GPUs not working anymore, any help would be very much appreciated

[senordingdong]

Hello,

I'm stuck with my pc for nearly a week and can't figure out what's wrong.
I'm using an i5 6500, with an asrock b250m pro 4 mainboard, a gtx 960, and a bequiet SU7 450w PSU;
If I use my graphics card output I only get a black screen. The card seems to be powered one, the LED is blinking, the fan is spinning.If I uninstall the graphics card the system works fine, without any problems. I can access bios via my graphicscard output but can't get it displayed properly (it looks like I would use the magnifiying glass, mybe just low resolution...) but as soon as windows starts the monitor turns black;

I already flashed my bios, tried a fresh install of windows 10, windows 7 and ubuntu 17.04. PCI as primary controller is enabled, I tried out another mainboard (a MSI b250m mortar) and an old graphics card (AMD HD 5770) having the same problem.

Does anyone can help me with my problem? Thank you in advance!

 

[Obakasama]

This seems to be certain from what I can tell.

i5-6500, ASRock combination works fine.
i5-6500, ASRock, GTX 960 doesn't work.

I'm not sure of whether the following combinations are true or not, but I assume you're using the same i5-6500 for all.

ASRock and HD 5770 doesn't work?
MSI and GTX 960 doesn't work?
MSI and HD 5770 doesn't work?
MSI works?

Given the hardware you have access to, there are a few combinations you could try to narrow down the potential issue.

One thing more specific to both motherboards is the other PCIe slot. Not ideal as it runs slower, but if the graphics card works in the secondary slot and not the primary then it would seem to suggest the PCIe slot is the issue (graphics card has to be working, of course).

If you have other cables to use then you could try swapping around to see if there's an issue there with the specific port of the graphics cards. Though from what I can tell, it doesn't sound like an issue.

 

[senordingdong]

Thank you for your reply!

Yes I'm using the same i5. i5 works with both mainboards. Neither mainboard works with gtx 960 or hd 5770, so i guess it is not an PCIe slot issue, as both mainboards (Asrock completely new) would would have a defect pcie port. Unfortunatley I can't try out the other pcie slot dew to my limited space in my case....

I tried out the dvi port of both graphic cards and the hdmi port of both graphic cards, neither is working....

 

[Obakasama]

Neither graphics cards are shown to work. That makes it trickier then. It is possible, though fiddly, to remove the motherboard and boot up on cardboard. This would allow you access to the other PCIe slot temporarily for troubleshooting purposes. I would agree it seems unlikely both motherboards to have the same fault, but best to eliminate it as a possible cause if possible. Though this would still need a known working graphics card.

You mentioned with the ASRock and GTX 960 combo that you can see BIOS (shrunken down?), but black screens when it tries to boot into Windows; do you have many USB peripherals plugged in? I would suggest removing as many as you can to see if it will allow boot up. (I had a USB stick in a particular port preventing Windows boot up before.)

On the shrunken BIOS screen: http://www.tomshardware.co.uk/forum/259042-30-screen-shrink-problem-bios-post
Something about autoscaling for the monitor itself. Might be worth checking out.

 

[senordingdong]

Hi!
It was possible to enter the bios with the Asrock and GTX combo and with the MSI and GTX combo but had graphical issues displaying bios properly with both.

In the meantime i tried removing one Ram stick and it worked again! I also tried the other one and it also worked , then i tried both stix together again and it worked for a short period of time until i plugged in the usb cable of my speakers (for power supply). Then my screen flicked (with funny colors - kind of "retro game" style) and turned black again.

Then I again removed a ram stick and put it in again and now it seems to work with speakers connected, allthough I am not sure how long it will......

Could there be an issue with my psu not having enough power to support all my components (though it worked for 1 1/2 years supporting them without any issues) or with the ram modules?

