent it back I tested it with MSI kombuster and after a few seconds the display cut out and would not return after restarting but then come back a few restarts later.
I got the PSU tester today and after testing each PCI power cables for 10mins I got these stable results and nothing seems wrongs with then:

-12V – 11.7
+12V2 – 12.0 or 12.1
5VSB – 5.0
PG(ms) – 290 or 300
+5v – 5.0 or 5.1
+12V1 – 12.0 or 12.1
+3.3 – 3.3

 

[alpenwasser]

Just a little update, I sent the card back yesterday and today they said there was nothing wrong with it found in the test they have done (tested all of the display ports with 3Dmark firestrike and they all worked).

The day before I sent it back I tested it with MSI kombuster and after a few seconds the display cut out and would not return after restarting but then come back a few restarts later.

Did the PC boot up when there was no image displayed or did the machine stop entirely? Because a defective GPU should not usually stop the machine from booting (I've had this happen) unless there's some sort of effect back onto the M/B or (unlikely) the PSU.

Did you ever try plugging the monitor into the HDMI port (or another graphics port) on the M/B and running it off of that while the GPU was not displaying anything?

Also (and I know this seems like a really stupid question): Are the monitor and the cable ok?

I got the PSU tester today and after testing each PCI power cables for 10mins I got these stable results and nothing seems wrongs with then:

Well the results don't look like the are killing your GPUs, and if they were, you'd think something else should be broken by now. It could still be very bad ripple (I doubt that the voltmeter or the PSU tester could pick that up), but with all other values looking ok I'm skeptical about that, and that too should probably have killed not just your GPUs.

This really is a :headscratch:

 

[russcatt]

Did the PC boot up when there was no image displayed or did the machine stop entirely? Because a defective GPU should not usually stop the machine from booting (I've had this happen) unless there's some sort of effect back onto the M/B or (unlikely) the PSU.

The PC was still on but the signal from the GPU had stopped (tried HDMI and DVI) then I restarted with no signal again, after that I booted it up again using intergraded graphics on my main monitor and using the GPU on a second monitor and the GPU wasn’t detected I even checked the screen resolution in windows to see if anything was detected as well.
And Im certain that the cables are fine because I have been using the HDMI with integrated graphics while this problem has been occurring and the DVI seen ok using integrated as well.

 

[alpenwasser]

Well in that case I think you are suffering from a severe case of "WTF-is-going-on-itis?!"

 

[russcatt]

I think that is it, fingers crossed that when I get it back everything is fine, think I will move it to the lower PCI slot for a while and move my sound card above it ruining my pretty cable management , will keep you guys posted if anything goes wrong

 

[alpenwasser]

I will move it to the lower PCI slot for a while and move my sound card above

I actually had to do that once, completely forgot! GPU was not working in top slot (although the slot itself seems fine, tested it with another card), so I had to move it 2 slots down :lol:

Don't know if it'll help in your case, but I had just totally forgotten about that.

 

[russcatt]

Well unfortunately I am back again with some problems, the GPU just cut out on me and seems to do so after a few minutes of being booted up or under load (tested with MSI kombuster),I have removed the card and put it in another slot but that didn’t help nothing. Any ideas of what could be wrong e.g thr GPU or mobo, im getting pretty sick of this happening and don’t want to waste money sending the GPU back to the retailer for them to tell me that’s its fine and charge me for it.

 

[computergeek]

Well unfortunately I am back again with some problems, the GPU just cut out on me and seems to do so after a few minutes of being booted up or under load (tested with MSI kombuster),I have removed the card and put it in another slot but that didn’t help nothing. Any ideas of what could be wrong e.g thr GPU or mobo, im getting pretty sick of this happening and don’t want to waste money sending the GPU back to the retailer for them to tell me that’s its fine and charge me for it.

This is very strange, my hunch would of been a defective PCI-E slot but now I am more convinced it could be a motherboard fault.
Are you able to test your GPU in someone else's PC yourself?:huh:
A quick search dictates it could also be a ram issue.
Do you have an error read out on your motherboard itself, if so what error numbers do you get when the signal cuts out?
Thanks.
I still believe that the most likely hood in all of this is your motherboard is screwed and needs replacing.