Sign in with
Sign up | Sign in
Your question

HD 5850 TWIN FROZR II display hangs, GPU resets, VPU recover

Last response: in Graphics & Displays
Share
October 8, 2010 7:35:04 PM

OS: WinXP 32bit

CPU: i7-920

MB: GA-EX58-UD3R

PSU: Thermaltake Toughpower Qfan 750W

RAM: 2 X 2GB OCZ 1333MHz

No overclocking.

EDIT: After giving the computer back to the shop and getting it back being told it's "fine" and wasting a lot of time looking around, I found OCCT, which apparently caused a lot of errors, flashing screens and freezes. Gave it back to the shop again and 24 hours later they replaced the 5850 with a 6870 with no extra cost.


First, note that this does not happen while playing Arma 2, which runs without any problems with GPU at 70-72 degrees. Playing other games temps rarely reach that (though they aren't much lower either, usually 65-58). Also note that Arma 2 is a DX9 game, though I don't think it should matter. World of Tanks beta also runs with no issues.

While playing Starcraft 2 (and on rare occasions while playing Dragon Age: Origins as well) I would have my display hang (sound would continue) for ~10-20 seconds, and then get a window of Cataclyst Control Center saying "VPU Recover has reset your graphics accelerator as it was no longer responding to graphics driver commands" with an option to send/don't send the error. The time between errors seems completely random, sometimes it would happen after playing for less than a minute and sometimes after playing for an hour+.

I limited SC2's FPS by adding frameratecap=60 and frameratecapGlue=30 in the Variables.txt file, and Dragon age already runs at 60 FPS with vsync enables, so it doesn't seem like the GPU is overloaded (not that it should reset when it does run at max! But just to verify).

I first thought this might be a driver issue of some sort, so I uninstalled everything related to ATI (including driver sweeper in safe mode), restarted and installed the latest (10-9) drivers. Still getting the same problem (at least in Starcraft 2, haven't tried Dragon Age). Temperatures are still higher while playing Arma 2 yet Arma 2 never causes any issues at all. No "VPU Recover" message though since Cataclyst Control Center is not install, but same thing (display hangs and then GPU resets). One time I also had SC2 crash but usually it just resets and starts working again.

Tried running 3dmark for a bit, FPS was in the hundreds, no resets (though I didn't go through the whole thing, as even if it did reset it wouldn't have given me too much info).

I had in that same machine an HD 4890 before, had no problems, but also haven't played Dragon Age nor SC2 with it.

Video Card Stability Test (http://freestone-group.com/video-card-stability-test/fr...) ran for a few minutes with no issues.


Now my question:

Anything I should do before RMAing the card? Any tests I can run that will prove that the GPU is faulty (so that I can RMA it without telling them to play SC2 for a few hours hoping it'll happen to them too)?


Thanks for any help
October 8, 2010 8:26:50 PM

Hmm well the only thing you might want to do is run FurMark at the biggest resolution possible with 4xMSAA. If it's fine after 10-15 mins then chances are the problem isn't card related, although certainly that does seem to be where the problem lies. These issues were pretty common and some people are just unlucky it seems, because my 5850 hasn't done this sort of thing since probably 10.3 drivers.

You may want to try 10.8 or wait for 10.10 to come out soon and see if those help.
m
0
l
October 8, 2010 9:01:24 PM

FunMark ran fine for 17 mins on 2560X2048 and 8X MSAA. max temp reached was 81 degrees yet nothing hanged or crashed.

I'm assuming that my PSU is more than enough, yes?

Played some Resident Evil 5 again too, no issues.

I would go complain at Blizzard if it was only SC2, but Dragon Age: Origins was causing that too every once in a while, so it's not a problem specific to only one game.

