Sudden black screen; after BIOS boot, no OS desktop boot, no cursor

[jellybutt]

Hi all,

Once again facing an issue, yet this time more problematic.

 

TL;DR: PC will not boot past BIOS and Windows 10 boot screen (the blue logo). Tried clearing RTC RAM, single RAM stick, different DIMM slots, Mem Ok!, Windows problem solver, 2nd BIOS setting on GPU. No results.

 

This issue arose yesterday evening. The situation was as follows:

I started BF4, played about 1 hour. Halfway through I set my GPU profile from stock to a mild OC, to see if it helped bump up the FPS.

After BF4, I stuck with webbrowsing (emails, youtube etc.). After 2 minutes after shutting down BF4 and switching to light usage, I got a sudden black screen. The monitor was still on, but no cursor and no desktop.

I rebooted by using the reset switch on the PC case, no difference. Also tried by holding down power button, also no result.

 

What happens is:

1. Press power button, PC bleeps and start as normal
2. Get BIOS screen, finishes succesfully
3. Get Windows 10 startup screen, finishes succesfully
4. End up with black screen, no cursor, no sound
5. No response to keyboard, mouse etc.

See video:

 

What I have tried so far:

1. Switch between BIOS settings on GPU (as I thought I fried my GPU somehow with the mild OC)
2. Single RAM stick
3. Different DIMM slots
4. Mem OK!
5. Clearing RTC RAM
6. Windows problem solver (the blue screen you get when Windows can't start up properly, right after BIOS screen);
   • Can't find a solution by auto-solve
   • Have NOT tried a recovery point yet
7. Booting from old Windows 7 Ultimate 64bit on WD Green drive, but it seems this OS does not work anymore somehow(!?) when setting this disk as boot priority. Windows files are still there though, so I don't get it...

System:

OS: Windows 10 Pro 64bit

CPU: AMD Phenom II X4 965 BE @ stock 3.4Ghz

GPU: Sapphire R9 290 Tri-X OC @ stock 1000Mhz Core and 1300Mhz Mem (mild OC @ 1075Mhz Core and 1450Mhz mem)

RAM: DDR3 GEIL 2x2 GB @ 1333Mhz + DDR3 Crucial Ballistix Sport 2x4 GB @ 1600Mhz (same timings, same voltage)

MoBo: ASUS M4A87TD/USB3

PSU: Corsair RM850

Storage: 840 Pro 256GB (boot; 2y 4m old); 850 Evo 500GB (7m old); WD Green 1TB (±6y old)

 

Possible problem-causer:

I have found a somewhat similar topic, of which the suggestion and possible outcome scare me... 

It could be that my SSD is fried, but I'd like to hear some other options I might be able to try.

I am not capable of trying the HIREN BootCD this weekend, as I don't have any empty disks at hand.

 

I would love to hear some other things I could try, as this of course happens a few days prior to project and essay deadlines...

Thanks in advance!

 

JellyButt

 

[NunoLava1998]

I think it's nothing that scary. Likely means that your Windows 10 has something wrong in a dll file or something.

[NeoFrux]

Yeah, it sounds to me like there's something up with your graphics driver, like maybe it's applying the OC at bootup and your GPU actually isn't stable with it. Try resetting the PC while Windows is loading (still have the spinning dots) a couple of times and eventually you'll get to startup repair, where you can get to Safe Mode. Load up into Safe Mode, uninstall your display driver and your OC software, and try booting Windows normally again. If it works, go ahead and reinstall your graphics driver and see if everything is stable.

[jellybutt]

21 hours ago, Runefox said:

Yeah, it sounds to me like there's something up with your graphics driver, like maybe it's applying the OC at bootup and your GPU actually isn't stable with it. Try resetting the PC while Windows is loading (still have the spinning dots) a couple of times and eventually you'll get to startup repair, where you can get to Safe Mode. Load up into Safe Mode, uninstall your display driver and your OC software, and try booting Windows normally again. If it works, go ahead and reinstall your graphics driver and see if everything is stable.

