XFX GF8800GTX XXX crashes

[speckbert]

Hi all,

I have a strange problem. My computer keeps crashing when i run 3DMark06, Overlord and Infernal. I never expirienced crashes running Everquest, watching films, writing texts etc.
The crash is not really reproduceable. It takes from a few seconds to a few hours until the app crashes, although it usually happens within a few minutes. My record so far is running 3DMark for 3 hours without crashing. It always looks the same, just with different colors sometimes:

http://i24.tinypic.com/2f0g0w0.png

My system specs:

XFX GF8800GTX XXX
Asus P5B-E (Bios 1601)
Intel C2D E6600
Mushkin DualChannel Kit 2x1024MB DDR2-800
Samsung SP2504C
LG GSA-H42L
Enermax EG701AX-VE (I also tried the Enermax Liberty 620W, but nothing changed)
3x Case Fans

Heat is not a problem, already checked that. I also gave the card back to my vendor, but got it back 4 weeks later with the remark, that they ran 3Dmark06 for 4 hours without a crash so everything must be fine.

Can anyone help me? Any known incompatibiltys between my hardware (e.g. the p965 chipset or so)?

Thanks in advance for your help,
speckbert

[Scott]

I would use Riva Tuner and lower the memory clocks by 5MHz and the GPU by 5MHz. If the problem goes away the card is overheating.

[speckbert]

The heat problem was my first idea, too. But it's not the reason.
I checked temperature using NVMonitor while running Overlord in windowed mode and it crashed at less than 70°C.
I also lowered the clocks down to standard non-OC (575/900) and 3DMark06 sill crashed. It also does not show the typical overheating problems like texture errors, artifacts etc before crashing.

I forgot to mention that I use WindowsXP and I have the latest bios on my mainboard as well as the latest drivers.

[Scott]

Hopefully Shannon my XFX tech guy will be chiming in soon. I do recommend you open a ticket # on the XFX site and then post the # here so Shannon can get you going sooner.

[xfxsupport]

Thanks Scott, yes start a ticket at www.xfxforce.com and let me know the ticket number. I take it you have tried different driver versions for the video card? Also if possible a great way to test would be to try this video card in another system as well. That picture is not your normal "my system crashed pic". Do you get any error codes when your system crashes?

[Miker]

At least that screen is not as bland as the blue screen of death. We shall call it "The Ranbow of Death". :wink:

[speckbert]

Yes, I tried different driver versions (almost every driver that was released the past months). The picture I posted is what I always get on the crash. There is no bluescreen, no crash to desktop, no reboot. Suddenly, out of nothing, the screen turns into this freeze image, a few (very few) pixels are changing color, sound keeps running for a very short time and then also stops (guessing it's playing whats still in buffer).

I will create a ticket later and let you know the number. Thanks for the fast answers :)

speckbert

[speckbert]

I just started a ticket, the number is 113888.

[xfxsupport]

Did you have a chance to test any of the components in another system? Thanks for the ticket number I will let the Europe team know about it.

[speckbert]

I will take my graphics card to a friend next week. He has a similar system I have, just a 8800GTS and a different PSU.

I did some more test, here are the results:

When I start Windows, do some stuff with clocks set at 630/1000 and then clock down to 575/900 using nTune, I still get the crash, although it feels like it takes longer.

Using RivaTuner to automatically override the clock speed at startup to 575/900 (+ setting some compatibilty checkboxes) seems to eliminate the problem. However, some other things happened with these settings. I ran 3Dmark06 in loop the last 2 nights, to find out that both times my PC rebootet at some time when I checked in the morning. But I have to say that I do not know what happens if you leave the PC with the "Rainbow Crash" for a longer time, since i always rebooted manually after some time when i happened.

Next thing I will try is having nTune log the temperatures and run the system at 630/1000 to see if I can find out anything.

I also did some research since i wondered why there are games i can play for 12+ hours (yes, call me crazy, but I do that sometimes) without any problems, and some crash after a short time:

Games that run without problems at 630/1000:

- Vanguard (checked temperature, its at 90+°C)
- Everquest
- Oblivion

Games that crash:

- 3DMark06
- Infernal
- Overlord

I found out that the games that run perfectly neither use anti aliasing (played Oblivion with HDR enabled, what automatically disables AA) or shader model 3.0. Only shader model 2.0 is used (Oblivion has 3.0 shaders disabled unless you edit some ini files, what I did not due to not knowing that until now.)
The games that crash all use anti aliasing and 3.0 shaders.
I'm not sure about texture filtering right now, also might have an effect.

