Radeon 5xxx Instabililty/Corruption Issue

[swaaye]

I bought a notebook last week with a Mobility 5870 and of course stress tested it to find out if it worked fine before the return window ran out. I quickly found that with Catalyst 10.3 that it will get the grey lines / screen corruption / black / white screen freeze (it varies) quickly in 3D apps. It's stable with the notebook manufacturer's drivers (9.12-based) though, but I've found that other people are fine on 10.3. I'm in the process of doing an exchange for a new notebook now.

Over the past few days I've done some searches for these grey screen / black screen / stripe / corruption issues with the 5xxx cards. I am surprised to find that there are many posts around the web on a lot of different sites about this. Some news sites have articles about it too. There's even a hotfix to "alleviate" the issue (ie reduce it) for 58xx Cypress-based cards, but it's not always effective.

I'm inferring from this bit of research that ATI is having problems with their GPUs in some way that they can't debug effectively. Notebook graphics chips and desktop addon cards are passing QA testing and making it to customers and then showing serious, unsolvable problems. Replacements seem to solve the issue, meaning it is some sort of hardware flakiness with the graphics card. Some people seem satisfied with tweaking Powerplay settings or downclocking, but that's not exactly a reasonable cure.

Here is a Tom's Hardware article I found that collates info about the problems:
http://www.tomshardware.com/news/ATI-Radeon-Gray-Screen-Crash,9529.html

This is the worst that I got from my notebook during testing. It would usually do some flickering before crashing completely and requiring a manual power off. This was a windowed Furmark session, but I initially ran into the problem during some Serious Sam HD. The problem didn't seem to be temperature related either, because it would crash almost immediately sometimes, before it could heat up.


How many here have run into these kind of instabilities with their GPUs? Silent Buddha, I found some of your posts when I was searching B3D.

 

[Valzic]

No problem here... XFX 5870 factory o/c to 875 Mhz. I have had for 2 months now.

 

[eastmen]

HardOCP has articles and a very active topic on it. Ati seems to be working on it and fewer and fewer cases are reported with newer drivers. I believe they are now having some users send affected cards that recent drivers have failed to fix to them so they can test it on problem hardware.

http://hardforum.com/showthread.php?t=1497931

 

[swaaye]

It's nice to see that AMD is acknowledging the issue.

 

[Tchock]

Is it the Asus G73?

I've been hearing RAM issues through the local hardwarezone forums.
And yes quite some of the GSODs were due to unstable/problematic RAM configs that didn't really manifest themselves with prior card (but would with stress testing)

 

[Silent_Buddha]

How many here have run into these kind of instabilities with their GPUs? Silent Buddha, I found some of your posts when I was searching B3D.

Well for me, the problem is better but still not completely solved/fixed. This is with a desktop 5870.

Now my issue boils down to.

Sleep mode. Come out of sleep mode. That's the trigger to allow the problems to surface on my system.

Then if I run a 3D application (say Dawn of War II which I'm playing now) I'll get the grey stripped screen, but VPU recover will prevent total system lock (which was a problem in the past).

Unfortunately, while it prevents system lock, I have to reboot the computer after due to random graphical corruption on the desktop even after exiting the game.

Workaround for me is to just use hibernate mode (no sleep or hybrid sleep) in which case the problem never manifests, but my computer is slower to come out of hibernate than sleep.

My last test was with the preview 10.3. So there's steady progress being made. Almost got the Asus laptop with the mobility 5870, but couldn't justify the price for use 3 months out of the year while I work in Japan. I'll just have to stick with my aging Shuttle SFF with a 4890.

Regards,
SB

 

[swaaye]

SB, that sounds extraordinarily annoying. Sounds like defective hardware, plain and simple.

 

[Mintmaster]

SB, that sounds extraordinarily annoying. Sounds like defective hardware, plain and simple.

That's seems highly unlikely if the card is fine out of hibernation and only shows problems going into sleep mode. Maybe the motherboard is at fault, but it looks like a driver issue to me, e.g. some GPU configuration registers not being set properly while the machine wakes up.

 

[Silent_Buddha]

It's been a combination driver/MB bios issue.

Problem was initially quite bad when I first got the 5870 along with the Gigabyte P55-UD4P. F4 BIOS revision for the MB cleared up quite a few issues (extra large mouse cursor, less frequent grey screening).

Since then the issue has gradually gone down and down with each MB BIOS revision (up to F8 now) and with driver updates.

At least now, VPU recover can recover enough that the system doesn't lock. Initally locked after 1 grey screen, then either a driver update or BIOS update made it so VPU recover attempted to recover, but would still lock after multiple successive grey screens in quick order if you didn't very quickly close the application when the normal screen flashed on before grey screening again.

Now I can safely test new driver's + Sleep mode without fear of computer lock and the potential problems associated with a forced power off.

Interestingly, back on either 10.1 or 10.2 it would take longer before grey screening after sleep (sometimes taking 2-3 sleep on/off cycles), but VPU recover wasn't able to properly recover and computer would hard lock. Now, it is back to grey screening after the first sleep cycle (if stressed by the 3D accelerated application), but VPU recover can recover enough to allow a proper shutdown and restart.

