x1900 screen corruption/reboot (only on first run fullscreen apps)

[Sobek]

Tried to put as best a description I could into the title...

I bought this 512mb Sapphire x1900xt midway through 2005, so it's still under warranty. What happens is (and bear with me here as I try and cover every aspect), after I boot my system up, I can't play any fullscreen games, because as soon as the card switches into fullscreen mode, my system 'locks up' and i'm presented with a slightly-checkerboard effect screen that cycles through countless colour variations. I haven't been able to take a picture of this, printscreen naturally fails, but I just haven't had my phone ready at any stage to take a picture...

There is an exception though; Windowed mode. If I run a game (any game) in windowed mode straight up, it loads just fine. If I play any game in Windowed mode for a few minutes, then fire up any old fullscreen app, it works just fine from thereon out. At first I thought it might be a temperature related problem, as i've seen this sort of thing before where systems refuse to do fullscreen / 3d clocks right after the system has booted, needing time to warm up before operating normally. Thing is, my system will do this regardless of how long it's been on for. At one stage, I came in, turned my system on, hopped on MSN, then had to go do some things at the house...I came back in about 2 or 3 hours later, and without thinking just fired up CSS, only to be greeted with this same technicolour checkerboard screen.

I can almost always recover from this screen though... When it goes to this screen, if I spam CTRL-ALT-DEL to bring up task manager, in about 10-15 seconds it does, and I basically get Task Manager overlayed on a black background...when I reach this point all I have to do is kill the app's process, and i'm back at my windows desktop. I can then fire up a game in windowed mode for a few minutes, then return to CSS and have it function normally. This making sense? When the corrupted screen goes to switch to task manager after I spam CTRL-ALT-DEL, my monitor switches off for a few seconds then back on again, during which time I hear my videocards fan rev up as if i'd just turned my system on (is this a video reset a-la VPU recovery? I have VPU recovery disabled, by the way).

Hope that all makes sense.

Basically, what i'm wondering is, does this sound like something that would warrant having the card replaced under warranty? If so, the store I bought it from no longer stocks this particular card...what do you think would happen? Would they try and force me to receive a replacement card of less value as mine originally cost (as the best they now have are x1950 pro's, with the exception of HD2900xt's) or would I get something 'better' (like..oh I don't know, an HD2900xt which costs almost half what this x1900 did when I bought it). Or do you thin they'd simply send the card back to Sapphire or something and get a replacement from them? Any thoughts much appreciated. If I can get a screenshot of the corruption i'll post it :smile:

*edit* Oh and I forgot to add; about the 'reboot' part of my title.

Sometimes if i'm playing a game that's a persistent fullscreen game (such as Vietcong), when I spam CTRL-ALT-DEL for task manager and the screen blinks off momentarily, Task Manager can't actually steal focus so it goes back to the corrupt screen...this happens twice usually, then I get a super-corrupted BSOD that I can't read and my system reboots.

 

[Sobek]

I was scouring Sapphire's official support forums to see if anyone else had seen the same corruption I had. In the end, I found just one that was similar...;

That checkerboard-like corruption is EXACTLY what I see, though not while ingame like this user seems to have. It's normally just the black screen with the checkerboards overlayed, cycling through the various colours in the spectrum.

Is this something like the cold bug people always talk about? Or is that something else entirely. I also remember reading in a thread, here I believe, somewhere where someone mentioned that on some cards the memory controller (I think that was the culprit) just flat out refuses to work unless it either gets warm enough first (not my problem) or something else (sorry, can't remember what else was said). Sounds likely though.

Anyway, my real thoughts are whether or not this could be covered by warranty, and what the likelihood is of getting either a (cheaper) replacement 2900xt, or whether they would simply go to Sapphire for a new card. :smile:

 

[Jawed]

Looks like it is memory related. Which might imply memory clocks.

The 3D-windowed, then 3D-fullscreen is OK, sequence might be pointing to a too-high default clock setting for the memory on your board. The windowed mode may be running your game at desktop clocks (I can't remember if desktop clocks are lowered for both core and memory or if it is just core) and somehow that setting is having an effect on what clocks are being used for fullscreen gaming.

It's worth trying to strip out all your ATI drivers and reinstalling. I noticed yesterday that Guru3D has a thing called Driversweeper:

http://www.guru3d.com/index.php?page=driversweeper

but I've no idea if it's any good.