Alright, I'll give that a shot. I had the same thought, hence my tries of switching the hardware switch on the card itself. My only comment on this is that, with a 'defective' GPU, wouldn't the screen display nothing at all? Or is the BIOS+Windows boot screen not displayed/powered by the GPU...

 

Will update here with the results.

Thanks!

[jellybutt]

22 hours ago, Runefox said:

Yeah, it sounds to me like there's something up with your graphics driver, like maybe it's applying the OC at bootup and your GPU actually isn't stable with it. Try resetting the PC while Windows is loading (still have the spinning dots) a couple of times and eventually you'll get to startup repair, where you can get to Safe Mode. Load up into Safe Mode, uninstall your display driver and your OC software, and try booting Windows normally again. If it works, go ahead and reinstall your graphics driver and see if everything is stable.

Alright, I made some progress.

1. Booting in safe mode works. Uninstalled all AMD driver related stuff with their driver uninstall utility. Uninstalled MSI Afterburner. Checked the clocks with GPU-Z, they were back at normal (1000 and 1300).
2. Rebooted, this time in 'normal' mode. Reinstalled the latest drivers from AMD (16.10.3). Also note that, prior to this problem occurring, I had something like 16.8.2.
3. While installing, I got the same problem. Normally you get a few black screens, I'm used to that. But the third time the black screen transition was different. I saw some white lines and strips. Similar to this effect (http://wall-papers.gallery/images/cracked-screen-background/cracked-screen-background-3.jpg), but in a lesser quantity. Again, black screen and a non responsive PC.
4. Repeated the process, but tried installing an older version Driver (16.1). Install process was completed this time, but fairly quickly and with the message that it was installed. However, I cannot find any Crimson or driver related programs. I'm guessing the installation failed.
5. I needed to get some work done, so I fired up SolidWorks. In the meantime I finished up something else, leaving the program open/idle for about 5 minutes. When I came back, I had the black screen again...

I am currently using my PC without any drivers, and boy can I tell the difference when trying to set up a render scene... Video playback (YouTube) is also heavily affected, as it is now processed by the CPU. And an eight year old CPU with 4 cores is not that good at 1080p60, I can tell you that much.

 

So I started looking online, and one of the first hits on Google was this topic: https://community.amd.com/message/2743721

On the bright side, apparently I'm not the only one. That does not solve my problem though. On the darker side, it appears that there is no real fix for this. Leaving me with a relatively unusable PC in times when I need it the most.

 

If anyone has any suggestions, fixes or other helpful information, I would be delighted. But for now it seems like this problem is beyond fixing, with a simple fix that is. I will be trying some of the suggestions in that thread. If I manage to make some progress, I will post the results here.

 

Thanks!

[NeoFrux]

It sounds like a video memory issue, or an issue with the memory controller on the card... Ugh, what a nightmare. I had no idea this would be so widespread... Did AMD's OEM's pick up a bunch of bad memory or something?

 

Anyway. Do you have another GPU you could test with in the meantime just to rule out the board or some other problem?

[jellybutt]

No I do not. I had to move in quite a hurry and was limited in the PC stuff I was able to bring with me. Otherwise I would have had an old HD5850 to test. I might be able to ask the campus IT service center, they could have one laying around I can test. Otherwise the next possible date I might be able to test with a different GPU, is after 21st of November.

 

I agree, this is not an ideal situation. Also, I do not have warranty on the card anymore. So no RMA'ing this time around. I could perhaps contact Sapphire to see if they could be of use with this problem, but I think I already know the answer to that question.

 

But, the PC has been running somewhat stable now for 90 consecutive minutes. No real GPU power though.

 

And here I was thinking I was better off buying a relatively good card when I had the money, reducing the risk of any faulty things. But in retrospect, you never really know with technology.

[NeoFrux]

Sadly, even the highest end has its issues, although QC is usually better after a certain point. Hopefully the campus IT centre has a spare they can try in it and rule out a larger issue, not that a bad GPU is a good thing.

[jellybutt]

Agreed, but that is more easily fixed. I have been thinking about upgrading anyway, however that plan included everything BUT the GPU...

I will post the outcome when I get to it. These next few days/weeks are quite packed.

 

Thank you so far with all the replies!

