System sometimes booting with onboard gpu instead of discrete.

kevin1212

Distinguished
Oct 3, 2008
191
0
18,690
I have a bit of an odd issue here. Recently downgraded from Windows 10 back to 7 as I had to deal with too many bugs and preferred more stability (issues with hdmi audio as well as some games simply not working). What finally drove me to downgrading was constant game crashes in spite of updating and refreshing drivers. Back to 7 for about a week now and all seemed well when I encountered more crashing. When forced to restart I was getting no video but heard everything else running, swapped the hdmi cable onto the motherboard instead of the discrete card and got video, even though the discrete card was installed and properly seated (and was working 5 minutes ago), couldn't understand it. Windows recognized no discrete graphics card.

Checked the BIOS, nothing odd (btw I'm by no means a pro but have been building my own systems for years). Opened the case to see if the card fans were at least still spinning when I powered on, now to take my side panel off I have to disconnect the fan attached to my side panel from the motherboard. When I did that then booted I got video from the discrete card, figured I'm dealing with a problem with either the card, motherboard or maybe power supply.

I then uninstalled the gpu drivers and installed the drivers that came with the graphics card, no more crashes. I figured that was at least a step in the right direction. So following that I shut down, reconnected the side panel fan and closed up then booted one more time to make sure all was well. Again it booted from the onboard gpu. Knowing the only thing I did differently was reconnect the side panel fan I took it back off and tried one more time, then it booted with the discrete card, coincidence I don't know. As it stands I am booting with the discrete card while my side panel fan is not connected.

I am wondering if I have a power supply problem, had it for years, 600W Thermaltake Purepower, not too shabby but not at the 80plus standards today. Now the discrete card I'm referring to is a recently purchased ASUS r9 390 (about 4 months old), I even had to use the molex / 8 pin adapter that came with it in order for it to work with my power supply. Other info: Motherboard - Gigabyte Z68XP-UD3 LGA 1155, CPU Intel Core i5 2500k, no OC.

Would appreciate any help before making unnecessary purchases just to hope that my diagnosis is correct.

 
That sounds as if you have an issue with the motherboard. If everything was working normally and the power output degraded to below the operational limits of the graphics card, then you would likely have received an error. But it seems to me that the problem is in the detection of the presence of a card in the PCI Express x16 slot. Most likely the BIOS setting for the graphics source is set to Auto. That would result in the system switching to integrated. One thing that you can do is to set the graphics source in the BIOS to the PCI Express x16 slot. But if the problem is the motherboard PCI Express x16 slot, then you can move the graphics card to another slot.
 

Valnac

Reputable
Jun 28, 2014
168
2
4,760
Certainly seems to point to a PSU issue. You did what I would suggest to test that theory (unplug all unnecessary draws on power) and got the results I'd expect if it was a bad/weak PSU. Not enough power so motherboard doesnt see discrete, so boots from on die.That is a power hungry card.
 

kevin1212

Distinguished
Oct 3, 2008
191
0
18,690
@ terry4536
Yea I tried that already, noticed it was set to AUTO and switched it to PCIe x16, still didn't make a difference until I took the panel off and then powered on. I suppose I can try a different slot. I will be forced to update my drivers soon and even if I don't I still expect some crashes forthcoming. The problem is sometimes when a game crashes the whole system crashes with it, that's when I first encountered the problem. I'll consider switching slots the next time it happens.

@Valnac
PSU likely being the problem were my thoughts as well. I'd prefer it to be that rather than the motherboard, quicker and cheaper fix as my sockect type would now be rare. I suspect if the PSU is the issue the symptoms may get worse so I'll keep an eye open, although the safer thing to do would be to replace it so as not to risk damaging any other component. I'll power through for a bit longer in case I get any other clue that it could be something else. Thanks
 
I think that you are correct. Even quality power supplies degrade over time. The power out put after seven years is almost certainly not 600 watts. And 600 watts is marginal at best for that graphics card. Those graphics cards are well known for being power hungry. I would recommend a new power supply in the 750 watt range (an 850 watt PSU would not be out of the question). I would recommend a quality power supply brand like Seasonic, XFX, or EVGA.
 

kevin1212

Distinguished
Oct 3, 2008
191
0
18,690
Hey Guys, haven't posted about this issue for a while, had to be certain. I replaced the PSU about a month ago, bought an EVGA SuperNOVA 750. All seemed well so I figured I solved the problem. Not long ago however I encountered some errors/crashing again. What I can say is that the system is recovering from the crashes without reverting to the onboard gpu, so at least that problem seems to be solved.

