ATI Driver Crashing

I've read a thread or two on here already (that I found from Google search) that seemed related but they didn't really help much. Considering different cases are usually unique, I wanted to go ahead and ask about this specifically.

I put together a computer for a friend using my old parts (since I upgraded). It worked fine here, though I didn't really do anything on it like watch YouTube. Here are the parts in it:

OS: Windows 7 x64 (a fresh install on a formatted HDD)
Video card: ATI Radeon 4890 - HIS was the brand, I believe.
Motherboard: Newegg.com - ASUS M4A77TD AM3 AMD 770 ATX AMD Motherboard
CPU: AMD Phenom II X4 945
RAM was OCZ Platinum DDR3. I had to manually set the BIOS settings for the timings (since ASUS motherboards apparently don't automatically config some OCZ RAM) for it to work properly (computer would randomly lock up completely otherwise, had to manually power off). Mentioning in case this is a factor.
Monitor is a discontinued LG 22", resolution of 1680x1050 (one of the threads I read here pointed out that they were using this resolution too and that it could be incompatible with windows 7?)
Note that all these parts I used to use, except for the motherboard and RAM (I used OCZ Gold). I also used that same resolution, on dual monitors, and never once had any issues. That monitor was actually mine.

What's happening is that the video driver keeps ending unexpectedly (so the screen appears to freeze, flashes off/on, appears black and white, then gives that popup in the system tray about it crashing). Sometimes, it does more than that though, and the colors get all messed up on the screen. Like the desktop background will be a bunch of different colors or other things will just be the wrong shades. The computer has to be rebooted then for it to correct. It also adds to the Event Log about the driver crashing (as type Warning). It's been doing this since they got it, but they mostly just use it for those Steam games since they use their laptop for everything else so it was never completely looked into yet.

The odd thing is that this doesn't happen while they're in-game (on Steam, such as Left 4 Dead and Audiosurf. just those two games actually, don't have any others). It only happens when they're doing normal stuff which is mostly browsing the internet. It started happening very frequently, like every several seconds, once they went on YouTube to listen to some music. I've had them download the latest video driver from AMD's site and install it (didn't do an uninstall of it at first either, just wanted to mention that in case) but it didn't help. As for flash, I had installed the latest version from their download page, which should've been 10.0.* and not 10.1 in case someone suggest that as being the problem. I did this a few weeks ago.

Knowing all this, what are the possibilities that it's a hardware problem like the video card or motherboard dying? Could it just be a corrupt install of the video driver that cleanly uninstalling it first could solve? Could it be something to do with adobe flash? Any other ideas?

Thanks.

Hello Zerxer - welcome to the forum.

Most likely you are getting a Timeout Detection & Recovery error - "Display driver xxxx has stopped responding and has recovered" - which means that the video card is not responding as expected, and that condition will cause your symptoms.

And yes, memory issues are a primary cause of TDRs. You must get the system to play nice together. I would focus on getting some stable settings with that memory on that board or look at getting some compatible memory.

You should run Memtest Memtest86+ - Advanced Memory Diagnostic Tool to test the memory under various settings until you get a clean pass 5 or 6 times.

Check also that you have the correct type of memory for that motherboard.

There are of course several other factors that can cause TDRs, including defective components and poorly installed or incompatible drivers, but yours sounds like a real case of memory timing issues.

Test that out and let us know. We can look at other factors from there.

Have a look at the Qualified Vendor List for that board: ASUSTeK Computer Inc.-Support-

There are only a few OCZ kits that board likes. My guess is that in order to get that board to play nice, you're going to have to find a kit on that list?

Zerxer said:

(computer would randomly lock up completely otherwise, had to manually power off). Mentioning in case this is a factor.

Click to expand...

As mentioned by TVeblen, that sounds like a prime candidate for a memtest run. Incorrect memory settings can play havoc at any given time.

Monitor is a discontinued LG 22", resolution of 1680x1050 (one of the threads I read here pointed out that they were using this resolution too and that it could be incompatible with windows 7?)

Click to expand...

Not an issue.

Knowing all this, what are the possibilities that it's a hardware problem like the video card or motherboard dying?

Click to expand...

Not high on the likely list.

Could it just be a corrupt install of the video driver that cleanly uninstalling it first could solve?

Click to expand...

If the memory passes memtest, then a fresh re-installation of the drivers would be my next port of call.

At this stage, I'd err on the side of it being a memory issue. I was getting this issue using some aggressive RAM timings and Cat 10.4a preview drivers. The issue got increasingly worse the more I adjusted the RAM until it was unusable because of the time outs.

However, I cleaned and installed that driver and used a newer beta version (8.74) and the issue went away :huh:

Thanks for the replies, guys. It's good to know that RAM can be such a likely suspect for an issue like this. I've been suspicious of that RAM from the start because of my initial problems. That just seemed basic though since it was just that the motherboard wasn't setting the proper timings automatically. All I did to correct that was manually input the timings listed on the product page on newegg for the RAM and then that problem went away. That problem being the computer randomly freezing at time and having to be shut off.

