Dying GPU or broken drivers?

[belkun]

I've been running into an issue with my GTX 970, which I own for around a year and a half, since last week, and it's getting me extremely worried. One day I was watching YouTube with no issues, when suddenly my audio devices stopped working, which I found weird, so I restarted, and was greeted to only one monitor working on 800x600. Checked the device manager, and it showed my card with the yellow triangle and 'Error 43'. I tried reinstalling the latest drivers and that seemed to fix it. A couple of days later, after restarting to install a software, the issue happened again, this time it seemed to be a lot harder to get it to work again. I updated my BIOS (I'm on a MSI Z97-G45), refreshed Windows 10, and after lots of different tries, DDU cleaning and reinstalls I got it to work again, and it seemed fine for a while, I was playing Overwatch with no overheating, no artifacts and good framerates. Yesterday, again after another restart, I got the apparently broken drivers again, with the same issue: just one monitor working on 800x600. This time I again tried everything I had tried before: CMOS reset, different slots, DDU cleaning, Windows refresh, different computer, but nothing seemed to work. I then actually got it to work again briefly, uninstalling the device and everything related to NVIDIA, restarting, installing GeForce Experience and letting it update the driver, and again, I got it to work, played Overwatch with no issues, shutdown the system, and today it was broken again. Today I went ahead and wiped clean my drive, with a fresh Windows 10 install with Microsoft's own tool, disabled Windows device installation, installed GeForce Experience, it downloaded and installed the latest driver, all of my monitors turned on, things were working, but as expected, as soon as I rebooted it was back to a single monitor on 800x600.

 

I really don't know what else I can try, or what is actually wrong, because whenever the GPU works, it works flawlessly, but as soon as I reboot everything breaks. I thought it could be Windows, but it's happening on a fresh install, I thought it could be my mobo, but updating it and clearing its CMOS didn't work, I even thought it could be my SSD, but it also happened on a completely different computer. Is it safe to assume the card itself is dead or dying? I really don't want to drop a massive amount of money on a new one without trying everything I can and being sure about the issue.

[Mr.Meerkat]

Drivers, new Nvidia drivers have been pretty terrible from what I've head, you've gotta roll back  

[Energycore]

16 minutes ago, belkun said:

I've been running into an issue with my GTX 970, which I own for around a year and a half, since last week, and it's getting me extremely worried. One day I was watching YouTube with no issues, when suddenly my audio devices stopped working, which I found weird, so I restarted, and was greeted to only one monitor working on 800x600. Checked the device manager, and it showed my card with the yellow triangle and 'Error 43'. I tried reinstalling the latest drivers and that seemed to fix it. A couple of days later, after restarting to install a software, the issue happened again, this time it seemed to be a lot harder to get it to work again. I updated my BIOS (I'm on a MSI Z97-G45), refreshed Windows 10, and after lots of different tries, DDU cleaning and reinstalls I got it to work again, and it seemed fine for a while, I was playing Overwatch with no overheating, no artifacts and good framerates. Yesterday, again after another restart, I got the apparently broken drivers again, with the same issue: just one monitor working on 800x600. This time I again tried everything I had tried before: CMOS reset, different slots, DDU cleaning, Windows refresh, different computer, but nothing seemed to work. I then actually got it to work again briefly, uninstalling the device and everything related to NVIDIA, restarting, installing GeForce Experience and letting it update the driver, and again, I got it to work, played Overwatch with no issues, shutdown the system, and today it was broken again. Today I went ahead and wiped clean my drive, with a fresh Windows 10 install with Microsoft's own tool, disabled Windows device installation, installed GeForce Experience, it downloaded and installed the latest driver, all of my monitors turned on, things were working, but as expected, as soon as I rebooted it was back to a single monitor on 800x600.

 

I really don't know what else I can try, or what is actually wrong, because whenever the GPU works, it works flawlessly, but as soon as I reboot everything breaks. I thought it could be Windows, but it's happening on a fresh install, I thought it could be my mobo, but updating it and clearing its CMOS didn't work, I even thought it could be my SSD, but it also happened on a completely different computer. Is it safe to assume the card itself is dead or dying? I really don't want to drop a massive amount of money on a new one without trying everything I can and being sure about the issue.

Hi!

 

Nvidia drivers have been problematic since the 364.xx versions. Please try to see if your card is recognized in versions 362.00 or 361.91. If that works, it was drivers, otherwise you have a GPU issue, probably in its firmware chip.

[belkun]

Thanks for the replies!

 

I had already tried some older drivers, but they were just a couple of months older than the latest. I will try the versions you recommended and report back.

 

In other news, on (yet another) fresh Windows install, again I got it to work, with all monitors running fine, with MSI Afterburner detecting my card and reporting its status. Currently 38C at 0% usage, Fan RPM 1540, Core clock 900MHz and Memory clock 3500MHz, if any of that matters. I actually even managed to run Fire Strike from beginning to end with no issues whatsoever. Got the GPU to 100% usage and around 60C, with no artifacts or anything of the sort. 9340 score, and the result for that matter: http://www.3dmark.com/3dm/12518627

 

That really seals the deal for me on believing the GPU is working fine and the drivers are broken, or as Energycore mentioned, maybe it's something to do with its firmware? And if so, how to even diagnose it and/or fix it?