After you've tried that and made sure there's no OCing utilities gettnig their mucky paws on your GPU, you might as well RMA it if the problem persists.

Jawed

 

[Sobek]

Hey, a reply! :smile:

Between when this first started happening (whenever that was, I honestly can't recall...a while ago) I've done numerous driver cleanouts/reinstalls (and i'm extremely thorough on that regard). Anyway, just earlier I did another cleanout of all my drivers, and reinstalled with the absolute latest versions (was using Catalyst 7.1's to fix some issues with SC Double Agent, but i've finished that so it's back to 7.5, also using the latest 9.35 Nforce4 remix package). I have a good, thorough routine that I use, which is to uninstall my drivers, reboot into Safe Mode, run DriverCleaner, reboot, run DC once more (for good measure), then I install my Chipset drivers, reboot, then install my Videocard drivers, and reboot once more.

Everything is zipping along nicely...though the first thing I did upon entering into Windows again after installing all my drivers was fire up Vietcong, for the hell of it. Nada, same black screen with that technicolour checkerboard effect, it would seem driver issues aren't the cause (a short while back I even tried a really early 6 series Catalyst driver just to be sure, and it showed the same results). I'm also 1000000% (recurring) sure that there's absolutely no Overclocking buggery going on. I'm 100% against videocard overclocking as it shows no appreciable gain for the stress it puts the poor thing through, and as a result of that I have overclocking disabled in every way possible. I don't use the CCC, just ATI Tray Tools, and while installing I tick every option available to disable/remove/limit overclocking. I'll double check my clockspeeds both at desktop settings after a cold boot, desktop settings after a windowed game is run, and so on, to see if there are any discrepancies. The strange thing is that once I run a game in Windowed mode and 'fix' things, rebooting still results in games running fine (you'd think if it were a clockspeed issue, that rebooting would 'reset' that issue again, but it doesn't). It seems my system has to be off for quite some time before the problem returns.

I guess I might contact the store I bought the card from first thing Monday and see what kind of a deal I can hammer out. Surely replacing my $800+ card (at the time) with a measly $550 2900xt isn't outside the realm of possibility?

 

[Jawed]

I think you're obliged to try a bog standard ATI install before calling it quits, i.e. CCC and no ATT. I know that's paranoic, but PCs induce this. On some forums you'd be told not to come back until you've done a fresh OS install, lol.

Anyway, it looks like it's a duffer. Fingers-crossed a replacement card fixes it!

Jawed

 

[Sobek]

Hahaha, irony at its best...I actually reinstalled XP about 3 weeks ago. Thus making my current install technically 'fresh'

I don't know if ATT is affecting things in any way, if I don't run it at all and just leave everything at stock (ie; install all my drivers and just leave ATT alone), it still happens.

Oh well, i'll call up the store on Monday and harass/sweet talk them into a good deal

 

[Albuquerque]

Not sure that this makes you feel any better, but...

I just finished assembling a "new" PC for my younger brother. His current rig is a P2/333 with onboard Intel video, an 8Gb drive and 256mb of ram. I pulled together all my spare parts to build a P4/3.0ghz rig, 2gb of ram, a 160gb hard drive and a 550W power supply.

All I needed was a good video card, a half-decent case and a DVDRW drive. A it of shopping at Newegg resulted in purchasing a nice mid-tower case, a great DVDRW drive and a Sapphire AGP x1950GT 256mb card. With the processor at 3.6ghz and the memory at 480DDR, it should be a great rig, right?

So, card comes in, I assemble the whole machine taking care of routing the wiring, making the layout nice and clean, blah blah etc. I start at all-stock clocks for everything, boot the system, install windows, install drivers, blah blah and it's all great.

Overclock the box to 3.6Ghz / DDR480 and run Memtest86 for a day with no issues. I run dual sessions of P95 for 20+ hours with no issues. At this point, I haven't even touched the video card -- it's stock drivers, stock clocks, even stock AGP speed.

I start installing some games, and I start up FarCry -- it hard-locks the system before it even gets past the splash screen. ATI tool hard-locks the system after about 1/16th of a revolution of the "fuzzy cube" in the 3D Window. DXDiag even chokes the system. I wonder if it's the Cat7.5s, so I uninstall the video drivers -- and in the process, the machine hard-locks even trying to unload the drivers.

