f13.net

I finally got fed up with AoC crashes, so I canceled and picked up BioShock. Great game, but it also crashes :( ! Well I am starting to suspect my PC. When I put it together I briefly tested everything, but this time around I need to find out what is wrong. I am not overclocking anything. Every driver and all firmware are up to date.


So far I tested

a) overnight memory test - passed
b) 30 minutes CPU test - I will set it to test tonight overnight

I need to test GPU and raid (HDDs), any suggestions how to stress-test these? Thanks!

GPU I let it run the FF XI benchmark or 3DMarks set on a loop overnight. If it's still running when I wake up I figure things will be pretty stable (and I have had cards that have failed this test which I returned).

Final Fantasy benchmark? O.o

http://www.playonline.com/ff11us/download/media/benchmark01.html

Back in the day this thing was brutal -- it was worse than running EQ II on your machine :awesome_for_real:

Dunno how it runs on something like an 8800 or 9800 GT these days.

Edit: oh hey they are up to version 3 now. Probably still a good stress tester then.

Do people get these strange diagonal lines across the FFXI demo? I get great FPS with my 8800 on the demo, but there's this near constant effect that I vaguely remember from the actual game when I played it, what, 5 years ago? FFXI seems to use a very different rendering engine than most western MMOs...is it an effect of it being a console port?

Nope never seen diagonal lines before.

With FF XI benchmark I get Err Code =3 within hour of running it :( no idea if I just forgot to disable some sleep/power safe mode setting or video card is actually crapping out. I used to run 3DMark in repeat mode but cheap bustards disabled it in *all* free versions.

Did you try both resolutions?

Well crap that's annoying.

Do you have something like UT 3? I believe that can be setup to run demo benchmarks in a loop.

Ya I have UT3... I have to look into how to set it up into loop benchmark.

Prime was stable overnight, 2 cores 100% load. Processor got to 70C but no errors. So now I need to determine if this is OS or GPU instability.

I tried number of UT benchmarks and none of them let me loop it!!!

From what I observed crashes almost always happen well within playtime, never right away. It also doesn't ever blue screen, it just crashes to desktop. FF demo gives err =3 some-time into test run, couldn't find anything on error codes.

What are the GPU temps like right before/after it crashes?

75-77C at max load, I have good air flow and additional slot fan cooling GPU. FF benchmark only gets it to 60C.

I run semi-attended benchmark test with FF. I think error 3 is related to going to power-safe mode (even if I disabled everything I could find).

Wow. That's hot.

not for a gpu. that's on the high side of good.

I did "light" WinXP reinstall, where I just installed it on top of old one without wiping any drivers or settings first.

BioShock and Age of Conan still regularly crash to desktop. I think I will start looking into error messages, maybe it will tell me why.

Logs are:

The application, C:\Program Files (x86)\2K Games\BioShock\Builds\Release\Bioshock.exe, generated an application error The error occurred on 08/03/2008 @ 18:16:05.312 The exception generated was c0000005 at address 4E0B759B (d3d9!Direct3DShaderValidatorCreate9)

Faulting application bioshock.exe, version 1.0.0.0, faulting module bioshock.exe, version 1.0.0.0, fault address 0x001bed69.

So now you know its something to do with shader handling, which suggests a gpu failure. Did you use a fresh driver set, or one previously downloaded?

I am re-downloading Nvidia drivers right now. I am also going to reinstall directx.

Reinstalled *all* Nvidia drivers and DirectX. I doubt it will help.


At this point my guess is that BFG overclocked my video card too much - shader error I am getting is probably related to shader clock speed and only shows up at full load.

You can download NTune from Nvidia that will let you underclock your Nvidia video card. May be worth a shot, but to be honest, I think the damage may already be done.

You need a clean install.

Download all your drivers, Firefox, Adaware, spybot search and destroy, put them all on a CD...

FORMAT

Reinstall windows
install everything off the CD

Install 1 game/benchmark that was failing prior to the format


try again

Also try ATITool (http://www.techpowerup.com/atitool/), ignore the name as it works for all video card vendors. Will let you know your GPU temp and is built for stressing the card with artifact testing for reliability.

Check your vendor's website for downloadable diagnostic utilities to check the hard drives (Also your raid controllers diagnostics). If you want to bench the disks try something like IOmeter (http://www.iometer.org/) which has binaries for Windows/Linux.

I'd also like to add, I'd reset your bios to defaults before you reinstall as well.

You could have some mem timings off or something.

