Absolutely pulling my hair out over these lockups

[Dstoop]

System Specs:
OS Windows 7 Ultimate x64
CPU Core i5-3570K (stock speeds)
Motherboard Gigabyte Z77X-UD3H
Memory 8Gb (2x4 Gb) Corsair Vengeance 1600mhz
Graphics Card MSI Radeon HD 7870 HAWK (Ghz Edition)
Sound Card N/A
Monitor(s) Displays ASUS VW266H 25.5"
Screen Resolution 1920x1200


Keyboard Microsoft Sidewinder x4
Mouse Logitech G600
PSU OCZ Modxtream Pro 700w
Case Silverstone KL04
Cooling Corsair H80
Hard Drives Crucial m4 256Gb Western Digital Caviar Blue 7200rpm 500Gb
Internet Speed 50D/20U Cable



Brief history: I put together a brand new rig almost a month ago, which has had some odd bluescreens and lockup behavior. Sleep and Hibernate are totally disabled, but the monitor is set to turn off after 10 minutes of idle. Occasionally, the computer will go into this idle state and will completely refuse to exit it. The video card stops sending signal to the monitor, it refuses to "wake up". The LEDs on my keyboard/mouse will work right up until I unplug them and plug them back in, then nothing, and eventually after trying to get it to wake up long enough it bluescreens with a DRIVER POWER STATE FAILURE bluescreen. But the odd thing is that you can still access shared files on the PC over the network, and open/transfer them all fine, so *something* is still working right up until the total failure.

When this first started happening, the video card would also sometimes have odd fan spinning behavior while it was idle, making a half spin at 100% then stopping constantly until you kicked it out of idle when it would immediately work normally again, and the first time it froze it kicked the fans up to 100% on the video card and kept them there as soon as I tried to move the mouse to wake it up. Eventually the video card failed completely to the point where any system I put the card in would not POST.

I RMAed the card and saw on newegg that it had an almost 25% 1-egg review rating for DOA/failing cards. When I finally got my new card on Monday the box was a little dinged up but everything looked OK and I didnt have any issues since, until just today when it froze again while idle. I couldn't get the bluescreen to trigger and eventually just had to hold the power button until it hard booted.

Here's the WhoCrashed analysis of the first three freezes, not that it's much help. It didnt record a BSOD event today for some reason.

I've got nothing on this one outside of tearing the whole system apart and RMAing everything, it'd be great if you guys could help me out. I've done repeated stability testing on the GPU, CPU, and RAM and *everything comes back fine*. No errors with MemTest after 2000% coverage, Furmark and OCCT can run for an hour and a half with no errors or artifacts, and LinX passed a full 20 passes with flying colors. The system works great right up until the second it doesnt 🙁

On Mon 9/17/2012 10:18:03 PM GMT your computer crashed
crash dump file: C:\Windows\Minidump\091712-4773-01.dmp
This was probably caused by the following module: ntoskrnl.exe (nt+0x7F1C0) 
Bugcheck code: 0x9F (0x3, 0xFFFFFA800AB01AB0, 0xFFFFF800045183D8, 0xFFFFFA800B3D42C0)
Error: DRIVER_POWER_STATE_FAILURE
file path: C:\Windows\system32\ntoskrnl.exe
product: Microsoft® Windows® Operating System
company: Microsoft Corporation
description: NT Kernel & System
Bug check description: This bug check indicates that the driver is in an inconsistent or invalid power state.
This appears to be a typical software driver bug and is not likely to be caused by a hardware problem. 
The crash took place in the Windows kernel. Possibly this problem is caused by another driver which cannot be identified at this time. 


On Mon 9/17/2012 10:18:03 PM GMT your computer crashed
crash dump file: C:\Windows\memory.dmp
This was probably caused by the following module: ntkrnlmp.exe (nt!KeBugCheckEx+0x0) 
Bugcheck code: 0x9F (0x3, 0xFFFFFA800AB01AB0, 0xFFFFF800045183D8, 0xFFFFFA800B3D42C0)
Error: DRIVER_POWER_STATE_FAILURE
Bug check description: This bug check indicates that the driver is in an inconsistent or invalid power state.
This appears to be a typical software driver bug and is not likely to be caused by a hardware problem. 
The crash took place in the Windows kernel. Possibly this problem is caused by another driver which cannot be identified at this time. 


On Fri 9/14/2012 10:26:26 PM GMT your computer crashed
crash dump file: C:\Windows\Minidump\091412-5600-01.dmp
This was probably caused by the following module: ntoskrnl.exe (nt+0x7F1C0) 
Bugcheck code: 0x9F (0x3, 0xFFFFFA800AF87AB0, 0xFFFFF800045143D8, 0xFFFFFA800CBCB010)
Error: DRIVER_POWER_STATE_FAILURE
file path: C:\Windows\system32\ntoskrnl.exe
product: Microsoft® Windows® Operating System
company: Microsoft Corporation
description: NT Kernel & System
Bug check description: This bug check indicates that the driver is in an inconsistent or invalid power state.
This appears to be a typical software driver bug and is not likely to be caused by a hardware problem. 
The crash took place in the Windows kernel. Possibly this problem is caused by another driver which cannot be identified at this time. 



--------------------------------------------------------------------------------
Conclusion
--------------------------------------------------------------------------------

3 crash dumps have been found and analyzed. No offending third party drivers have been found. Consider configuring your system to produce a full memory dump for better analysis.

 

[Puppies04]

Can you go into control panel and check the advanced settings in your power options to make sure that your SSD is set to sleep never. I have come across a couple of issues with SSD's, seems they go to sleep fine but can have trouble waking back up.

 

[postmortemIA]

can you check for bent pins on the LGA 1155 socket? I've been building PCs since 2000 and I haven't seen more fragile socket than this one.

