GPU possibly dead - how can I test this?

[Cadwah]

Heya folks,

So my GPU seems to have died, for the last couple of weeks one of my monitors flickered now and again leading to like 1-2 second drops in signal, I assumed I had a loose connection or faulty monitor but today on 3 occasions while playing ARMA all 3 screens froze then went black, I could still hear music playing so the audio was fine but the displays were non-responsive. Temperatures were in the normal range so 46C when idling and max 80-85C when under heavy game load. I had to restart the comp on each occasion. Then on the final restart when the comp started all screens remained blank. When I plugged a monitor into the motherboard port it works fine, I have tested all three monitors and they all work like this without flickering etc.

Can anyone help me troubleshoot this to confirm it's the GPU, it's just past it's 1 year so I want to confirm before I fork out for a replacement GPU?

Specs:
Case CORSAIR GRAPHITE SERIES™ 380T YELLOW GAMING CASE
Processor (CPU) Intel® Core™i7 Quad Core Processor i7-4790k (4.0GHz) 8MB Cache
Motherboard ASUS® H81I-PLUS: Mini-ITX, LG1150, USB 3.0, SATA 6GBs
Memory (RAM) 16GB HyperX FURY DUAL-DDR3 1600MHz (2 x 8GB)
Graphics Card 4GB NVIDIA GEFORCE GTX 970 - 1 DVI, 1 mHDMI, 3 mDP - 3D Vision Ready
2nd Graphics Card NONE
1st Hard Disk 120GB KINGSTON HYPERX 3K SSD, SATA 6 Gb/s (upto 555MB/sR | 510MB/sW)
2nd Hard Disk 2TB WD CAVIAR BLACK WD2003FZEX, SATA 6 Gb/s, 64MB CACHE (7200rpm)
3rd Hard Disk 2TB 3.5" SATA-III 6GB/s HDD 7200RPM 64MB CACHE
Power Supply CORSAIR 650W CS SERIES™ MODULAR 80 PLUS® GOLD, ULTRA QUIET
Processor Cooling Corsair H100i GTX Hydro Series High Performance CPU Cooler
Thermal Paste ARCTIC MX-4 EXTREME THERMAL CONDUCTIVITY COMPOUND

OS - Win 10

 

[Oussebon]

How do you get your audio? It's not HDMI via monitor speakers / headphones plugged into the monitor I assume?

 

[Cadwah]

How do you get your audio? It's not HDMI via monitor speakers / headphones plugged into the monitor I assume?

No the audio is direct from the mobo.

As an update, this morning, I've taken out the GPU and although it is regularly given a blast of air I gave it a good clean, then reseated it and tested. All seems to now be working as it was although I am still getting one of the monitors dropping occasionally. Temperatures are all below 40C. I ran a Firestrike test and temps went up to 80C and I didn't crash. I have to work now but I'll test it in ARMA later.

 

[NilSatis]

No the audio is direct from the mobo.

As an update, this morning, I've taken out the GPU and although it is regularly given a blast of air I gave it a good clean, then reseated it and tested. All seems to now be working as it was although I am still getting one of the monitors dropping occasionally. Temperatures are all below 40C. I ran a Firestrike test and temps went up to 80C and I didn't crash. I have to work now but I'll test it in ARMA later.

Is it a reference model 970? Even so, temps of 80 on Firestrike seem pretty high, but then you are somewhere nice and hot :sweatdrop: They aren't too high that it should be causing problems, but opening the card up and repasting it would probably give you better temperatures. Having said that, if it is just over a year old many manufacturers will offer a longer warranty on the card than that; it may be worth getting in touch with them with regards to an RMA. In this country they like trying to pass you off to the reseller of the card to do this; but you can normally pursue this directly with the manufacturer depending on what is easiest for you. You can get in touch with PCS and see what they say with regards to the manufacturers warranty. They might be willing to send it back for you; but this will obviously make the process longer.

If you go down the warranty route, don't open the card up as this may void the warranty. Having seen some of the crap stock paste jobs on gpus it would make a big difference I should think, but if you are getting this problem its more likely that some of the memory/a solder joint has failed; and once the card gets up to temperature, things will go wrong. To make sure its still working ok, you need to leave it running during a benchmark such as Firestrike for several passes in a row. They can fail after a few once a temperature has been reached on the vrms etc.

Sometimes the card simply needs reseating like you have done, if it has come slightly loose or there is tarnish or dust on the pci express socket; or you can try cleaning all Nvidia software with something like DDU and downloading the latest driver. If you are error checking etc, I would strongly advise against installing Geforce Experience when installing the new driver; should be able to custom install and not include this. In any case, good luck with further testing!