I'll try to walk them through running memtest (unfortunately it was shipped to an out-of-state friend so I can't be of help in person). Now, is it pretty much 100% that memtest will fail some of its pass throughs if the RAM is the culprit? Or is it possible that it'll pass with flying colors and still have the possibility that the issue is the RAM?

Zerxer said:

Now, is it pretty much 100% that memtest will fail some of its pass throughs if the RAM is the culprit? Or is it possible that it'll pass with flying colors and still have the possibility that the issue is the RAM?

Click to expand...

The annoying answer is Yes and Yes.

For memory to be good it should pass Memtest at least 5 or 6 times to 100% without any errors. The longer you run it, then the more thorough the test, and the more likely to turn up any errors. I like to run it over night whilst sleeping.

Unfortunately there have been instances where memory that has passed Memtest with flying colors has ultimately turned out to be bad.

That is why it is important to check to see if your exact model of memory is on the motherboards tested and recommended list. If it is then you can be more confident that a Memtest pass will eliminate memory as a cause of the TDRs.

Zerxer said:

I'll try to walk them through running memtest (unfortunately it was shipped to an out-of-state friend so I can't be of help in person). Now, is it pretty much 100% that memtest will fail some of its pass throughs if the RAM is the culprit? Or is it possible that it'll pass with flying colors and still have the possibility that the issue is the RAM?

Click to expand...

It's very hard to say mate.

I'm not sure how familiar you are with memtest, so I'll err on the side of caution and provide more information rather than less, just in case.

Memtest should be run with a minimum of 8-12 passes. Tell your friend to run it overnight, as it as exciting as watching paint dry. However if it starts throwing up errors straight away, halt the test. (Although as TVeblen mentioned, it's still not a 100% guarantee that the RAM is okay)

If it does show errors, then it's a case of determining if it's the modules themselves, or the timings.

If no errors appear on the RAM, your friend may have to uninstall the drivers, run driver sweeper in safe mode and then re-install the drivers.

One thing I forgot to mention was that I also had a 4890 with the Cat 10.4a drivers and never had an issue with display driver failing and recovering, even with the aggressive RAM timings.

It only started to occur when I switched to a 5970.

If the problem persists even after a fresh driver clean / install you may be looking at a faulty card.

To eliminate the card as faulty, you could get them to download MSI After burner + Kombuster, or Furmark to stress test the card. If it artifacts badly, the card may be a dud.

(Got bored of doing estimates - so I just looked up that motherboard on Newegg.)

Thinking about the statement: "ASUS motherboards apparently don't automatically config RAM"

First, know that the native speed for the memory on that board is 1066. Anything you set higher than that is "overclocking". If you bought faster memory (1600 for example) it will run just fine throttled down at 1066. For diagnostics purposes you should run the memory at the default motherboard speed - 1066. You can play with "overclocking" it to 1600, or whatever, later.

The people who were complaining were having issues with getting that board to run faster memory and keep a stable system. It just makes more sense when you are having stability issues to run everything at default settings. The simplest way to do this and reset everything back to "normal" is to "load BIOS defaults" in the EXIT menu of the BIOS.

Some people also complained about voltages on that board. Meaning they needed to manually change the voltages to get correct (spec) readings. Not many though. There were surprisingly few negative reviews for that board. But it is one more thing to check against spec, because poor voltages will also cause memory errors.

A note about memory: manufacturers do not make modules exclusively for each specific speed. They make a big batch of modules and then test them at various speeds. The modules that start to fail when clocked above 800 go in the 800 bin, the ones that start failing above 1066 go in the 1066 bin, etc. A few modules seem to be able to be clocked infinitely and they go in the 2000 bin.
So often any stick of memory that shows errors at a certain speed can theoretically be "throttled down" to work at a slower speed. So don't be afraid of testing that system at 800 to see if it becomes stable. Then start ramping up from there.

Hope that helps

smarteyeball said:

Memtest should be run with a minimum of 8-12 passes. Tell your friend to run it overnight, as it as exciting as watching paint dry.

Click to expand...

Haha, yeah. I've ran memtest myself already back when I was having problems with my own computer before I discovered it was the motherboard.

TVeblen said:

Thinking about the statement: "ASUS motherboards apparently don't automatically config RAM"

Click to expand...

Well, it was something I read somewhere when I was Googling around back when I put it together. People were saying that some ASUS boards don't use the correct timings for OCZ memory (not all memory in general).

Here's the RAM, by the way:
Amazon.com: OCZ OCZ3P13334GK PC3-1066 DDR3 1333MHz Platinum Series 4 GB Kit: Electronics

If it does turn out to be faulty, I'm not going to mess with trying to get 'more stable' timings. Because that would just be really hard to walk them through. I'll just order them better/more compatible modules and have them return the old ones to me and then I can RMA them (hasn't been all that long since I purchased them).

Alright, time for an update here. I finally got her to uninstall the video drivers. Before she had a chance to restart the computer (but after the uninstall said it completed successfully), the computer crashed and restarted itself. By this point then, the drivers should be uninstalled. After it booted back up, it crashed again shortly after. After the 3rd bootup (while I had her downloading memtest), she got the popup notification from the system tray again saying that the ATI Driver has crashed. By this point, we didn't reinstall the fresh drivers, so there were no ATI Drivers installed (except the generic ones I guess). I'm not sure if this means something much worse.

