The long running GSOD issue

[Phage]

Just thought I'd like to share that I have been chasing this problem down over the last few weeks and it does seem to be the RAM on the card that's causing the issue. (i'm summarising a stack of threads on various forums including the AMD Game)

There's a bloke who can reproduce the issue at will by overclocking his RAM.
http://www.youtube.com/watch?v=FnxNSi8YTZ0

After going though 10.2 10.3 beta, 10.3a and 10.3 WHQL versions of the drivers, and clock settings, things were marginally improved, but not cured. At my wits end they gave me a new BIOS for the card and an update tool. This seems to have worked !! (touch wood)


I have asked XFX and they said the following.
"if others with the same XFX card have the problem they should register, and they will get the update. This is only for the XXX version of the XFX 5870, no other version or model. kind regards, james "

I can recommend giving it a whirl, or if you have a faulty XXX model let me know and I'll send you a copy.

[Syllopsium]

Has anyone checked to see if the new BIOS is underclocking the RAM..?

[Phage]

CCC and GPUz tell me still at standard 875/1250 on this card. Currently 157/300 on idle - so I have no idea what they changed.

[Syllopsium]

Well, a fix is a fix.. Looks like ATI's cards can now be bought without a worry now the GSOD and HD decoding errors have been solved.

[Phage]

Looks that way. Thumbs up here (touches wood again)

[caveman-jim]

I speculate wildly they changed a power link state threshold or memory training point, specific to their versions of the VRM's used on the cards.

[andmar]

Phage, Are you still floating about on these forums? I've bought a xfx 5870 xxx myself and having the same striped vertical line crash, is your card "still" ok? I haven't tried underclocking yet as its taken me ages to finally conclude it must "just" be the graphics card at fault. I also notice that now on the support downloads they list a 1.7 bios update marked as "BIOS Update - For HD-587X-ZNDx models only" is that the tool they sent you? That tools says they adjusted the fan speeds.... I also tried flashing this on my card on advice from support and it hasn't solved my problem. I'm interested to know if the flash you have if the same as this one.

[Phage]

Phage, Are you still floating about on these forums? I've bought a xfx 5870 xxx myself and having the same striped vertical line crash, is your card "still" ok? I haven't tried underclocking yet as its taken me ages to finally conclude it must "just" be the graphics card at fault. I also notice that now on the support downloads they list a 1.7 bios update marked as "BIOS Update - For HD-587X-ZNDx models only" is that the tool they sent you? That tools says they adjusted the fan speeds.... I also tried flashing this on my card on advice from support and it hasn't solved my problem. I'm interested to know if the flash you have if the same as this one.

Yup - still here.
The card is A-OK now, but I can't tell you anything about it until I get home tonight. All the necessary tools and files are on the machine at home.
If you have flashed the BIOS already, and it didn't work, I'm happy to share my BIOS and tools. If that doesn't work you should ask for a RMA code. They gave me one when I hassled them.

[andmar]

If you could send me the tools that'd be really helpful (I can PM you an email address?), I reckon I'd binary diff the bios to the one on the xfx site and see if it's the same. Did you ever find any recognisable difference after you flashed it? i.e. did it report a different version of the card in some form or another somewhere inside windows? I've had a suspicion that when I tried flashing the tool hadn't applied the new bios as I couldn't verify any change.... I just assumed it had done "something" or completely "not done something".

Interested to know if you had the same problem though, I find pretty much any game does it (I usually play bad company 2, settlers 7, gta iv) although I can sometimes have sessions as long as 4-5 hours straight with rock solid stability and the all of a sudden you can't make the damn thing run a game for 10 minutes. I'm currently on 10.12 from the xfx site and (weirdly!) find turning my virus checker off during gaming seems to help slightly (yes very weird!), which made me think it was something to do with timings and/or chunks of memory not getting through in time as they are being scanned.... Anyway, basically what I tend to find is it often doesn't just happen straight away, it tends to "warn" me, I'll get my screen flicker black for a split second, or a very quick flash of artifacts, or the weirdest one, I'll actually have it display the coloured vertical stripes but then recover (!!!) for a brief time, but it nearly always goes on to fully hard lock with the stripes displayed (ctrl-alt-del is useless) sometimes the music/sounds continues, sometimes it starts to loop.