Before using 10.9 by drivers were 10.6, both had same problem.
m
0
l
October 8, 2010 9:03:32 PM

Have you tried changing Catalyst AI settings while playing those games, or have you tried the Crossfire Profiles (not sure why but I hear that can help with just a single card)
m
0
l
October 8, 2010 9:06:59 PM

Nope, no idea what those are supposed to do. Anything specific I need to be trying to change?
m
0
l
October 8, 2010 9:09:19 PM

Well you can try Catalyst AI off, standard, and advanced. Set it to off, or if already on off, try standard.
m
0
l
October 8, 2010 10:07:12 PM

All 3 settings still had the card reset. Advanced took the longest before it reset but it could be just luck.
m
0
l
October 8, 2010 10:20:06 PM

Hmm yeah. I guess you can see what they say over on the AMD forums, I'm sure others have had this issue but yeah it might be RMA time. :( 
m
0
l
October 8, 2010 10:22:50 PM

I don't know, how will I RMA it if it goes through something like FunMark without causing any issues? They'll just test it, say it's fine and blame the game (which may or may not actually be correct), so I need to find a way to reproduce this somehwat reliably.
m
0
l
October 8, 2010 10:41:53 PM

True. Well in SC2 are you enabling AA? If yes, does it run well without it?

Can't say much about DA:o , because I have it on PS3 (and don't really like it).

I dunno... I guess you'll just have to try scouring the net for any solutions because right now I'm out of ideas. Sorry.
m
0
l
October 9, 2010 12:06:27 AM

No AA for SC2 - Didn't touch any of the driver settings, only set everything in all games to max (except Arma 2 where everything is max except post processing which is ugly and view distance which kills even the most powerful PCs on max).
m
0
l
October 9, 2010 1:16:14 AM

I'm having a similar issue with my Asus 5850 Direct Cu.

I play Dragon Age, Fallout3, Dirt2, and a few other games.

My system will lock randomly (usually doesn't recover) when playing Dragon Age, while running 3DMark Vantage (never makes it past 3 runs) and also locked while playing Left for Dead 2.

Furmark ran twice for 30 min each time without issue though.

My system was stable for about 3 months with the card, and it's still stable if I use my old 3870. I had the card in another PC and it played BFBC2 for several hours without issue, and I already had the card RMA'd once to which Asus found "no issues".

In my situation I've already tested by removing all hardware and software (other then the absolute necessary), moved the card in to another PCIe slot, and swapped power supplies with no change.

Only thing I can think of is maybe there's something wrong with my mobo. Maybe power from the PCIe slots?

If you have the option to loop the 3DMark test, try to let it run for at least 3 passes and see if it fails.
m
0
l
October 9, 2010 11:13:08 AM

Well 3DMark03 ran 4 times with no problems (GPU temp didn't even go up much). I really need to find more games that reproduce this, so far only got this problem with Dragon Age (not often) and Starcraft 2 (quite often).
m
0
l
October 9, 2010 4:14:43 PM

I know this is a bit of a long shot, but it might be worthwhile to backup your windows and then format and do a fresh install. If you don't have a second HDD you can partition your current one to put the backup on.
m
0
l
October 11, 2010 4:16:36 PM

In SC2 I keep getting it to happen in the same place in the same mission, and as soon as I play for ~1 second it hangs again. This combined with a lot of others having those issues, make it seem like it really IS a driver/game issue rather than a hardware issue. But of course go figure how one should prove this to ATI/Blizzard...
m
0
l
October 18, 2010 12:05:08 AM

The more I read the more this seems to be a SC2-specific issue (though I also get this in Dragon Age, I haven't really found much talk about it. Dragon Age simply isn't as popular and doesn't cause this issue as often). Some people claim that even a 1MHz overclock/underclock of their GPU solved it. A friend of mine had to downclock his HD4850 from 650MHz to 500MHz (when he tried to downclock to 600MHz it wasn't enough).

I reduced my clock speed to 749MHz (down from 750MHz) and it still hadn't crashed, but I also haven't really played enough to tell. I'll post next time it gives me the black screen BS again.