[NeoFrux]

Look forward to hearing the verdict on the card! If it's not that, we have a lot more digging to do.

[jellybutt]

On 30/10/2016 at 4:11 PM, Runefox said:

Look forward to hearing the verdict on the card! If it's not that, we have a lot more digging to do.

Alright, time to share the progress of this investigation.

 

TL;DR: stress test on clean system revealed no black-screen crashes, but also very low performance. Testing it again on my system now includes the problem of not even detecting the GPU and using the Microsoft basic display adapter for everything, even with drivers installed etc.

 

I've had the chance to test at the IT service center at my campus. The did not have a open test bench or something similar, but were eager to help me with this test. So they found an unused Dell factory-system with a Mini-ATX and 3770(k?). Wonderful people, the were genuinely interested in this (and the massive card, they are not used to seeing that). Unfortunately, the included powersupply of 250W was not going to cut it for me GPU, so we used mine. Meaning that I was not able to exclude this variable when testing. We tested this on a relatively fresh Windows 7 Enterprise 64-bit OS, running the AMD Driver Cleanup Utility to be certain and installed the latest Crimson driver (16.11.3 Hotfix).

 

The results are as follows:

• The system recognises the GPU;
• We stress-tested it using FurMark on all extreme settings (1080p, full anti-aliasing etc.);
• It only outputs around 18-20 fps on 1280x800;
• It did not crash with a black screen during the testing period;
• Clocks and temps were as expected and normal.

After that, I thought I'd give it a shot with my system:

• I cleaned all drivers and installed an old Catalyst version (I believe 15.7);
• Installation successful;
• I started Unigine Valley, it started with 18 fps and a black screen and continued the benchmark itself at 2-3 fps (which was actually 0.2-0.5 fps, as it stuttered).
• Tried again with driver cleanup and installed the latest Crimson;
• Same story.

So I looked at Device Manager, GPU-Z and Catalyst, all stating that the recognised GPU was the Microsoft basic display adapter, not my ATI graphics (also not listed). Weirdly, GPU-Z does recognise that it is an AMD card, including all specs (hawaii chip etc.). Also got these messages when Valley crashed, confirming it is using the Microsoft basic display adapter:

  

 

I performed several potential fixes, suggested on the web:

• Cleaned up the PCIe slot and re-inserted the GPU properly;
• Unplugged and plugged in the 8+6pin power;
• Switched between PCIe and PCI priority, and back again, in the BIOS.
• No results, the system simply won't detect my GPU.

I have been able to film the black screen as well. It seems to happen every time it wants to install the AMD Display Driver part (not sure if that is the correct translation). Also, it sounds like the entire PC freezes, as I can hear the familiar sound of the fraction of a second of the last frequency and tone, looping for about 5 seconds. The system stays on though and does not auto-reboot or signal any errors (I have the little MB speaker attached).

 

 

I also found a small notch on one of the PCIe pins, but I'm fairly confident that this is not the cause of this problem.

 

 

So, what's the next step?

I am planning to stress-test the GPU again next Monday or Tuesday on the test-bench we made at the IT service center, letting it run for about 8 hours.

I can also only exclude the GPU as problem-causer after next weekend, as I'll receive my older GPU for testing then.

[0x1e]

4 hours ago, jellybutt said:

<snip>

I'm sorry to say (Mine might be different) but your GPU is dead.

I JUST had the same problem around 1-2 weeks ago on a 980ti Kingpin.

 

It would go into bios without a problem, safe mode no issue, but once it loaded Windows normally it would black screened exactly like yours.

It takes out the entire system, not just the GPU. Mine started as soon as I installed a new GPU driver.

 

What I tried,

DDU,

Default CPU clocks,

One stick of ram,

48 hours of memtest (that included all 4 sticks)

Windows format,

Another GPU (worked kinda)

Another PCI-e slot,

Different PSU cables,

PSU tester,

Another PSU,

Another computer (4690k, Asus Sabertooth Mark 1, 1050watt PSU, 8GB 2133Mhz ram)

Another SSD.

 

Now I'm on my iGPU waiting for EVGA to do their tests.

 

My gut feeling is the GPU is damaged and has taken the PCI-e slot with it.