 

[Obakasama]

I did think PSU at the beginning, but those are usually difficult to figure out. Ideally it would require electrical testing, but a tentative software check using HWMonitor may help to identify the worst case scenario. It can give figures for the voltage rails, and my hope is it would detect the largest deviations from the voltage it should be. A large deviation would suggest to me a problem with the power supply. But the wattage suggests to me that it should be sufficient to power the graphics card if working properly.

RAM sounds a bit suspicious. To eliminate RAM being a potential cause you could download MemTest and leave it to run overnight for 8 passes. That should suffice to discover any problems. (Technically it should flag up errors for the RAM and slot combination, so an error may require additional testing.)

With the speaker... I mentioned about my experience of a USB stick being plugged in causing boot up issues. You could try booting up with the speakers plugged in (and plug in after boot up), or another USB port to see if the problem persists. Also... not sure on the colours, but that seems to suggest possible electrical interference (at least that's how I see it).

Did you have a look at the monitor options on autoscaling? The BIOS screen still sounds like that issue in the link rather than a graphics card one. What happens is BIOS screen tends to look low resolution on full screen (because it is). Drivers load with Windows correcting the resolution for normal use.

 

[senordingdong]

Hi!
Yesterday it worked fine without any problems, I could even run a more or less graphically demanding game (ac unity) without any problems.
This morning I had the same problem again...

I may be able to test another PSU of my friends computer... If I would run the HWMonitor i would have to run it without my graphic card plugged in. Would it still be able to detect large deviations?
I also will try to run the MemTest over night.
Could it be that ram removal /reinsertion does an automatic bios reset. Because i also noticed that when i install/unistall the device and do a bios reset it seems to work for the first time. After some time it stops again.... (with MSI and Asrock Mainboard) and it does't work all the time, which makes me more unsure about this......

I didn't find autoscaling in my monitor (it's a fairly old Samsung 2333hd) not sure if it has it. Though when i was able to use my graphic cards I was also able to access bios without any resolution changes (so i had full resolution /no graphical issues in bios).

Thank you again for your time!

 

[Obakasama]

HWMonitor should detect the supply of voltage. My guess would be a heavier load would perhaps mean higher fluctuations (if any). If the power supply was faulty, I would hope it would be detected regardless of load and electrical draw. If able to, use this as a guide to see if the voltage you see are safe: http://i.imgur.com/iR93jqB.png

RAM removal shouldn't result in a BIOS reset (not to my knowledge). Removing the CMOS battery would reset the BIOS back to stock settings (at some point it might even be worth trying). Let's see whether MemTest finds anything. If it doesn't we can rule out the RAM being a possible cause.

Additionally, I'm not overly convinced it's a driver issue but it could be worth trying at some point. A clean installation of drivers is something often recommended: http://www.tomshardware.co.uk/faq/id-2767677/clean-graphics-driver-install-windows.html

Related to the driver, I would wonder if you have the same boot up problems if trying to boot into safe mode.

 

[senordingdong]

I ran MemTest and it didn't find any errors.

I used HWMonitor and apparantly I don't see my voltage output. I uploaded the textfile log here:
http://txt.do/dq5cw

I also used HWinfo64. I was not able to create a txt-file log, but I did a screenshot of the mainboard voltages but having trouble reading and to interpret them. Maybe could tell me if the stats are correct?
https://imgur.com/DiJCiFR

as for the driver. I did a clean install of the drivers (and i already tried an older driver) after using DDU) in the past. Now i am working on a freshly installed windows 10 system (and tried a freshly installed windows 7 and ubuntu 17.04 system). Though If i install a new windows version i have a graphical output through my graphics card during the installation process (allthough I don't get full resolution - I do however get the full resolution using the internal graphics.... like I do, when i access my bios). However when windows (or ubuntu) starts do boot up the monitor turns black again...

 

[Obakasama]

At least the RAM is of good working order then, so that can be ruled out as a probable cause.

From what I can see with the HWinfo64 screenshot, the +12V rail of the PSU is being read as being at a minimum of +4.8V and maximum of +6.336V, which are out of the tolerance allowance of voltage variation for a PSU. It should be within 5% either way, so the variation worries me.