EDIT: That didn't help. I'm trying 700MHz and 950MHz memory. Though to be honest I really don't see why the hell I need to downclock a non-overheating card to make SC2 work (when other games work great). Maybe SC2 does strain the card in some special ways and I really should RMA it... Anyone actually uses that card with SC2 successfully without downclocking?
m
0
l
October 20, 2010 12:38:42 AM

Seems stable with 700MHz. Does this mean it's RMA time (as in, does it mean it's actually the card's fault and how can I prove it)??
m
0
l
October 20, 2010 4:40:06 AM

Dude Open you case and use box fan i did that and it work on 2 computers that had the same problem.
m
0
l
October 20, 2010 9:34:36 AM

It's not overheating, already mentioned in previous posts. Was stable as a rock running funmark at 81 degrees. When it freezes in Starcraft 2 temps are usually ~65... Even Blizzard's initial response to such issues is that the GPU is overheating (as SC2 with default settings does tend to overheat GPUs), but I already set max frames and am using a great card with proper cooling so the temps don't even get close to what they can get. So either the card is broken or it can someone make some very weird faliure in the card trigger only when running at stock speeds and not when running at lower speeds, that have nothing to do with temprature (or at least nothing to do with temprature shown by ATI tray tools...).
m
0
l
October 21, 2010 10:37:16 PM

I ended up putting a new system together (thanks to the lightning strike that insurance paid for) with a new mobo, cpu, memory, and another Asus 5850 Direct Cu.

My system is running fine now with the card that was suspect. It's running over clocked too. I really think it was a motherboard issue.
m
0
l
November 3, 2010 12:45:47 PM

Still unresolved. Downclocking to 725MHz doesn't even seem to be enough. Only 700MHz seems stable (though only time will tell). It seems to be running on 1.088 volts which is supposed to be fine.

Any ideas? If something needs to be RMAed I need to have a way to show that it indeed needs RMAing. I can't just tell them "take this PC play Starcraft and see". :( 
m
0
l
November 3, 2010 12:58:57 PM

I think the only way you'll know for sure is to put the card in another system and test it.
m
0
l
November 3, 2010 1:03:12 PM

These sound like driver or OS issues to me. Try chaning your drivers with a full driversweep. If this doesn't work try reinstalling the OS and just installing the games to test.
m
0
l
November 3, 2010 1:07:24 PM

Drivers already sweeped and reinstalled (a lot of people having these issues with Starcraft 2 were also told to sweep reinstall drivers but it didn't help them either, only downclocking GPU solved their issue).

I'll try a reformat when I have the time to do it, haven't reformated in a while.

Is there no useful infromation that can be attained from the fact that stability is gained when the GPU is downclocked (to lower than factory settings)? From googling it a lot, it seems that for other people having pretty much the exact same problem and tried everything the only thing that worked for them was downclocking GPU as well.
m
0
l
November 3, 2010 1:14:27 PM

As I may or may not have mentioned, I had a very similar problem (save for mine was unstable in any 3d app) and it wasn't resolved until I moved the GPU into another system. It's the only way to rule out a hardware issue somewhere else.

Like you, I re-installed, changed drivers several times, even bought a new P.S. and ultimately it was my mobo that was the issue.
m
0
l
November 3, 2010 2:53:01 PM

If I do get it working on another PC, then how do I prove that it is indeed a mobo issue? Is there any way to actually test those kinds of things?
m
0
l
November 3, 2010 3:34:58 PM

Well, if you put the card in another PC that is already working and the suspect card works fine, it's something else.
If you put it in another PC and you run in to the same issues it's the card.