[kiwiplayer]

I had a similar problem with my stock 8800GTX, i tried many different drivers and even bought a new PSU as well as adding extra fans, underclocking the 8800GTX, fresah install of XP and disconnection of all unnessasary hardware (case lighting, most HDD's, some case fans, etc).
most of the time i would get a throttling effect or a BSOD, but one of the RMA cards had the same " rainbow of death" like yours every time.

after a few RMA's, my 8800GTX is working.
the previous cards were both the black or green PCB versions. they were also PCB versions from 1.7 to 1.8.
the current card is the black PCB version 2.0 which seems to run a few degrees cooler and i score a few hundred points better in 3dmark06 than the previous cards.
I was just unlucky enough to recieve a few bad 8800GTX, the store i bought the card from went out of their way to search for a newer PCB version and also to test the card before sending to me and that is how i ended up with a great 8800GTX that works well.

do not misunderstand me, i am not saying their is a widespread problem or that there is a problem with PCB version 1.7-1.8 but i am saying that 2.0 appears to work very well indeed. so what version PCB is your 8800GTX?

good luck.

[speckbert]

My PCB is green, but I don't know how to find out the PCB version.
There is a sticker on it with "Ordering code: PV-T80F-SHD9 V1.0 W12/06", the serial number and some other stuff. The only other thing that looks like a version thingy is the sticker with the bios version (its 60.80.08.00.40).

Thanks for your post, gives me hope :)

[kiwiplayer]

Ordering code: PV-T80F-SHD9 V1.0 W12/06 the V1.0 is the version and the W12/06 is when it was packaged and sent of to the supplier (i think:grin:)
so yours is the first released PCB. this is not a problem but i assume you've had the card for quite some time, atleast 6 months.
how long ago was it having problems?

I believe the error your getting is a Vram issue. see if you can find a graghical test which runs at a high resolution. unfortunatly 3Dmark06 is probably the best.

In my opinion, running 3dmark06 for 4 hours does not sufficiently stress test a VGA card.
I was able to run 3Dmark06 for 5 hours fine, but once i started a game, i locked up. my thoery is that there is so much Vram on these cards, loop testing an application probably means it uses the same amount of Vram in the same allocation.
Try this,run ATItool scan for artifacts for 2-3 hours, then run aquamark, then 3Dmark03, then 3Dmark05, then 3Dmark06, THEN start a game, play for 10 minutes then try a different game.
I found supreme comander, UT2004 and world of warcraft to be the ones that caused the error more often.
once the error appears, reboot and start again, ATItool for 20minutes this time. does it keep failing with the same benchmark, does it fail more with a certian game?

I'd suggest to stick with XFX RMA rather than the store you bought the card from.

[speckbert]

I believe you are right kiwiplayer.

When I played a game and then run 3DMark, it crashes within a few seconds usually, while i can run it for a few hours after a fresh reboot. I noticed that already but didn't realize it until you posted it.

I bought the card in April this year. Overlord was the first game that crashed. I usually play Everquest what never crashes (thank god :)). Even running 2 clients at the same time (which uses ~98% of the video ram according to NVMonitor) does not cause any problems. For the first few months I only played Oblivion besides Everquest. Oblivion also has a rather old engine I believe and never crashed. As i wrote before, games without anti aliasing and only shader model 2.0 run perfectly without any problems. So I can't tell if the problem was there from start, since I did not play any "new" games at first.

I think I have the same problem you had, since you describe exactly what happens with my PC. I hope XFX will help me :).

[kiwiplayer]

I had a bad run of 8800GTX:grin: so i've done alot of testing with alot of video cards.
the store i bought the original card from also tested all the returned cards and reported only one was definitly faulty but as they told me, they didnt have the time to run 3Dmark06 and then playgames so they were trusting i was doing and testing things correctly. i love that store :icon_tiphat:

I think its now the weekend for XFX so you may not recieve a reply for a couple of days (as far as RMA goes).
as a temporary fix, try this. (assuming your using XP)

go to "control panel", open "display" then click the "settings" tab.
make sure your not using dual screen, if you are, turn it off.
click the "advanced" button, then click the troubleshoot tab.
UNtick the "enable write combining" box and drop the hardware acceleration slider one notch. click okay then when prompted, do a restart. if the computer still has problems, move the slider back alittle more. do this untill you have no more problems.

this is by no means a permanent fix, this is just a work around. with the slider all the way to "none" you have now disabled most (if not all) driver settings but it will hopefully allow you to play games with low detail until the issue is fixed. the microsoft way is to start from "none" and move up, i found it easier to go from "full" and move down.

have you tried the latest driver? have you completly removed the old drivers?
to do this, restart the computer, just after the bios post and when windows starts, hit F8, choose VGA mode (or safe mode if you cant choose VGA mode). when the computer has started, remove the Nvidia display driver then reboot. again go into VGA mode and run driver cleaner, (http://downloads.guru3d.com/download.php?det=745) i always went to tools, cab cleaner, ran that then cleaned the drivers. all is with the one program). once you have run the progam, empty the rubbish then restart normally and install your display driver.
i suggest doing this everytime you update or change a display driver or video card.
doing this should also undo what i had suggested above.

enjoy and i hope XFX are able to fix the problem for you.

