GPUs failing.. what's wrong?

guil3

Reputable
Jun 17, 2014
4
0
4,510
Config
CPU: Intel i5-2500
Mobo: ASRock Z68 Pro3 Gen3
RAM: Corsair Vengeance 2x8GB
HDD: Seagate Barracuda 1TB
PSU: Corsair CX500

I've been using a Radeon HD 5830 for the past two-and-half years. It was meant for my gaming and video editing, but the card wasn't really ever pushed to its limits. There was no instance of graphics crashing or overheating.

In the meantime, I discontinued using a UPS sometime last year. It kept tripping off occasionally, and it became a big deal to have it working normally after a powercut. The electricity supply is not so erratic, and hasn't damaged anything in this house.

A few weeks back, I returned after a month's vacation, to boot the system. It booted into Windows, and that is when I saw a few dots starting to appear. The dots spread in a few minutes of usage. Then, the screen tore up and froze. Reebooting would not help - the tearing was too much, and after a while the card stopped sending any display to the monitor (monitor would fall into standby). The card would give a glimmer of hope when it would output normally in 1 try out of 20, but that's it. It's as good as dead.

Since, I had been relying on the inbuilt HD2000 graphics on my CPU. It worked fined on lesser loads. But anytime I opened a graphic-hungry app like Photoshop or Sony Vegas, it would trip the Intel Graphics adapter, which would either reset with a warning message, or simply lock the system forcing a restart. Even Google Maps, which uses OpenGL would do that.
Monitoring the temps, I couldn't find an instance where it would've gone beyond 52C.

Two days back, I replaced the HD 5830 with a nVidia GTX 650 Ti. Power-wise, the 650 Ti draws lesser power than the 5830, and has a lower TDP, with a single external power connector needed (5830 needed 2). The card worked perfect upon install. However, into two days of use, and during normal browsing the screen froze. That's when the graphics driver started crashing. Inside a coupla restarts, the worst happened - screen started tearing. Now this card is also unusable like the previous one. Dead again.

To eliminate external factors, I got the UPS fixed and its back on the grid. One change I noticed that the onboard HD Graphics are crashing a tenth of the time than they used to. The 650 Ti also started working fine soon as the UPS was installed.
I though the problem was fixed by the simple addition of a UPS, but the screen again tore up (and became the same condition) after an hour's use. So I fell back to onboard HD graphics, which work very slow but stable.

Before I purchase another graphics card, I need to figure out what's wrong. Anybody have clues which component could be faulty here? GPU? Motherboard? PSU?
500W should be enough for my machine (TPU requirement adds up to ~300W); even the PSU is 80+ certified.
 
Solution
The power supply voltages aren't too bad. 12V is a touch over where I'd like it (12.25) but OK.

The only thing I see wrong is your cpu temps are too high (57C). They should be less than 40C for non arduous use.
Blow out your case - especially fans, filters and coolers.
I couldn't see your cpu cooler fan rpm. Make sure its around 1500rpm.

If that doesn't help. you need to re-seat your cpu cooler -
Remove the cooler. Clean the old thermal paste off it's base and the cpu with isopropyl alcohol.
Put a pea size drop of new Arctic Silver 5 thermal paste on the cpu.
Refit the cooler.

If that doesn't help.
Replace your stock cpu cooler with a custom cooler eg Coolermaster 212 Hyper X - if it'll fit in your case.

snowctrl

Distinguished
There have been concerns voiced on these forums that the Corsair CX PSUs are maybe not such a good choice... from what you've said I would be suspecting that it is reaching the end of its life. Do you have an alternative PSU you can try?

Also, rather than using the UPS, you might try a simple surge-protected power strip? And make sure you try plugging your whole rig into a different wall socket, just in case....

Consider this list when sizing up a new PSU:

http://www.tomshardware.co.uk/answers/id-1804779/power-supply-unit-tier-list.html
 
Sounds like it'll run long enough for you to download HWInfo and run Sensors and post the results - 3 screenshots onTinyPic or similar.

I'm suspecting your power supply as we've had trouble with cx's before here. But just to make sure - post the results and we'll check.
 

guil3

Reputable
Jun 17, 2014
4
0
4,510
Hi, I've posted HWInfo screens for my current (surviving) setup, with two cards, of amd and nvidia, dying in similar fashion inside a coupla weeks.
http://pho.to/6E3Ol

I also suspect the PSU to be to blame here. But what could be causing the sudden burnout? I will be switching to a bigger wattage PSU, a friend's, and try a new card (nVidia) tomorrow. Will keep updated.
How much probability could there be of a faulty PCI port or the Motherboard itself?
 
The power supply voltages aren't too bad. 12V is a touch over where I'd like it (12.25) but OK.

The only thing I see wrong is your cpu temps are too high (57C). They should be less than 40C for non arduous use.
Blow out your case - especially fans, filters and coolers.
I couldn't see your cpu cooler fan rpm. Make sure its around 1500rpm.

If that doesn't help. you need to re-seat your cpu cooler -
Remove the cooler. Clean the old thermal paste off it's base and the cpu with isopropyl alcohol.
Put a pea size drop of new Arctic Silver 5 thermal paste on the cpu.
Refit the cooler.

If that doesn't help.
Replace your stock cpu cooler with a custom cooler eg Coolermaster 212 Hyper X - if it'll fit in your case.
 
Solution

guil3

Reputable
Jun 17, 2014
4
0
4,510
Hey guys, I'd just like to update on the issue.
Since losing the GTX 650 Ti, I have kept my system under check. I have been running a GTX 645, a card which no reference benchmarks, but one which has been running smooth. It has TDP of mere 56W, hence not needing an extra power connector, and reducing the power load.

Personally think taking it off the UPS was biggest mistake. Since reconnecting the (annoying) UPS the card has stayed sane. The only time I had my heart skip a beat was when I took the system off the UPS, and predictably enough the display corrupted. It was saved because of switching off the mains immediately. It is understandable that the power supply/phase to the system isn't reliable, and spikes and surges were what was doing the graphics card every time. Why only the graphics card, I dunno.
 

snowctrl

Distinguished


If a UPS is annoying, will a simple surge protected powerstrip not do the job?