You really need an entire spare system to change out individual parts until you determine what is the cause. It takes time and can be a pain, but there is no simple, quick solution.
m
0
l
November 3, 2010 3:36:49 PM

What I mean is that even if I do that and find out which part is bad, I'll know what needs replacing, but still won't really have a way to show the shop the MB or GPU is bad, and they probably won't take my word for it (though they might, so I guess it's worth a try).
m
0
l
November 3, 2010 3:44:06 PM

They would try to recreate the problem. If you bring the whole PC in and they can recreate the issue, they also will have test systems there so they can check each component. Just make sure to explain exactly the issue and how it comes up.
m
0
l
November 3, 2010 3:46:33 PM

The problem is that it occurs only in Starcraft 2 and only after you play a while (and in Dragon Age: Origins after you play for a very long time). I'm not sure they would be willing to judge the MB's failure based on a single game. Especially when SC2 requires an account to actually log in and play (though I could just tell them to watch replays which can hopefully be done in offline mode).
m
0
l
November 3, 2010 3:48:07 PM

You can always try to RMA the card and see how that turns out. It would mean you'll have to wait a week or two and pay for shipping.
My PC shop charged me $20 to diagnose the PC, and had it for 2 days just to tell me that they didn't have the parts to swap in, and they basically just told me what I already knew....
m
0
l
November 3, 2010 3:51:43 PM

Well my PC is fully on warranty, bought all parts at the same place, so as long as there is a way they can prove that there is a hardware issue... But, since all tests come out absolutely fine, I don't see how that can be done without them actually sitting and playing starcraft, and again since it's just 1 game they can always blame the game.
m
0
l
November 3, 2010 9:45:16 PM

galzohar said:
Well my PC is fully on warranty, bought all parts at the same place, so as long as there is a way they can prove that there is a hardware issue... But, since all tests come out absolutely fine, I don't see how that can be done without them actually sitting and playing starcraft, and again since it's just 1 game they can always blame the game.


I would just give them a call and tell them the situation, hopefully they can just tell you if it's worthwhile to bring it in or not.
m
0
l
December 17, 2010 8:36:34 PM

It seems with 700MHz (which completely solved the very frequent issues with SC2), when playing Unreal Tournament 1999 I still get occasional freezes and recoveries (about once every 30-60 minutes, though I had 2 about a minute apart as well).
m
0
l
December 17, 2010 8:52:08 PM

Necrooooooo

Well, UT 99 is almost 12 years old. Can't expect it to run perfect on todays advanced hardware.
m
0
l
December 19, 2010 11:38:47 PM

Except it does run perfect other than this issue that is exactly the same in some modern games. Oh well, I guess I'll have to take advantage of my brother going on a trip and play mix&match with his PC to figure this out. But it's seriously one of the most mysterious problems I've ever seen.
m
0
l
December 20, 2010 4:01:18 PM

For now it seems like everything works fine with the GPUs swapped (5850 in my brother's PC and 4890 in my PC) even in UT99 and SC2 on my PC (only tested SC2 on my brother's PC). Both PCs used the same (most recent) drivers when using the same card.

Does this necessarily that this is a MB issue, or can it be something else?

I guess the last thing I have left to try is to place the HD5850 in the other slot... Will try that after I spend some more time making sure the 4890 is truly stable. But anyway I really need to be able to be very specific when I give the PC back to the shop if I actually want it to be fixed (not to mention if I actually want it done in a timely fashion).

One more thing I noticed, is that I can still talk on teamspeak while the game is frozen.
m
0
l
January 15, 2011 5:33:40 PM

I've been having the same issue (and a friend of mine too) with the 5850. I think it started happening when i did an update of my drivers around the start of December. Before that, no problems. Reverting to previous drivers doesn't help afaik. Haven't tried a whole lot of them though, and i do not remember what driver i was on before. But i've read about this problem on some other forums too, and no solution so far. Doesn't seem like a hardware issue imho.
m
0
l
January 15, 2011 9:12:47 PM

Did you also manage to get it to stop by underclocking?
m
0
l
January 15, 2011 11:22:24 PM

When i underclock to 700Mhz my 2nd screen starts to act weird while running a game. A horizontal line comes down and "refreshes" the screen. Also some flickering sometimes.
m
0
l
January 16, 2011 6:08:57 PM