Anyways, memtest is currently running. It's done about 2 passes so far and 0 errors. I'm really hoping it does error at some point. Replacing the RAM would be easy enough.

EDIT: Also, figured I'd mention that it seems to be getting worse. It's never just crashed right away like it's doing now. And it's even messing up on her in-game, see this screenshot: http://i.imgur.com/jrq2j.png. The weird thing about the screenshot is that the game was actually black for her when she took it. When she pasted it in paint though it was like that.

Just in case too, I'd fall back to Catalyst v10.3, I had to for Future Pinball, 10.4 and later, doesn't render the table graphics.

Always fall back a driver version or two when weird issues occur is my motto.

All current and old drivers and video card utils can be found at guru3d.com

Last update (hopefully). Left memtest run all day and night. It ran 20 passes with 0 errors. Noticed that her resolution was still at 1680x1050 even after uninstalling the drivers, thought that was a bit odd. Got her to get driver sweeper and run it, which knocked her resolution down, and had her watch some YouTube vids before reinstalling. No crashes or errors or anything. Got her to install the latest drivers (since there were newer ones out now anyways) and well so far so good, no problems yet.

Thanks to all the responses here. Seems to definitely have been corrupt drivers. At least, we're still hoping. It's only been maybe 45 minutes since doing it but still no crashes or messed up colors.

Ugh, I hate to do this, but another update... she checked back on the computer after a couple of hours (left it on to download a game on Steam) and the colors were messed up again. Rebooted and went into a game and it crashed shortly after. Did this two more times as well. So, it's not the RAM and not the video drivers (I'm guessing?). Is it now possible that the video card is just dying? Let's see.. I bought it last August and then replaced it with a 5870 in December. Between December and May, it had been doing nothing except sitting around in one of those static bags that they package the video cards in. Then I put it in that computer. I don't know what else to think of.

What psu are you using in the rig?

Fumz said:

What psu are you using in the rig?

Click to expand...

Uhhh.. I'm pretty sure it's this. I know it was an MSI TurboStream and I'm pretty sure 600w.

Newegg.com - MSI TurboStream 600W 600W ATX 12V V2.2/ EPS 12V/ BTX SLI Ready Active PFC Power Supply

I never had any problems with it though. Even used it in my current rig until I swapped it for a 750 corsair.

Do you have access to a power supply tester? I'm not saying the problem lies with the psu, but it's possible. This one has an exceptionally short warranty of only 16 months.

Here's the psu: MSI Turbostream 600 W Power Supply Review | Hardware Secrets, and here's some background on the actual makers of the psu, Solytech, which is just one of their many aka's: Allied AL-D500EXP 500W power supply review Review

If I were you, I'd see if you can a) test the psu and b) if possible, test the machine out with a different (read: better) psu, like the one you're using now.

It could still be your RAM? Just because it passes memtest doesn't necessarily mean it's all good. Have you run any benchmarks/burn in apps/stress tests in Windows? Also, I'm not sure if you mentioned it before, but, having tried several driver versions on this machine, have you run driver sweeper in safe mode to clean out all traces of previously installed drivers?

I personally can't do anything. I'm in Pennsylvania and she's in Texas so I've basically been trying to walk her through everything so far. I'm not sure if she has access to a tester, most likely not. I had replaced that PSU because I was having issues with my own PC and thought that might be the problem. My PC would randomly shut off, but not completely. Anyways, turns out my problem was a known problem with my old motherboard.

Testing the computer out with different hardware (PSU, video card, RAM) would be very ideal but since I can't physically access it, we've been trying to pin it as a software problem. I don't have any spares of either of those components to just send to her to test out either.

I'm still kind of hoping it's just the RAM. I would think colors going screwy and sometimes completely crashing would make the RAM a highly likely suspect, just wish memtest would have failed as to make this easier. Haven't done any other kind of tests like that though.

She ran driver sweeper in safe mode before installing the latest drivers (10.6). She hadn't run it when reinstalling the 10.4 though the first time we tested different drivers.

You said you had to manually set the timings; what did you set them to and what voltage were you giving it?

1.8 volts, 7-7-7-20? OCZ DDR3 PC3-10666 Platinum Edition OCZ Technology

**edit**

Also, did we find out if the memory ever made the qualified vendor list? Some boards just flat out do not like certain kits/timings. This may very well be the case... both products are just fine independently, but together, These timings are a bit high for 1066MHz, which you typically see running 5-5-5-15 or lower... or are they running at 1333MHz?

Yeah, I set them to whatever OCZ listed. I didn't have to change the voltage since it was already correct. You can see the RAM info from this picture she sent me of memtest working: http://i.imgur.com/9nU9W.png

Someone on the first page mentioned that it's not in the qualified list for that motherboard, actually. I figured memtest might fail if that were the case though. As I mentioned earlier in the thread too, I did have problems with this RAM initially. The computer would just lock up and I'd have to turn it off. Did the same thing in my PC. But that was because the bios didn't have the correct timings entered so I guess that's sort of moot.