I did mention that the best way to check a PSU is actual electrical testing, but the variation shown here is too great for 'margin of error' software readings (as I would believe). The things you've seen does fit with an underpowered PSU. If Windows requires a bit more graphical power, then the system might be drawing more power than can be supplied.

I would personally look towards a new PSU at this point for peace of mind. At the very least, see if you could borrow a good working PSU to see if the boot up problem occurs with the graphics card installed. I suspect you won't see the problem with a good working PSU.

EDIT: on HWInfo64, look at the motherboard info. Those yellow lightning marks indicate power, the first one (+12V) is the one I'm looking at with concern.

 

[senordingdong]

Ok, thank you!
I will try to get another PSU to test and report afterwards. maybe will take a few days.....

 

[senordingdong]

Reporting back...
I tried out another corsair 650w psu. I reinserted the graphics card, booted up and it worked fine. HWinfo and my device manager recognized the card with the drivers installed and didn' show any problems. After shutting down my pc and booting again, same problem again.
I ran HWInfo64 again and it shows more or less the same 12v values as with my old psu....
I also tried out an old hd 6950 of a friend and it seems to work for now, though I don't know if it will keep on working...

 

[Obakasama]

Hm... running out of suggestions I can make.

I would find it difficult to believe this Corsair PSU also has the same voltage issue, though not impossible (just highly improbable). It would be tempting to conclude it's the motherboard power connector being at fault (but that would be hasty without confirming the viability of the power supply). I would be tempted to see if the other motherboard gave the same readings but that would be very fiddly...

The other point I can think of being a potential issue is the power cable to the graphics card itself (if indeed the card itself requires one). Nvidia's site seems to suggest the GTX 960 requires a 6-pin power connector, so I would expect it would be the case for the variations of the chip. That could be something to check on (but I am grasping at straws).

 

[senordingdong]

I don't think that the Corsair PSU has some issue, I am also not completely sure that it is the motherboard connector because i did have the same issues with the older MSI motherboard (though i didn't run hwinfo to check the voltage with this motherboard).

Yes the GTX 960 requires a 6-pin connector.
The HD 6950 is running now for four days straight without any problems, so maybe it is a problem with the 960 power connector, allthough this wouldn't explain why the hd 5770 didn't work for me.
I will borrow my gtx 960 to a friend so he can try it out to see if it works for him.

 

[Obakasama]

That certainly sounds plausible with the GTX 960 power connector, which would narrow it down to a graphics card issue. Certainly puzzling. And I am still hoping someone can identify something I've missed to provide more help.

But if your friend can test the 960 in their PC, it would help see if it is indeed a problem with the graphics card.

 

[senordingdong]

So my friend has been testing my GTX 960 for the weekend and it ran fine. Didn't have any problems at all. Likewise I have been running his hd 6950 for a week now without any problems.
Could there be a problem with the Monitor i am using? Maybe that the graphics card supports different frequences or anything like that? It is a fairly old Samsung Syncmaster 2330hd and it worked with the gtx 960 before....

 

[Obakasama]

The HD 6950 purportedly draws more power than the GTX 960, so it should mean the PSU is clear of being the cause.

I came across a review for that monitor model stating set up can be problematic. http://www.techradar.com/reviews/pc-mac/monitors-and-projectors/monitors/samsung-b2330-hd-924251/review/2

Other threads suggest the frequency of the monitor can also play a part in how something is displayed (though why it should work sometimes doesn't make a lot of sense). One thread (for another monitor) mentions underscan of ATI graphics cards: https://www.cnet.com/forums/discussions/syncmaster-sa350-screen-resolution-problem-524162/

Insofar the hardware seems to be fine, we may have to look at drivers (again) and settings (monitor and graphics card). At least, that's how I would proceed.