Hmm interesting I had something similar when I tried to put my old 4890 in, before I changed the ATI Tray Tools settings for the correct clock rates (4890 runs at quite higher clocks than 5850 so it was underclocked by quite a bit), but with the 5850 downclocking by 25MHz doesn't seem to cause it for me.
m
0
l
January 16, 2011 8:54:30 PM

Hmm i noticed i said 5850, but i do own a 5800, sorry. I've done a clean install of my drivers now since the flickering was really annoying. It has stopped now but the gray screen of death keeps on popping (although only one did so far).
m
0
l
January 23, 2011 12:48:46 AM

After reading about someone who updated his PCIe drivers to solve this problem, I looked for something that seems appropriate on the MB's page:

http://ee.gigabyte.com/products/page/mb/ga-ex58-ud3r_10...

But could not find anything that seems like it should help.

Then I went to driver agent which was recommended on that page, and got those results:

http://driveragent.com/0/driveragent_results?hwid=f7e2c...

Of course to actually download anything they ask for money, and I wouldn't trust them with a penny, and besides, I already paid for my hardware, I shouldn't need to pay for drivers. So I went and googled for " Intel(R) ICH10 Family PCI Express Root" and found this:

http://www.pcpitstop.com/drivers/download/Intel%28R%29~...

Now I wonder if I should just go and update all of those things. To be honest I don't really know which driver websites I can trust and which I can't. But it is interesting that DriverAgent showed mostly PCIe and SATA problems, especially after this: http://www.tomshardware.com/forum/285555-30-computer-st... (though then again windows didn't even start when I had that problem so I'm not sure it can actually be related).
m
0
l
January 26, 2011 8:38:08 AM

I tried reinstalling CCC (which is a project in itself, took me a while to find out I need to also remove a bunch of stuff from windows\assembly folder), setting a profile, and removing the higher 2 "stepping" options in several places the xml file (which was also quite hard to find).

So far I haven't played enough to really tell if it hangs or not, but it definitely fixed my "takes forever to alt-tab back into the game" issue.
m
0
l
January 26, 2011 9:56:09 AM

Alt tab problem was fixed (which is at least one good thing), but I'm still getting VPU recovers.
m
0
l
January 30, 2011 9:23:59 AM

Well apparently even the alt-tab problem wasn't truly fixed, plus I still get the GPU core speed jump from 157MHz to 700MHz (700 is after I lowered it in order to be able to play Starcraft 2, factory setting is 725MHz) in 2d applications for no good reason (GPU%=0). And of course still getting VPU recovers when GPU is set to 725MHz.

Issue still unresolved...

I'd RMA it but they did run all tests and those went through fine, so I need to find a test that fails first.
m
0
l
February 10, 2011 7:56:35 PM

Reformatted and installed win7. Still same problem (except this time it didn't recover, just stayed frozen).

Still can't find a test that fails, and without one there won't be much use bringing the PC over to them again...

At least win7 did make my G35 headset work properly and I'm happy with it so far.
m
0
l
February 10, 2011 9:50:31 PM

Your problem sounds very similar to what I had before (as I've mentioned...) and my only resolution was a new motherboard. My 5850 just refused to play nice with my Asus Maximus Formula (x38).

I put the same card in a new system and it ran like a champ.
m
0
l
February 23, 2011 8:05:11 PM

Can anyone point at some less common stability tests I can try?

After trying just about anything, I'm rather convinced it's some kind of a stability issue that only happens on very specific things that the standard tests don't reproduce and thus I can give my PC back 100 times they'll just give it back to me saying it's fine, unless I find a way to reproduce the issue reliably somehow.
m
0
l
February 24, 2011 11:27:10 PM

Oh and with win7 it seems to actually never recover - just freeze.

OCCT fails miserably. Artifacts, errors, flashing screen, computer freezing/slowing to a halt, black screens, etc etc.
m
0
l
!