 

[jerpers]

Whilst we are still in the EU...just, there is a 2 year guarantee that should be honoured

http://www.thisismoney.co.uk/money/bills/article-1677034/Two-year-warranty-EU-law.html

It is a difficult one to get through as it has to be shown that it is due to a manufacturer defect. I have a fitbit surge and have been very glad of this directive as the replacements they gave me only lasted around 4 months before the same fault developed and manufacturer warranties only cover from the initial purchase date.

 

[Cadwah]

OK, so I've had no further issues of the screens going black. Although I'm still getting the flickering on one screen. Temperatures are within normal ranges.

Advice here and by email from PCS advises this is most likely a driver issue however I have GeForce Experience and I update the driver regularly and it is showing as the latest version.

In both instances I have been advised to remove GeForce Experience, just for my own education, why is this?

Advice from PCS:
This Issue appears to be related to your graphics card drivers, to resolve this I recommend attempting a full purge and re-install of the drivers, to do this please download the following files:

Display Driver Uninstaller:
http://www.guru3d.com/files-get/display-driver-uninstaller-download,20.html

Intel Graphics Driver: https://downloadcenter.intel.com/pr...4600-for-4th-Generation-Intel-Core-Processors


NVidia Driver: http://www.nvidia.co.uk/Download/index.aspx


Now to do a complete purge and re-install the instructions are as follows:

1. Download The three files above
2. Extract the folder that is downloaded by Display Driver Uninstaller (right click, Extract here)
3. Double click the 7-zip file it creates labelled DDU and click extract
4. Run display driver uninstaller, please reboot into safe mode when prompted
5. Ensure NVidia is selected at the top of the DDU windows and click clean and restart
6. Once the system has restarted repeat steps 4 & 5 ensuring Intel is selected in the DDU window
7. Once the system has restarted locate the intel Driver that downloaded and run the .exe file clicking through the installer (ensure the WinSAT box is ticked)
8. Restart your computer once the installation is complete
9. Locate the NVidia driver installer and run it, clicking through the installer until completion and then restart your system once again

 

[Cadwah]

OK, I found this, it's not great news, I'll be uninstalling tonight.

http://forums.evga.com/What-do-you-think-of-Geforce-Experience-m2153170.aspx#2153182

Thanks everyone for the help.

 

[SpyderTracks]

OK, I found this, it's not great news, I'll be uninstalling tonight.

http://forums.evga.com/What-do-you-think-of-Geforce-Experience-m2153170.aspx#2153182

Thanks everyone for the help.

I used GFE to apply graphics profiles in game, but the suggested settings are usually complete tripe especially when at 4K. Reading that link, I didn't realise there was any data harvesting going on which puts me off entirely so I will be deleting it as of now. Thanks for the link, rep given.

 

[Oussebon]

Q: What data does GeForce Experience send to NVIDIA?
A: The application collects data needed to recommend the correct driver update and optimal settings including hardware configuration, operating system, language, installed games, game settings, game usage, game performance, and current driver version. If a user is not signed into an NVIDIA account, this data is not personally identifiable. If a user is signed into an NVIDIA account, the data is personally identifiable. All data collected is protected by NVIDIA's privacy policy.

http://www.geforce.co.uk/geforce-experience/faq

 

[NilSatis]

Personally, the data collection wouldn't worry me too much, but quite honestly with Shadowplay being the exception, its a completely unnecessary bit of bloatware, that does nothing but cause problems and add something else to be running in the background all the time. When trouble shooting, its good to eliminate anything in the past that can affect the gpus operation. Also as builds in the past have caused a few problems with some games etc. and with that level of hardware....really you want to be making your own decisions on graphics settings for games. Most modern games that are a decent port have a system detection system now, but for me half the fun with good tech is playing about with setting specific to pc gaming and Geforce Experience eliminates that; sometimes wrongly picking settings. Im not sure if they have changed it now, but it used to add a fair chunk of time onto startup, and use a lot of ram and cause cpu spikes. This could have been solved a while ago; but worth getting rid unless you are desperate for Shadowplay....which is very decent and convenient to have!

That advice you were given seems perfect from PCS, exactly what I would do. Its hard to pinpoint an intermittent problem like this but if you are sure the problem goes away when display is plugged into the motherboard and there is nothing else going on in event manager to suggest there is a problem anywhere else (check this too-- in Windows -administrative tools, event viewer) then it does point to a driver problem or an intermittent fault with the card. Only way to be sure is to keep testing games and benchmarks once you have done the above. My last faulty card would be fine when gaming, then on low clock speeds in Windows 10 it would crash...i.e. when temperatures went back to normal. They can all fail in different ways! Only by running Furmark would it show a crash at high clocks too; my memory on it had developed a fault meaning it would also lose signal on these crashes.

Hopefully your drivers are the problem and a clean install will help. Good luck!