Ya know, I just realised that its a 64 bit OS...its probably worth reinstalling everything from scratch with the absolute latest drivers you can find for all components involved. 64 bit technology is still pretty new in terms of consumer products, and lord knows what compatibility issues are festering beneath the surface, just waiting for that one memory call to that one bit of address space.

Thanks!

60 min of artifact scanning - no errors. Temp is stable at 79c. WTF could it be?

How much ram are you using? Its possible that only on games that use every inch of your memory are you getting errors, since the upper end may not be used by casual use. Burn a copy of Ultimate Boot CD and run a memory test, or find another bootable disk that has Memtest86 and run that.

I bring this up because just today I was troubleshooting a machine that had two gig sticks of ram, and one was bad, as in totally borked. However, it would boot to the OS and run just fine till you ran something that took more ram to run. Then the program crashed, and occasionally, BSD.

c0000005 is a memory access violation. It's either poorly written software (that accesses memory where it shouldn't) or a hardware RAM defect as mentioned above. If you have multiple sticks of ram, try booting up with only one and see if it solves the problem (alternate as necessary). The memtest Engels suggested is very good too but can take quite some time.

My experience with Memtest86 is that if its a seriously broken piece of ram, it finds it in the first few seconds. I think that regardless of what phase the test it is in, it populates each stick of ram right out the gate.

Try http://forums.2kgames.com/forums/showthread.php?t=6356

Memtest was first thing I run - overnight and it was all fine.

well fuck

With the hardware checking out alright, it looks like you are down to software.
If you still want to spend time and energy on this I'd recommend a fresh windows install in a different folder (without wiping your harddisk so you can still revert to your current os later). Upgrading it with the latest drivers and patches and seeing what that does for your gaming stability.
It's quite a lot of work but it will get you a squeaky fresh windows installation. It's something I do every six months just to stop windows from cluttering up and getting gradually slower. Pretty much all software will run without the need to reinstall but you will lose most your settings.

Alternately, you could use a single cheapo empty hard drive to install a baseline OS rather than gumming up your current config.

Have you uninstalled the graphics card drivers and run a driver cleaner app, then rebooted and reinstalled? I ask because Nvidia, for example, is notorious for leaving old driver files operational despite the driver update, and if one of those is corrupt, you'd not see an improvement.

This man speaks the truth!

DriverSweeper (http://downloads.guru3d.com/download.php?det=1655) is one of the best little utility's for this.

 A friend was playing around with his nvidia card's multi monitor handling with his LCD tv and eventually put his computer in a state where past the Windows XP bootup logo the video card wouldn't display on his desktop LCD. Tried to reset it 20 different ways with safe mode, removing the driver, registry settings, reinstalling a driver and it never worked. DriverSweeper (http://downloads.guru3d.com/download.php?det=1655) fixed it and put the computer card back into a VGA driver/mode and reinstall of the drivers worked flawlessly.

My sentiments exactly.

I will try DriverSweeper.

You really should reinstall WinXP from scratch.

If you are running more than one stick, swap the sticks and run it again.

Also, do you have another video card you could try?

Reinstalled OS with HD format, all new drivers... BioShock still craps out.

Faulting application bioshock.exe, version 1.0.0.0, faulting module bioshock.exe, version 1.0.0.0, fault address 0x001be4e9.
The application, C:\Program Files (x86)\2K Games\BioShock\Builds\Release\Bioshock.exe, generated an application error The error occurred on 08/10/2008 @ 19:04:14.281 The exception generated was c0000005 at address 10ABE4E9 (Bioshock)

Run with only one of the sticks of RAM. If you have another video card try that too.

Wouldn't memtest or video test expose problems? About the only thing I didn't stresstest is RAID, no idea how to do it reliably.

If memtest finds a problem. well then, you have a memory problem. If it doesn't find a problem you could still have one.

Basically you need to ignore the tests since they are giving you conflicting answers and just isolate components one by one until you can (hopefully) narrow down the culprit. Unfortunately it could also be the motherboard, which would be a bummer.

The more I think about it more MB seems to be likely cause - every component checked out when tested alone. All errors are random, things go wrong in a very different way every time. What would be good replacement for P5N-D, ideally something that supports DDR3 so I can upgrade that as well?

I will try swapping video cards next, then I will try to stress testing RAID. Any idea how I can max out HDD data transfer and verify data integrity at the same time?

SO's M3A AMD rig is surprisingly stable, 1001 firmware upgrade and not a single problem. Also only 5-10% behind in all benchmarks from my rig, and mine cost nearly twice. This is the last time I am picking up Intel, should have stayed with AMD, regardless of it being slower it is more reliable technology all around.