[speckbert]

I have a XXX version (overclocked to 630/1000), but I have clocked it down to standard clocks (575/900) for a few days now. It seems to run a lot more stable, but I have yet to test what you said at that speeds (running one game/benchmark after another). Will do that as soon as I have time and post the results.

[speckbert]

Ok, after some testing, here are my results:

1. Setting RivaTuner to put clocks to 575/900 at boot, I could not produce that crash again.

2. AtiTool seems to be the best Tool to use in my case, since it almost instantly crashes when I use the default clocks (630/1000).
Here is video I made, its not the best quality, but all I do is log in, start AtiTool and click "Scan for Artifacts". So finally this crash IS reproduceable.
http://www.stud.uni-karlsruhe.de/~uwiw/crash.mpg

3. When I boot with default clocks (and NOT having RivaTuner modify them), and then clock down in AtiTool before scanning for artifacts, it takes a bit longer to crash, but still does.

4. Doing like described in 1., scanning for artifacts does not crash (or it takes too long for me to remain patient). However, turning the memory clocks up to 1000 while still running the test instantly produces the crash.

I am 100% sure now, that it is a VRAM issue. The VRAM somehow can't get along with the increased memory speed.

Thanks a lot kiwiplayer, you really helped me here :)

I learned a few other things from using AtiTool (oh damn I hate that name, else I might have tried it earlier):
The card seems to have 3 modes: 2D mode, low power 3D mode and performance 3D mode.
Playing the super-stable-never-crashing Everquest, the card is only in low power 3D mode, while all games/tools that crash use the performance mode. All three modes however have the same clocks.
I have no idea, what makes the card pick the mode, maybe its the shader model used or the anti aliasing setting or something like that.

I also got a reply to my ticket. I should try to update my DirectX, what I did of course. Still no change (as expected). Will keep you updated.

[kiwiplayer]

great news, good luck with the RMA :)

ATItool is the best program for testing, i use to use it alot for my older ATI 9800XT :wink:

[speckbert]

Ok, some news. I got a reply to my ticket saying I should contact the reseller since they hold the warranty. I'll do that next week.

Now some more testing results (including the old ones):

At 630/1000 I can run 2D-apps and older 3D-games. 3DMark06 crashes sometimes, sometimes runs for hours. Newer games (Infernal, Overlord) crash after some time. AtiTool crashes instantly on "Scan for Artifacts".

At 575/900 Infernal, Overlord and AtiTool run prefectly. I thought these clocks were stable, until I downloaded the Crysis Demo. The Crysis Demo really seems to torture the card as it produces the crash even at these clocks.

At 500/750 Crysis Demo runs fine and hasn't crashed yet.

-------------------------------

I tested some more and set Bios to only use x1 PCIe. With standard clocks (630/1000) and PCIe set to 1x something interesting happens. The framerate decreases significantly of course (only ~30% of the x16 fps in AtiTool) but it does not crash. However, for the first time, I got Artifacts using the "Scan for Artifacts". The card will produce Artifacts at least every 30 seconds, but no crashes.

-------------------------------

I also took the card to a friend and tested it there. He has the same CPU and Board, but different RAM. While AtiTool crashes instantly on my PC, it took ~30sec - 1min until the first crash occured.
After rebooting we couldn't produce the crash again, but we didn't have much time to test.
This leads to the conclusion that there are other factors apart from the card.

-------------------------------

Leaving the PC after it crashed sometimes results in a reboot after some time. If this happens, my monitor stays black, and even after manually pressing the reboot-button it stays black. I have to switch the PC off and on, then everything works fine. After WinXP loads I get a window saying XP crashed and displays some error codes. The stop code is 100000EA, which is described here: http://support.microsoft.com/kb/293078/EN-US/

-------------------------------

Finally some questions:

Although it appears to be a problem directly related to the graphics card (the more demanding the app is, the more it crashes), could it be a CPU/Board/RAM problem? Bad Northbridge that can't handle the speed? Incompatible RAM?
I ran Prime95 for 10 hours without problems, memtest86+ for 4 hours also without problems.
I want to be sure I did everything I can before I turn the card in.

[kiwiplayer]

It's exactly the same problem i had with my card, right down to the error code and the microsoft fix (which i explained a few posts up).
the symptom are exactly the same.
my monitor would go blank then display a warning that the resolution was incorrect, this is with a CRT:grin:

I was unlucky enough to go through a couple of cards until i finally found one that worked properly.
I think this is the most common problem a faulty 8800GTX has.
of course i doubt there are too many faulty 8800GTX out there or this forum would be full.

no two components are the same, so your mates computer is different, maybe more stable. fact is, the card still failed after 30sec - 1min.
one thing i found was, the more the card failed, the harder it got to make it fail. however, it would still fail, usually after a few hours of gaming or web browsing.
All my cards were reported as faulty, all had the same problem (one card had a nother couple of problems too), so sleep well. your 8800GTX xxxis most probably faulty.
ask the store to test the replacement before they send it