Computer crashes consistently during some games. I've tried everything I can think of. Paying for a solution.

Erik765

Honorable
Jan 26, 2014
3
0
10,510
Hello all,

At this point, I have tried everything I can possibly think of to resolve an issue where my entire computer crashes (sometimes just freezes up, sometimes restarts after a BSOD that I can't actually see). It will consistently crash in Metro Last Light, in one particular scene in the game. Other games crash as well, and others are completely stable.

I will literally PayPal $100 to whomever has a working solution at this point. I'm so defeated on it, I would almost do anything.

As I've replaced parts on my build (almost to the point of it being a completely new build now), certain games have become more stable as well.

My current specs are:

ASUS Sabertooth Z77
i5-3570k (currently at stock, although my overclock is 4.8Ghz (on water))
2x EVGA GTX 670 FTW 4Gb in SLI pushing 3 Acer GD235HZ in 2D Surround (6030x1080 w Bezzle Correction).
GTX 210 for Accessory Display (4th monitor)
8Gb Gskill DDR3 @ stock 1600Mhz
2x Samsung 840 PRO 128SSD (in RAID 0)
Samsung 1Tb HDD
1350w Silverstone ZEUS PST (thing's a beast)
Windows 7 Ultimate x64


What I've tried so far:

Replace motherboard
Replace PSU
Re-do all wiring from PSU (fully modular)
Replace SLI bridge
Try different SLI ports on GPUs
Try no SLI bridge (driver won't load at all)
Remove CPU overclock (ran everything at stock speeds)
Run games in Windowed mode
Remove all Startup items for a clean boot
4 different driver versions (all WHQL- 332.21, 327.23, 320.49, 320.18) (using Driver Sweaper to Remove all traces of driver/software inbetween each new installation)
Try different PSU rails for GPUs
Re-seat GPUs
Test RAM with latest version of Memtest86 (passed)
Move RAM to different sockets
Replace RAM
Try games using different DirectX versions
Remove accessory GPU (GTX 210 used for accessory display)
Keep CPU at 4.8Ghz (remove Turbo boost)
Control Panel - Power Options - Change Plan Settings - Change Advanced Power Settings - PCI Express/Link State Power Management set to 'Off'
Nvidia Control Panel - Manage 3D settings/Global Settings tab- Power management mode - set to 'Prefer Maximum Performance'
Uninstall EVGA Precision X
Set TdrDelay to 8 in registry
Underclock GPU RAM by 100Mhz
Set Vsync to adaptive in Manage 3D settings/Global Settings tab
Reinstall OS

At this point, I have a fresh copy of Windows 7, fully updated, fresh graphics driver (latest version), fresh installation of Steam and a fresh installation of Metro Last Light. The only other thing I have installed is Whocrashed (to read BSOD logs). EVERYTHING at stock speeds right now.

Here is a link to a video showing what the game looks (and sounds) like when it crashes:
(keep in mind, the humming sound is not my camera, that's the buzzing sound the game makes when it freezes my whole box up):

http://eriksplace.com/shared/VIDEO0104.mp4

Here is a pic of my temps at the exact moment of the crash(no, I'm not overheating), keep in mind, these were my temps with the CPU running at 4.8Ghz as well: (fortunately I have a few seconds before it restarts to grab a pic of this from my accessory display)

IMAG1451.jpg


If anyone cares, here is the config file showing the in-game settings on the fresh install (default settings, not even my actual resolution) for Metro Last Light: http://http://eriksplace.com/shared/user.cfg

Here is the Whocrashed output after it crashed from those settings (again, everything at stock speeds on a fresh install of everything):
http://eriksplace.com/shared/WhoCrashedOutput.htm

And, last but not least, by DxDiag report (64bit version):
http://eriksplace.com/shared/DxDiag.txt


What works (keeps it from crashing):
Disable SLI (which disables Nvidia Surround)
Run Nvidia Surround off one card (only possible if you physically remove 1 GPU from the system, effectively removing the benefit of SLI)


Some more notes I've taken along the way:
One card has 80.04.5C.00.70 (P2004-0005) BIOS version
One card has 80.04.4B.00.70 (P2004-0005) BIOS version
Both cards run at the same memory speed (1502.3Mhz, to get 3004.6 combined)
GPU clocks are different. One card runs at 1045 toggled into a game and 1084 when toggled out. Another card runs at 1202 toggled into a game and 1215 when toggled out. (speeds seem backwards here, shouldn't speeds clock up when toggled into a game, instead of down?) like clicking in and out of a game running in Windowed mode).

TL;DR? Solve my issue, I pay you money (as long as it's not against the rules here).

As you can imagine, I'm desperate and completely bummed out after having spent this much time/money on this.

Thanks in advance.

:(
 

Erik765

Honorable
Jan 26, 2014
3
0
10,510
Thanks for the replies so far. I could certainly try matching my GPU frequencies. Remember though- things I've tried #1 (replaced motherboard), so I could do a BIOS update and I'll keep you posted (the CMOS has been cleared several times), but it acts identical so far on two different motherboards (everything else being equal).

Again, it doesn't crash with one GPU installed ('What works' section...). Either card will run fine by itself. The problem is, that defeats the whole purpose and I spent $600 (for another GPU and water block) for nothing (SLI), is what it feels like at this point, and that much more horsepower is almost a requirement to run anything with decent frame rates at this resolution.

The thought has crossed my mind about the power into the house. It is an older house, but I'm pretty sure the wiring has been re-done in the majority of it. So far I've tried two different outlets, but they're likely on the same circuit (not very far from each other). I may try running an extension cord out to my garage just for the hell of it.

Would a different PSU make a difference here though (by the time it even gets to my components)? I've ran two different PSUs (my prior one was an Antec 850).

Another thought I had is that this PSU has pots on it where you can adjust the voltage amounts for the rails. From what I've seen, it always looks like the 12+V rails are slightly above 12V, so I'd be worried to really mess with these much, but what do you think? Maybe just a little, and see?

Thanks for the input so far. I'll see what a BIOS flash does.

More thoughts?