I first got some bugs in Starcraft 2, it would black screen with the audio still on. With the old PSU this was the part where I would be forced to reboot with my fingers crossed. Instead it would black screen and then resume working after a few seconds, very unusual. That started happening more frequently and what was even weirder is that after closing the game the GPU load was still at 100%, and you could tell something was up as the system would move sluggishly, I could only get it back down by restarting.

I did some research and thought maybe it was some malware/bitcoin miner but after many clean-sweeps I still had the issue. Not long after that it started happening on other games, meaning it wasn't just Starcraft. NFS Rivals first crashed with a message "DirectX function "GetDeviceRemovedReason" failed with DXGI_ERROR_DEVICE_HUNG..." Eventually it would just crash in a similar fashion to Starcraft. Through out this period I would have been testing various drivers for the graphics card with no success. Some games also just crash with a message saying the driver has failed and recovered.

Leaning to a GPU issue I tried Furmark to stress test the GPU and sadly that crashes as well. After a bit of researching I tried underclocking the Core and Memory Clock, kept an eye on temperatures while running Furmark and I still crash with the temperatures not exceeding what I believe to be acceptable levels (72 degrees give or take).

With nothing left to try I am assuming my GPU is faulty, perhaps damaged from the old PSU. I am simply asking if it could be something else, hopefully software related, that could be the issue.

Would appreciate any suggestion.
 

kevin1212

Distinguished
Oct 3, 2008
191
0
18,690
Just tried the clean uninstall and reinstall. Had high hopes as I actually went a whole 5 minutes into starcraft but crashed all the same, albeit after a longer time. If it is a software problem the last resort would be to do a clean reinstall of windows I guess, unless there is anything else I can try.
 
Have you tried the Windows troubleshooters? The device driver troubleshooter might be of help. You can also check the device manager. The more serious conflicts will show up there with an error icon. Check the display adapter and make sure that there is only one display adapter listed. If it is listing the integrated graphics adapterr and the graphics card adapter then that is a problem.

A fresh install of the operating system can solve many headaches. The discrete graphics card will work much better if it is installed prior to the operating system install. Then the graphics card is incorporated into the OS installation as opposed to adding it in at the end.
 
Another area to look into is system resources. That is a combination of limiting the number of programs loaded at startup and making sure that the memory is sufficient and setup properly.

Many times just reseating the RAM sticks will help immensely. Remove the sticks one at a time , clean the contacts, blow out the slot, and reinstall firmly. Make sure that the RAM is in the primary color keyed slots (consult your manual). Make sure that all of your RAM is showing up in the operating system. Also make sure that it is running at the correct frequency (or overclocked).
 

kevin1212

Distinguished
Oct 3, 2008
191
0
18,690
Ok I'll just check the ram on the system briefly before I reinstall windows, if the reinstall doesn't work I'll reinsert the ram and see what happens. Will work on that later today and report.
 

kevin1212

Distinguished
Oct 3, 2008
191
0
18,690
Yea I have a case speaker, post is fine every time, I checked the ram just to see if all is available and that was ok so I proceeded with the windows reinstall and still crashed with furmark.

So I opened up to reinsert the ram and video card and do any cleaning in between. I noticed however that the extra 2 pins that turns the psu's 6 pin to 8 pin couldn't have been attached properly. I realised it when attaching them after reinserting everything. The extra 2 pins definitely went right down this time. Booted up and furmark is going strong for 15 minuntes so far, will probably go for half hour to be safe. I'll test the games afterwards but it's looking positive. Looks like my own carelessness at fault. I'll report as soon as I've confirmed all is well. Thanks for everything.