It must be noted that I also upgraded the MB BIOS at the same time. From F6 to F8. So I don't know if it's the MB BIOS that has had mixed results or the driver, or a combination of the two. Since hibernate works, I'm not in a hurry to revert the BIOS to F6 to try to isolate which change was caused by which update.

Regards,
SB

 

[swaaye]

It still wonder if a replacement video card would work differently. Have you used other video cards on that mobo?

I've been using sleep mode with my 3870, 4670, and 8800 on 3 separate mobos, Intel and AMD, for many months (years for the Intel mobo) without problems. P35, GF 8200, and 780G chipsets. Come to think of it, I can't remember having sleep/standby problems unless I think back to before Athlon 64 / P4.

 

[Broken Hope]

Can't say I've ever had a single GSOD with my 5870, that's over 2 different systems, my old X38 Core 2 Quad system and now my Gigabyte X58 i7 system. I'd be perhaps looking at the video card if it had persisted this long.

 

[nate240]

I have the same issue with my 5870. I had grey screens left and right until I updated my bios to the latest version. Unfortunately it's old enough now that I don't think they're updating it anymore (abit IP35 pro). The issue mostly went away with the bios update and hotfix drivers, but has been happening more lately (10.3a). It used to always happen coming out of sleep, never anywhere else. Now it has happened a couple times when browsing/etc in addition to coming out of sleep. It is a LOT better, but still annoying. I have absolutely no problems during gaming.

 

[swaaye]

P35 is a ~3 yr old mature chipset. Abit is long gone too so I doubt anybody's writing BIOS updates for their boards. It sounds like 5xxx is sensitive to something, but I'm not sure I'd blame motherboards.

I've read that the issue boils down to some sort of GDDR5 memory "training" problem. Some 5xxx cards work perfectly, some are horribly unstable, and everything in between. I've read of grey screen issues all the way back to 4870.

With the notebook I just RMA'd, with the Mobility 5870 (that's a desktop 5770 essentially), it was stable with the ASUS ATI drivers available (9.12 beta), but with any 10.3 set it would flicker and corrupt quickly (within minutes) in several 3D apps I tried. It did none of this on the 9.12 beta set. So something very strange is indeed going on and I'm sure it varies between cards which is why they can't nail it down. I sent the notebook back because I'm not going to bet on it working in the future on my dollar.

 

[willardjuice]

I still get them from time to time.

 

[Silent_Buddha]

It still wonder if a replacement video card would work differently. Have you used other video cards on that mobo?

I've been using sleep mode with my 3870, 4670, and 8800 on 3 separate mobos, Intel and AMD, for many months (years for the Intel mobo) without problems. P35, GF 8200, and 780G chipsets. Come to think of it, I can't remember having sleep/standby problems unless I think back to before Athlon 64 / P4.

Since there aren't any problems with using Hibernate, I haven't been inclined to test out another video card. If NI isn't out by the time I get back from Japan, I may throw the 4890 in there just to see. I doubt it'll have any issues though as I don't think I've seen any issues with 4890 + this MB.

Right now, I just don't have the time, so a longer startup time from hibernate is a good enough workaround for me.

I have a theory that it might be something related to memory transfers to/from main memory. As that's one of the things constantly changing in the P55-UDP4 BIOS. The original F4 BIOS that fixed many of the issues was released mainly to fix a corruption and instability issue in GOM player and other media players. Whatever it fixed also had an impact on the 5870 problems.

That might explain why sleep mode is somehow exacerbating the problem since the machine state is saved to main memory when going into sleep mode. No idea if video memory also gets swapped out to main memory when going into sleep mode however. Although wouldn't surprise me as Vista/Win7 have unified memory that includes the video card now right? Or something like that?

Regards,
SB

 

[Psycho]

I remember my old P35 based motherboard had a bios option to reset the graphics card when returning from S3, and without that I would get 3d problems after sleep (on the 4850 back then). Maybe that's the thing missing from some (atleast older) MB's?

 

[stevem]

I remember my old P35 based motherboard had a bios option to reset the graphics card when returning from S3, and without that I would get 3d problems after sleep (on the 4850 back then). Maybe that's the thing missing from some (atleast older) MB's?

Yes, I recall the same. Updated BIOSes removed this option for some reason, so I edited it back in. Problem solved. My new boards don't have the option, but I haven't had a problem and am not running Radeons ATM.

 

[Silent_Buddha]

I remember my old P35 based motherboard had a bios option to reset the graphics card when returning from S3, and without that I would get 3d problems after sleep (on the 4850 back then). Maybe that's the thing missing from some (atleast older) MB's?

Interesting, next time I reboot my computer, I'll have to see if there's anything like that in my BIOS and give it a shot if there is.

Regards,
SB

 

[air_ii]

I haven't had any issues. I had GSOD only once, but that was after hitting 1.05GHz on default voltage and running FurMark (1GHz run just fine).

 

[nate240]

I remember my old P35 based motherboard had a bios option to reset the graphics card when returning from S3, and without that I would get 3d problems after sleep (on the 4850 back then). Maybe that's the thing missing from some (atleast older) MB's?

Ok, so I went and downloaded the manual for my motherboard to see if there was this magic switch, and as I was scrolling through the pdf what happens...? Yup, another driver has stopped responding crash. Can't remember if there was a gray screen though.