I run Prime95 again, this time for two solid days. Nothing. I reinstall windows, no luck. I install three different versions of drivers, to no avail. Three different power supplies result in no fix.

I drop my 7900GT-on-AGP video card into his rig -- and it works the very first time. I'm about to throw the 1950GT back in the box and ship it, when I get a wild hair and completely un-overclock the entire system.

The 1950 works now. Even 1 fsb overclock on the CPU results in the ATi card going absolutely bonkers, but if I'm at 100% perfectly stock clocks on the system CPU and memory, then the video card is fine.

I've never had this problem, ever. The processor and ram and system board are 100% stable at those speeds, but the video card for whatever reason can't stand it. I have no clue why, other than I can hypothesize that either the AGP lock isn't as good as it should be on the 865 chipset (entirely possible) or there's just enough extra "noise" at overclocked speeds to irritate the ATI card (but not my NV card)

I spent a week chasing that problem before I figured it out.

 

[Sobek]

That's a shocker of a story...don't you just love random hardware oddities? I sure do :smile:

But, you have given me some ideas. See, I run a DFI NF4 Ultra-D with a modded BIOS, and my CPU (X2 4600+) is currently clocked at (and has been since the very first day I got it) 2.90ghz. The thing is, I remember spending whole days going through god knows how many iterations of tweaking in various modified BIOS's, so as a result, I may have inadvertantly changed a setting that caused this (I know for a while there I ran my PCI-e speed at 105mhz instead of 100, but that was shortlived and I don't recall it causing any problems). The thing i'm having trouble remembering though is if this all started either at around the time I got my new CPU, or the time I switched to this BIOS...I just don't know. It's a real shitter to try and sort out...I've got an Ultra-D based system in here at work as my work system, so I think I might just bring my x1900 in here and try it out (this one is using the stock BIOS).

Guess that's all I can do at the moment...if that fails, and perhaps, if I can replicate this behaviour on an entirely different system, then i'll definitely be phoning the store I bought it from. Guess i've got all weekend to try out various things eh?

So Albuquerque, in order to keep using that x1950, you have to refrain from using any overclock at all? That's shocking...especially since you'd naturally want to clock the living shit out of that P4.

 

[Colourless]

Not sure that this makes you feel any better, but...

I just finished assembling a "new" PC for my younger brother. His current rig is a P2/333 with onboard Intel video, an 8Gb drive and 256mb of ram. I pulled together all my spare parts to build a P4/3.0ghz rig, 2gb of ram, a 160gb hard drive and a 550W power supply.

All I needed was a good video card, a half-decent case and a DVDRW drive. A it of shopping at Newegg resulted in purchasing a nice mid-tower case, a great DVDRW drive and a Sapphire AGP x1950GT 256mb card. With the processor at 3.6ghz and the memory at 480DDR, it should be a great rig, right?

So, card comes in, I assemble the whole machine taking care of routing the wiring, making the layout nice and clean, blah blah etc. I start at all-stock clocks for everything, boot the system, install windows, install drivers, blah blah and it's all great.

Overclock the box to 3.6Ghz / DDR480 and run Memtest86 for a day with no issues. I run dual sessions of P95 for 20+ hours with no issues. At this point, I haven't even touched the video card -- it's stock drivers, stock clocks, even stock AGP speed.

I start installing some games, and I start up FarCry -- it hard-locks the system before it even gets past the splash screen. ATI tool hard-locks the system after about 1/16th of a revolution of the "fuzzy cube" in the 3D Window. DXDiag even chokes the system. I wonder if it's the Cat7.5s, so I uninstall the video drivers -- and in the process, the machine hard-locks even trying to unload the drivers.

I run Prime95 again, this time for two solid days. Nothing. I reinstall windows, no luck. I install three different versions of drivers, to no avail. Three different power supplies result in no fix.

I drop my 7900GT-on-AGP video card into his rig -- and it works the very first time. I'm about to throw the 1950GT back in the box and ship it, when I get a wild hair and completely un-overclock the entire system.

The 1950 works now. Even 1 fsb overclock on the CPU results in the ATi card going absolutely bonkers, but if I'm at 100% perfectly stock clocks on the system CPU and memory, then the video card is fine.