 

I'm even sort of worried about rebooting because I know it will break again, so I'm not sure if I want to try more stuff like older drivers or just try to get it to work right now with the latest, as at the moment it is again running properly.

[0x1e]

34 minutes ago, belkun said:

Thanks for the replies!

 

I had already tried some older drivers, but they were just a couple of months older than the latest. I will try the versions you recommended and report back.

 

In other news, on (yet another) fresh Windows install, again I got it to work, with all monitors running fine, with MSI Afterburner detecting my card and reporting its status. Currently 38C at 0% usage, Fan RPM 1540, Core clock 900MHz and Memory clock 3500MHz, if any of that matters. I actually even managed to run Fire Strike from beginning to end with no issues whatsoever. Got the GPU to 100% usage and around 60C, with no artifacts or anything of the sort. 9340 score, and the result for that matter: http://www.3dmark.com/3dm/12518627

 

That really seals the deal for me on believing the GPU is working fine and the drivers are broken, or as Energycore mentioned, maybe it's something to do with its firmware? And if so, how to even diagnose it and/or fix it?

 

I'm even sort of worried about rebooting because I know it will break again, so I'm not sure if I want to try more stuff like older drivers or just try to get it to work right now with the latest, as at the moment it is again running properly.

I have no idea how to fix this TBH so I'm shooting in the dark here,

 

Are your drivers starting on start up, or is something stopping it? Like an Anti-virus? Or maybe for some reason it's not enabled on start up.

When you next reinstall the drivers disable any and all anti-virus, malware programs. Enable them after the install then just before you shutdown disable them again so on next start up they will still be disabled.

 

You could also try making Nvidia files read only after you install them?

[scobie56]

6 minutes ago, Mr.Meerkat said:

Drivers, new Nvidia drivers have been pretty terrible from what I've head, you've gotta roll back  

ive had a problem with the drivers as well, my rez keeps changing to 800x600 and sometimes when i play games my card will heat upto 90c and the fans aint spinning 

[belkun]

5 minutes ago, 0x1e said:

I have no idea how to fix this TBH so I'm shooting in the dark here,

 

Are your drivers starting on start up, or is something stopping it? Like an Anti-virus? Or maybe for some reason it's not enabled on start up.

When you next reinstall the drivers disable any and all anti-virus, malware programs. Enable them after the install then just before you shutdown disable them again so on next start up they will still be disabled.

 

You could also try making Nvidia files read only after you install them?

The issue is happening even on a clean Windows install, with absolutely nothing installed apart from the OS itself and the drivers. I'll look into the NVIDIA driver files, not sure if it will really help but it's definitely worth a shot.

 

Just tried running FurMark and checking values on OC GURU II and GPU-Z, and everything is still running perfectly, and again, I know that if I restart everything will break, but I guess I'll do it anyway just to be sure, and then if/when it breaks, try cleaning everything with DDU and installing an older driver.

[belkun]

So, just rebooted, and as expected, as I got back on Windows it was broke again, with Device Manager throwing up Code 43, I rebooted into safe mode, ran DDU, when it rebooted I installed the 362.00 driver, it fixed it, I rebooted again, surprise surprise, it broke, ran DDU on safe mode, rebooted, installed 361.91, and the issue is still present. Whenever I reboot the driver breaks, but if I run DDU and reinstall it, it's fixed. I really don't know what's going on.

[belkun]

Sorry about the triple post, but I managed to nail down when exactly the issue happens. I tried shutting down my system multiple times and every single time I booted it back up again, it was working fine, but as soon as I try to use restart instead of shutdown, things break, and doing the Safe Mode DDU clean, reboot and reinstall gets it to work properly again. Can this mean it's an issue with the motherboard or perhaps even the SSD? It's really puzzling me.

 

As an added note, whenever the GPU is in its 'broken' state, for some reason the Sleep option isn't available on the Power menu.

 

EDIT

 

Well, I'm almost sure I fixed it! Did multiples reboots and shutdowns, and all my 3 monitors are working fine on their native res, Heaven Benchmark is working fine and the Device Manager isn't reporting any issues. It was a BIOS problem. I got a friend's monitor with a DVI port (mine is only DisplayPort/HDMI), downloaded Gigabyte's VGA @BIOS tool, downloaded an updated BIOS file from their website (GV-N970WF3OC-4GD F3, I was running GV-N970WF3OC-4GD F2), followed the included instructions, backing up the current BIOS, connecting the DVI cable to DVI-I, flashing the P file, rebooting, connecting the DVI cable to DVI-D, flashing the D file and then rebooting. I also noticed while doing this that the issue wasn't present while using the DVI monitor, but happened as soon as I tried to boot with any of my DP/HDMI monitors, which makes me believe the BIOS was somehow messing with the ports, however, right now I got 2 DP monitors and one HDMI plugged in and I'm being able to reboot and boot up back again without breaking the drivers.

 

Edited June 16, 2016 by belkun
Fixed

[0x1e]

2 hours ago, belkun said:

As an added note, whenever the GPU is in its 'broken' state, for some reason the Sleep option isn't available on the Power menu.

 

 

Good that it's finally fixed!

 

Sleep is only functional if all drivers and hardware on that computer support it. At the time you had no GPU driver so your GPU would not be able to enter or exit a sleep state, so Windows would disable that function.