Possible failing PSU but not sure

[Ted Striker]

My GPU (GTX 570) has been losing signal every time I try to play a game. It works fine when I stress it with Furmark but games cause it to lose signal. Everything else like audio still works when the signal is lost, and the GPU fan continues to spin (but reverts to lowest RPM), it's only the signal to monitor that fails.

I underclocked my GPU significantly to test some possibilities - after making sure it's not an overheating issue - and games work without losing the signal. Could that indicate a failing PSU? It's a few years old now (Corsair HW 750 or HX 750 I forget which).

To me it's still possible something is wrong with the GPU but since it runs fine even after 40 minutes of Furmark stress testing it has to be something else, like an old failing PSU.

I want to upgrade the PSU anyway, as well as the GPU, but if I can hold off on the GPU for another few months I will. My 570 is fine for now (assuming it's the PSU that's the problem) and I'm hoping for a price drop on the 970.

 

[jimmysmitty]

The only problem with Furmark is that it is a single point of stress. It is great for temperatures and trying to find possible faulty memory but it doesn't fill all the VRAM nor does it give the GPU a wide range of actual tests, as in what the GPU can do.

If the PSU was faulty and the GPU going to a higher load caused this issue it would replicate in FurMark because FurMark will push the PSU to use the most power of all the things you can do to a GPU.

However, if a PSU was going bad and a GPU drawing more power would more than likely cause your PC to shut down rather than cause the monitor to lose signal.

If I were a betting man I would say the GPU is what is going bad. Down clocking it might be causing it to work because higher clock speeds might be unstable, maybe the GPU needs more voltage or maybe it is just old enough to be failing. No way to be sure.

What conditions were you able to get it working at?

 

[ubercake]

Running Furmark should put the GPU in a state of using almost max power for the duration of the test. Have you updated the drivers lately?

The only real way to test the PSU is with a multimeter on the 6-pin/8-pin connectors or use a power supply tester to see if enough power is being sent to the GPU, but it is odd you can run Furmark without any issues.

I agree with jimmysmitty on this one.

 

[geofelt]

See if you can't test with a known good psu.
The HX-750 is a good unit with 7 year's warranty.
Corsair is good about warranty replacement if you have tested with another psu and were successful.
( have had to do this with two CX430 units)

A GTX970 is a nice upgrade. You could even install a GTX980ti on that 750w psu.
Do not get your hopes up that the new price of a GTX970 will plummet. They will still sell at retail prices for those wanting a sli upgrade and must match what they have.

On the used market, the price will drop in time, to be price/performance competitive with newer products.

I am guessing that the gpu is going bad, based on it running when undervolted at lower speed.

 

[Ted Striker]

So you guys think it's the GPU? Hmm.

When I ran Furmark I used 1920x1080 with 8x AA. That's what I use for games as it's my monitor's native resolution. It's not exactly the most stressful res to run Furmark with but I wanted to see if could indeed handle what I do normally.

1920x1080 is a bit stressful for a 570, and here's the thing: last week I tried to run Skyrim for the first time at that resolution (this monitor is new for me), and while it ran OK for a day - with a low framerate because it's Skyrim at 1920x1080 on a GTX 570 - the next day I started getting that "no signal" problem. So I may have pushed the GPU to failure.

That's not the end of the world because I want to upgrade it anyway, but without knowing if it's the GPU or PSU that's failing I don't want to buy a 970 or r9 390 and find out it was the PSU and then have to wait even longer while a new one ships.

 

[jimmysmitty]

All signs point to a faulty GPU over a faulty PSU.

Here is what I would suggest:

1. Do you have a friend whose system you could take a GPU from to test with or test your GPU in? Or maybe use their PSU to test with or test your PSU in?

2. What stores are near you? Best Buy has a pretty decent return policy and while they normally carry crap PSUs they sometimes have half decent PSUs. I would suggest buying one, keeping the receipt and packaging and testing your system with the new PSU. If it has the same problem then it is probably the GPU. If not then it is probably the PSU. Then return the PSU to Best Buy and buy the part you want online.

I still would say GPU over PSU based on your information but the only true way to tell is to swap out one of the possibly faulty parts and test.

 

[turkey3_scratch]

I'm almost thinking a software issue. If the GPU handles Furmark fine, it should handle some games fine, though it may have to do with the fluctuating transient loads while gaming whereas Furmark is a constant max stress.

 

[jimmysmitty]

I'm almost thinking a software issue. If the GPU handles Furmark fine, it should handle some games fine, though it may have to do with the fluctuating transient loads while gaming whereas Furmark is a constant max stress.

That is why FurMark and Prime95 are not a good measure of normal load stability. It is great for max temperature and full load stress but no system is under a 100% load 100% of the time, unless people are folding, so it is never indicative of how a system will behave when loads are changing consistently.

That is why I would assume a faulty GPU but it could be a faulty PSU too, even though all signs point to a GPU over PSU.

Software would be hard. It could be the driver or the GPU could be trying to push the monitor to an unsupported frequency, if so though it should do it on the desktop too.

 

[Ted Striker]

Driver is NVIDIA's latest one from the 28th I think? A few days ago. I also used DDU to clean out any old ones lingering. It's not a driver or software issue, I can almost guarantee it.

I also thought about the resolution but my games are set to 1920x1080, same as my desktop, and they work so long as the card is underclocked.

This GPU is a few years old and has seen a TON of use, and has been pretty heavily taxed since I got this new monitor. With my previous PC it was my PSU that failed (this was years ago), this time I think maybe it's the GPU. I don't have a friend with a PC I can swap components and test with so I think I just have to bite the bullet and buy a new GPU and see what happens. Worst case scenario would be I need a new PSU as well right?

(or worst case scenario could be a motherboard issue. I highly doubt that though. P8Z68 Deluxe/Gen 3 that's been flawless for me.)

 

[jimmysmitty]

The working when underclocked is what is getting me. What is the stock clock vs the underclock setting?

Your PSU can handle quite a bit actually, 750W is a lot of extra power. A 970 wont tax it. Even a 980Ti would run fine on it, in fact a 980Ti can run on 650W.

 

[Ted Striker]

Stock clock is 732 core, 1464 shader, 1900 memory. I underclock it considerably right now to 405, 810, 1325 (minimum in EVGA Precision). Haven't tried settings in between because it's so damned annoying to have to hard reset the PC all the time. If those minimum settings work for now I figure I'll keep them for now.

 

[Ted Striker]

EVGA GTX 790 is ordered and on the way. If it's not a PSU problem this new card will be all I need, and if it is a PSU problem I'll just get a new PSU and wait a few days. I can game with an underclocked card and lower settings in the meantime.

 

[jimmysmitty]

I wish you luck. I highly doubt it is a PSU problem although it is possible. It is just truly hard to tell without a good working piece of equipment to test it out with. Was the hardest thing about diagnostics, if I couldn't swap a part out it made it near impossible for me to say 100% what was the cause.

 

[Ted Striker]

It was my GPU. I uprgraded to GTX 970 and all is well...for now.

 

[jimmysmitty]

Glad to hear you have a solution and I am not surprised that it was the GPU but never know 100% until you test the suspected part.