 

[Dstoop]

Can you go into control panel and check the advanced settings in your power options to make sure that your SSD is set to sleep never. I have come across a couple of issues with SSD's, seems they go to sleep fine but can have trouble waking back up.

Turn off hard disk after: Never
Sleep after: Never
Allow Hybrid Sleep: Enabled
Hibernate After: Never

I turned off Allow Hybrid Sleep, though I would assume even enabled it would never activate because Sleep itself is disabled.

Though I did notice a setting PCI Express > Link State Power Management is set to Moderate Power Savings. Could this be the culprit? The description is "Attempt to use the L0S State when link is idle," maybe the PCI-E slot is going into its own wonky sleep state and doesn't like to come back up, forcing everything else to crap out too?

can you check for bent pins on the LGA 1155 socket? I've been building PCs since 2000 and I haven't seen more fragile socket than this one.

I'll keep it in mind, but taking the H80 off now that its on makes me cringe. The pins were all straight when I put the processor in, and it hasnt been taken out since. Wouldn't a bent pin cause the whole system not to POST, or at least have more noticeable effects?

 

[postmortemIA]

It depends which pin is bent. I am typing this now from mobo that I installed yesterday, and I had one bent pin, which I believe I had fixed, but not really sure. The pin that was bent was power supply for on CPU graphics, which I am not using. I've seen lots of people have this problem.

It is possible to do it right, no doubt about that.

I would also run windows memory diagnostics, hit F8 when BIOS finishes its loading.

Also I have Radeon 7850 and anytime I go to sleep mode in Windows XP it does not come back, but works fine in Windows 7.

 

[Dstoop]

Ill see about pulling the CPU off when i'm digging around in there today.

However, we're also back to really odd fan behavior while idle, and this time I got the chance to catch it in the act. There's two groups of indicator LEDs on the top of the card, two in each bank are always lit and the fans are always spinning. As soon as windows turns off the monitor, *these LEDS turn off and the fans stop spinning entirely*. Its as if the whole card has been turned off to save power, but the big glowing MSI logo dome still stays lit up like the sun. Every minute or so, these LEDs kick back on for between 2 and 4 seconds, the fans spin up to 30%, then it all turns right back off again. This is 100% reproducible at least since the crash yesterday, I don't know if I just didn't notice it before or if its new behavior. I grabbed a video of it with my phone that I can upload to youtube or something, and i'll take some pictures of the LEDs and which ones stay lit up and post them as soon as I have time to pull the whole thing apart today. I also have a GPU-Z log that clearly shows the behavior, when the LEDs are off it records junk data, and then records those 3-4 seconds of spin-up perfectly fine every time.

I've never heard or seen of a card completely turning itself off to the point where the fans dont even spin when the computer goes idle. My old 5870 sounds like a jet engine 24/7, it definitely doesnt shut the whole card down when the monitor turns off. These crashes might have something to do with the card not properly powering back up from this temporary off state, causing current devices to stay somewhat powered but not properly powering newly introduced devices, which might explain why the keyboard/mouse LEDs still work and you can still access the filesystem across the network.

 

[AnonymouseUser]

Update your BIOS. F16 = Improve VGA compatibility. F17 = Improve system compatibility. Both released within days of each other in August, so there were known issues. If that doesn't help try the following:

• Use the onboard video and see if the behavior continues.
  
• Try the other PCIe 3.0 slot.
  
• Try the PCIe 2.0 slot.
  
• Try another PSU if possible.

 

[Dstoop]

Ill try the BIOS, but with further testing I dont know if it'll make a difference.

I took my 5870 out of my old PC and put it in the new one, letting windows install just the basic drivers. It works fine, no odd fan behavior when the PC goes idle.

I put my 7870 in the old computer, letting windows install just the basic driver. *it does the same thing*. PC goes idle, fans turn off, lights turn off, every minute or so they pop back on for a few seconds and then go off again.

The odd thing? If I plug it into my TV or my monitor via HDMI, *it doesnt do it*. It only happens over DVI. I swapped out the cable (again) and it didn't make a difference. I guess because the HDMI signal has the power component to it, it just cuts the video signal and doesn't shut down the card itself. But why would the card shut itself off like that to begin with?

I've got those pics of the lights, i'll post them later. Both sets of lights that are lit when the card is on are labeld +-, whereas the unlit lights surround them are all labeled with error codes, so i'm guessing they're just the "yeah, the card has power" debugging lights.

The fact that it does the exact same thing in another PC entirely should logically rule out bent processor pins, motherboard problems, PSU problems, and RAM problems, i'd hope. I'm thinking this might lead to RMA #2, which according to the newegg reviews isn't unheard of with this card :/

 

[C1]

Load the driver and control panel for the affected video card then go thru the video card's control panel and inspect for available power settings. There's a chance that there is a check box available to alter the video card's power usage. (Note: I find such options in the software panels of my old cards which are less sophisticated than yours.)

 

[Dstoop]

No power settings anywhere to be found. The card should definitely not be awkwardly cutting its power entirely when the PC is idle, and it's doing it in more than one PC with both the full AMD driver suite and the basic microsoft vga driver. The failure rate on this card is considerable if neweggs reviews can be believed with people reporting similar bluescreens and card failures as mine. I already boxed it back up and initiated a refund RMA. I'll give another card/manufacturer a try, and be staying far away from MSI in the future.

 

[Dstoop]

Back to the drawing board. Replaced the faulty video card with a brand new card of a different model from a different manufacturer. No weird power issues, but the occasional freezing when it goes idle is persistent. The odd thing is that the night before it did it right around 10:15pm, and then last night it also did it right around 10:15pm. I'll see if it does it again tonight. I dont recall the freezes ever happening when I was using the onboard video between video card RMAs.