 

[senordingdong]

Ok, I maybe able to test my friend's Monitor with my computer next week...
In the mean time, how could i proceed to test the Drivers and Settings?
I already flashed my BIOS to the most current Version, I also flashed the graphics Card bios to the most current.
I did a fresh Windows 10 (and Windows 7) install and installed the most current Drivers directly from the nvidia website.

When I was able to use my graphics card, the current Drivers were displayed as correctly and "ready", I was also able to activate multi-GPU-Monitor in my Bios and start with my integrated Graphics card, then I don't have a Picture via my gtx 960 but in System Manager it is displayed as "ready and functioning".

Any ideas what i could try next?

 

[Obakasama]

I was thinking about whether there's a driver for the monitor itself. (I've seen mention of firmware update for other issues with the monitor, so I'm not sure how relevant it would be.) You could check using Device Manager to see if there's a newer driver available.

At this point it doesn't sound like a hardware issue (at least there's insufficient evidence to confirm it), so we're left with trying to tackle this via software. You've tried most of the things I'm aware of already which doesn't leave many options.

I do wonder whether the GTX 960 will work in safe mode though (would usually indicate a driver problem if it works in safe mode but not a normal boot up). Then there's the slight possibility of other software being the problem (in which case, a clean boot may help to identify it).

Grasping at straws of what can be tried and trying to see if there's anything glaringly missing from this troubleshooting...

 

[senordingdong]

When I got the gtx 960 back I reinstalled it and it booted fine (i was able to play for a whole day) the next day when i restarted my pc it turned black again.
I downloaded the lastest monitor drivers and it detected my graphics card and installed normally.
Maybe something with the bios settings? I made a photo of my current settings wondering if i missed something:
https://imgur.com/a/dUMK2

Booting in safe mode works with my graphics card, so it should be some driver issue? Though I don't know what to change.....

in addition: i used ddu to unistall the latest drivers, and windows booted normally, when i reinstalled the nvidia drivers it tourned black again... so i guess there must be something with the drivers, though i still don't know how to make it work....

 

[Obakasama]

I do recall you mentioning you clean installed drivers before but it didn't help. But if safe mode works fine... I would start thinking it's the drivers (though there have been cases where others disagree about the validity of this assumption). The other option for drivers is to install an older one in case it is the newer driver which is causing problems. I haven't asked, but did the system work smoothly before the problem? (Thinking if it did, then perhaps installing the last known working driver may help.)

I can't see much wrong in the BIOS settings (looks like PCIe graphics are prioritised) and the other options seem to be on, what I assume to be, default settings.

 

[senordingdong]

Yes the system worked normaly about half a year ago, i have tried to install older drivers from 2017 but the screen stayed black, through system manager recognizes the card and drivers as functioning. I also tried to install older drivers from 2016/2015 but i get an error report during installation (didn't try every driver though...) still wondering which of my components is causing the problem, because it worked with my friends pc....

 

[Obakasama]

It's very strange given everything we know so far. Each individual component has been verified to be working independently. BIOS has been updated to ensure compatibility. Clean driver installation method was used to be on the safe side. Yet a problem persists with your specific combination of hardware and software. Had your graphics card not worked on your friend's computer then we could say the graphics card is at fault.

As mentioned I'm running out of ideas to try and narrow down the issue. Last few things come to mind in which to try and see if there is something interfering with the system.

Clean boot: https://support.microsoft.com/en-gb/help/929135/how-to-perform-a-clean-boot-in-windows

The idea of using a clean boot is to see if there's another piece of software causing an issue. Not usually used for troubleshooting a graphics card problem but I can't think of any other common fixes (that I'm aware of) we can try.

The thought also occurred of a small possibility of a Windows update having caused some issues, but I would have expected those to have been fixed subsequently. Searching through my bookmarks: https://wccftech.com/fix-windows-10-nvidia-driver-conflicts/ I'm not sure how applicable that is now though.

We've sought a solution for the last few weeks with no additional input (which is rare). I would like to suggest posting on the graphics card board as well with a link back to this thread for reference of what we've tried so far. Perhaps someone there will have an insight I'm missing.