If it is your motherboard I suspect the problem is more the NVIDIA chipset than the Intel CPU.

Did you still want SLI or you going to give up on that for a while?

Heh, ya, Nvidia chipsets. I've avoided them after a bad spell with a NForce 4 back in 2002

What video card are we talking about here ?

GTX280 BFG OC, I believe.

Ok.  I just found out some interesting heat facts about my HD4870 at the weekend and wondered if it was the same one.

Namely, the BIOS Fan control is pretty much broken and won't go above 23%, leading to a GPU of 80.  The solution to this is to actually make your own profile and manually edit the XML file to get the fan to go faster.  As a result, I'm at around 45-50 degrees.

I found this whole situation a little bizarre, especially seeing as it took me a while to find out why the fuck it was happening.

GPUs have a much much higher heat rating than CPUs. 80s isn't that out of control for GPU. My Nvidia card, the 8800gt, for instance, won't kick in the fan above 30 until it reaches above 100. Naturally, I manually tune it.

Yeah, but that wasn't really the point.  Since the fan wasn't going properly, the thing was basically turning into a wee grill and heating the case up like an oven.

Ya, its pretty annoying. I have to wonder, however, if that's sometimes done by design. If you think about it, the GPU chip is going to put out high heat regardless of what type of heatsink and/or fan you have going on; its just a matter of how well your heatsink and fan dissapate the generated heat; If the GPU is rated towards extreme temps, wouldn't it make more sense to NOT dissapate that heat inside the case? In other words, let the heat 'hang out' by the gpu as much as possible without letting it spread upwards towards the CPU.

This is, of course, assuming that the video card has no back exhaust, and just blows the heat into the case enclosure. Speaking of which, does anyone have a good recommendation for a video card that does exactly that; doesn't blow heat into the case, but has it funneled straight out the back? I've got a 8800gt, but its just wacky hot all the time.

I swapped back from AM2 -> p35 was for chipset stability. After nvidia's wonkyness with the ethernet driver (lag generated from the packet tcp offload), and reading HardOCP's comments I swore off of their chips for this generation.

Nvidia is in a weird position anyhoo, AMD/ATI's chipsets are working well, Intel is switching its architecture soon, and dependent on board manufactures to use the nforce 200 on the x48/x58/nehalem(has it been confirmed yet?) to get SLI.
I really hope the technology dies as crossfire has really matured over the past few years without requirements for dedicated hardware/chips besides the x16 slots.


8800gt has the same issues with the initial release, the fan speed is messed up, atleast you could edit an xml.. nvidia's fix is flashing it (http://www.hardforum.com/showthread.php?t=1260590) :S

I just use NTune and manually jack up the card fan speed to 50 before playing anything serious, like AoC, Mass Effect, etc. It still kicks in at 100 to 100%. Its true to, Nvidia really screwed the pooch with the 8800 series cooling methods and bios profiles. Its a total shame, because otherwise the cards are champs.

The 4870s vent outside the case, and yeah, setting the profile fixes it. I keep mine set to 60% now. And the latest drivers (July 22 or so) default to the profile, so I don't have to set it manually every time.

Er, mine.

The HD4870 slings it out the back.  My case temp went down at least 5 degrees once I cranked the fan up.

Yep. I think I will go SLI,  price dropped to the point that you might as well.  Still at this point I just want it to fucking work. I went with P5N-D because so many people had it working.... Not me. So 2x SLI PCI-E2 DDR3 mobo, any out there? All online vendors seem to be into extremes - it is ether 3X SLI overkill mobo or last century junker, nothing in-between. Oh and it has to have optical out.

Before I give up on my current rig and start building another - is there a good way to test RAID? I think this will be the last thing I will test before junking this setup.

On more thing - sometimes WinXP won't load past loading profile. Generally it is instant after I enter password but first boot is usually sluggish or outright hangs. Once it boots once all consecutive re-boots are fine. Almost like something need to "warm up" for it even to work right.

Strangely - I noticed with age I am less likely to tolerate crashes. What used to be buisness as usual in UO/SB gets me walking away in disgust these days. I quit AoC because of crashes.

Could this be that Bioshock and AoC are just two garbage games that are... well just crash? Strangely enough I yet to have a crash with UT3, but then I don't play it for more than hour at a time.

I would recommend getting XP up and running without raid. in fact, if your motherboard has an IDE controller that doesn't utilize the raid chipset, all the beter.

Well unfortunately they just don't make what you want. So you have to decide what you are willing to compromise on.