I've never had this problem, ever. The processor and ram and system board are 100% stable at those speeds, but the video card for whatever reason can't stand it. I have no clue why, other than I can hypothesize that either the AGP lock isn't as good as it should be on the 865 chipset (entirely possible) or there's just enough extra "noise" at overclocked speeds to irritate the ATI card (but not my NV card)

I spent a week chasing that problem before I figured it out.

On those old chipset, the AGP clock rate and the FSB rate are linked.
By overclocking the FSB you're overclocking the AGP rate too. The AGP implementation on the 1950 probably isn't very tollerant of AGP being out of spec.

 

[hoom]

Some nForce mobos require setting the PCI clock to 101mhz to enable the PCI lock, otherwise OCing the FSB OCs the PCI & stuff like ATI AGP breaks.

I used to get a checkerboard pattern sometimes with my 9800np, it was from a cold boot & it was like every 3rd or 4th tile rendered black.
After a reboot it was invariably fine.

The screen corruption mentioned sounds more like trying to render at higher refresh rate than the monitor can handle though.

 

[Sobek]

The screen corruption mentioned sounds more like trying to render at higher refresh rate than the monitor can handle though.

It's definitely not that...

I have my refresh lock enabled, and in that I only have the few resolutions/refresh rates that I actually use (such as 1024x768 @ 100hz, 1280x1024 @ 100hz - generally the only one I use, and 1600x1200 @ 100hz), and the various games that I play each usually have their own Refresh Rate selection option in their launchers (again, Vietcong for instance is set to 1280x1024 @ 100hz in the launcher). Even it were some kind of refresh rate mishap, how would it explain the way running a game in windowed mode for 1+ minutes magically fixes it?

I'm leaning towards memory troubles...of all the Radeons that i've seen die or nearly die over the years due to memory faults, I have to say that the corruption pattern is pretty much identical to most of what I saw. Plus, i've yet to ever see, nor hear of, a refresh rate issue such as you suggest that causes a checkerboard effect that cycles through various colour spectrums...you really have to see it to appreciate it, just 10 seconds of looking at it and my eyes have a damned checkerboard imprint for a good 5 minutes after.

I suppose what I really care about at this point is opinion on what I might be able to get in regards to a replacement. If the store no longer stocks my particular card, or anything in the same class that's remotely as powerfull, would they be more likely to ship the silly thing off overseas to attempt some kind of replacement from Sapphire (if they even do that ), or would I be more likely to get a same-value-or-less replacement from instore?

 

[AlexV]

Remove ATT, it does weird stuff sometimes. If that doesn`t solve it...it may be RMA time

 

[Albuquerque]

On those old chipset, the AGP clock rate and the FSB rate are linked.
By overclocking the FSB you're overclocking the AGP rate too. The AGP implementation on the 1950 probably isn't very tollerant of AGP being out of spec.

Nope, this motherboard has seperate AGP, PCI and SRC bus lock speeds. I know it works, because there's no way in hell I've had this motherboard to 307FSB with a 2.4C and had my AGP unlocked 9500np working at the resultant bus speed.

The locks work fine, so that isn't the issue.

 

[Sobek]

I finally managed to capture a shot of it with my cameraphone (warning, quality is bad). But it does show the corruption clear enough... I was trying to launch Vietcong;

This was one of the few times where it just constantly cycled through various colours and checkerboard like patterns, but never did the 'reset' like before, so it never recovered...had to reset.

*edit* Forgot to ad; I did also try using just the drivers, with no ATT in sight. I uninstalled my display drivers, then uninstalled ATT, rebooted into safemode, cleaned twice, rebooted, installed my display drivers, rebooted once more, then fired up VC...same thing happened. I guess it's pretty much a given that the card is having troubles, i'm still just curious on what people's thoughts are on my possibilities with regards to replacement cards

 

[Sobek]

Argh damnit!

I contacted the store I bought the card from, but i'm outside of their 1 year Return to Base warranty...by 3, fricken', months. So now they're telling me that the only option is to return the card to Sapphire (who have a PATHETIC RMA system). They offered to take care of handling the card and so on for $25, so i'll probably take that route.

I just hope that the folks at Sapphire decide to be nice and replace it with something better, like my FX5950Ultra->6800GTOC BFG replacement. That was cool...

Goddamnit. 3 months!...