Another oddity, older games and windows tends to work flawless for as long as you like, this was the main thing that made me think it was "just" the gpu, as you seem to have to push the graphics card for it to happen, strongly pointing to some kind of over heat... which is interesting as you are also running an antec case (the super large and windy 1200 in my case!) and a gigabyte motherboard (with all it's des2 and gdrv.sys problems which I've tried to rule out)

Does any of that sound familiar to your experience?

[Phage]

That is exactly what I had. Although I never tried the AV ! All sorts of odd games, at no regular point in the game. Interestingly the problem was completely solved by the new BIOS, and ran like a dream, smooth and solid for several months. So good in fact that I bought another one.

Then, of course the machine developed a new way of crashing when I went to X-Fire. I installed another case fan with some homemade ducting.

http://forums.hexus.net/help-technic...ard-crash.html

I found that problem eventually. It was heat - but not due to the clocks on the 2nd card, but becuase the fans didn't accelerate with load. I don't know why this would be the case, but with a manual fan profile on MSI afterburner all is fine.

Since then I've dumped Afterburner as I couldn't get it to run at startup for non-Admin accounts. Instead I ennabled manual fans under CCC and find that 40-45% is enough for any game. I haven't had a crash since.

If I can find them, I will find the files tonight. PM me your email.

[andmar]

Fascinating, funnily enough, trying to set my fan speed to manual 100% for testing is on my list of things to try, along with...

Trying to flash the gpu... again, just to make sure
Turning off catalyst ai (although I "think" it's already not on!)
Down clocking the gpu/vram
Checking my psu rail set up, but it's an 850w psu and I'm sure I used two seperate hard wired PCIe connectors!
Looking into the "gauged memory fix" some people have been talking about

And probably some other stuff I forget.... I'd be surprised if this is a heat issue as I'm running an 9 fan set up in an antec 1200 unless the gpu fans are just a bit rubbish?...

2x 120mm fans push/pull intake through a h70
1x 120mm rear top vent
1x 200mm top vent
1x 120mm side intake
3x 120mm front intakes
1x stock gpu fan (obviously! )

[Phage]

It's definitely a VRAM timing issue. I did a ton of reasearch on this. In fact Google phage and GSOD or similar and I'll pop up all over the place.
Did you see the YouTube link ? Heat was only a problem in my X-Fire setup.
In the end this is last years tech - if the BIOS flash doesn't work - RMA it and get something else.

[andmar]

Yeah I saw the video, I was mish mashing the two conversations as I "had" suspected something along the lines of VRAM overheating.... slightly different; if it is the timings then that is a funny problem! I have to wonder if it's actually one of many problems being "exhibited" in the same way! i.e. the lines crash is just "how these cards crash" but what is "causing" them to crash could be one of many things.

[Phage]

Could be heat. But the new BIOS did fix the problem in mine without any new fans. It was only when I went to X-Fire did I hit a heat issue.

[andmar]

Some interesting info re: timings here...

http://www.nfohump.com/forum/viewtopic.php?t=57954

Seems to suggest that it can be to do with the gpu down clocking for some daft reason and providing too little power to the core for a short period of time. That would explain the "recover crash" whereby it crashes, recovers, then when it throttles down again it eventually fully crashes. The answer suggested here seems to be to "software fix" the clocks and voltages... which I "think" is probably what your bios update did....

I'm interested to try logging the clocks and voltages again, I have done this but didn't spot anything unusual at the time.... at least I didn't "think" it was unusual!

[Phage]

Grrrr. Gmail won't let me email an executable file. How can I get these to you ?