So I may have to upgrade my PC which my wife won't be happy about, but it's worth fighting for!

[revathetaker]

I have been trying to trouble shoot a black screen with a new GPU. So, I've been reading a lot on many forums. First of all, I'm kind of a noob. Was tempted to reply on your post because I've read a post on tomshardware with the same problem as yours and I cant remember if he/she managed to find a fix. I tried finding the link to it with out any luck, as the threads were pretty long. Though I'll try looking for it once I'm back home from work. Updating BIOS, switching between the PCIE slots on your motherboard (using the slower *8 instead of *16) or if you have more than one PCIE * 16 slots, were some of the fixes that I remember reading( not sure if the fix was for the exact same problem). I'm currently convinced that my PCIE slot is corrupt/ damaged as my new gtx 1050 ti is giving me the same problem that my old card ( which some one convinced was defective - gtx 650 ti) and a small probability that it might be the power supply. From your tests PCIE slot on the mother board as you tested your card on another board. PSU problem - maybe. I can say more but it's going to be very vague, so I'll try finding that link for you.  

[jellybutt]

13 hours ago, revathetaker said:

[...] I can say more but it's going to be very vague, so I'll try finding that link for you.  

Thanks for replying! Of some things I am fairly confident of ruling out, such as a bad PSU. But please read my findings below.

I'm still interested in your findings though, please post them!

 

 

In the mean time, I've had the opportunity to test again with the 3770 test bench. The FurMark test was actually valid, it was just incredibly demanding on the GPU with every type of setting fully maxed out. With AA off, it was around 40-45 fps.

 

Now, on to the findings. I tested again, this time using Afterburner as monitoring software. I tested both Valley and Firestrike. Valley first had a little hiccup, freezing at around 16 fps again (similar to the condition I had when testing with my own system). But then it came to life... I was amazed by the fps, as it was around 80-85 average. Even seeing numbers like 120fps with the less demanding scenes. The benches were done on Ultra settings, 1080p, but no AA. Firestrike also did not have any problems running the benches.

 

Which brings me to my conclusion. Apparently, the GPU as a separate variable is not the cause of this problem. The possibilities are:

• My PCIe 16x slot is broken / somehow stopped working properly during full operation and is now dead;
• Some part of my MB is broken (connected to the previous option, but maybe somewhere else);
• If possible at all, the PCI lane controllers are bad. I don't know if I'm saying this right, but perhaps something wrong in the PCI lane controller?;
• A combination of AMD CPU and AMD GPU is causing a problem (as the testbench had a 3770, so I cannot rule CPU manufacturer out as a variable).

My bet is on the broken PCIe 16x, as Windows does not detect the card being plugged in. I cannot confirm this, as I had too many variables between test systems. Plus, I only have one PCIe 16x lane on my MB. The rest is all PCI. Which brings me to my last question:

 

Does somebody know a way to verify the PCIe 16x lane is dead, besides undetected hardware according to Device Manager? So I can truly rule out software as a malfunctioning component.

 

The outcome of the benches kinda confirmed my suspicion of CPU bottleneck as well, as most people already pointed out. With the Firestrike tests, I got around 12-18 fps more on average with the 3770 on the Graphics tests compared to my Phenom II X4 965 (stock GPU clocks). Valley also gave spectacular fps numbers. The bench concluded in 78.1 fps @ stock clocks, while my X4 965 only got 51.2 fps @ stock clocks. But keep in mind that this 3770 test was without AA on, the X4 965 had 8xAA.

Overclocking a little bit gave me 85.6 fps @ 1075 core and 1400 mem (+20% PWL).

 

I know now which CPU is next in line when the time comes that I have to do another build. Unless the Zen rumors are true, as I prioritise core-count and MT/HT and for that price, Intel seems an illogical choice. Otherwise I am fairly certain to switch over to Intel, with something like a 6700K (would love a 6800K, 4 more logical cores. Dat price though...).

[jellybutt]

Alright. If no one else has anything to add, I'll mark this as sorta resolved for future needs. I bought a new Motherboard, no GPU problems with that one.

Thanks for the suggestions.