ratbastid

In the last couple days, my computer has learned a great new trick: rebooting itself. It's doing it every hour or so.

This machine is a 1.33 ghz Athlon with 1.128 gb ram running XP Home SP 2. I recently tried unsuccessfully to install a USB 2 PCI card, but the card wasn't recognized at all by the system, and I suspect the card itself is faulty--the PCI slot worked great with the device I pulled to make room for the new card.

I also just added the 1gb of ram. Prior to this incarnation, it ran Mandrake 9 and had 128mb of ram. Mostly it was a file server, and it was creaky, but it ran. That was a total no-go with XP, obviously, so I filled the two empty RAM slots with 512mb apiece.

That's the past. Here's the present: every time I leave the machine alone for a little while, it reboots itself. When it comes back, it wants to report to MS about having recovered from a serious error. When I let it do that, it brings up this page: http://oca.microsoft.com/en/response...8f97857&SID=11 about an error in the display driver. The specific error appears to be "Error Message: STOP 0x000000EA THREAD_STUCK_IN_DEVICE_DRIVER (Q293078)".

I have a GeForce2 MX AGP card in this thing. When I installed XP, Windows recognized it and installed Microsoft's driver. After I started getting these messages, I installed Nvidia's driver for this card, which went in just fine. The problem continued.

I then went Display Properties->Settings->Advanced->Troubleshooting and turned hardware accell right off. That helped the problem some--it started going longer between magic reboots, but not a lot longer.

Before I did that, it was rebooting itself every time I tried to install Palm Desktop, which I needed on this machine to sync my Treo. Tried it like five times with different settings on the display. It always crapped out right in the middle of the install until I turned off hardware accelleration.

Any suggestions about how to make the machine behave? Is my video card boned? Does XP just generally blow?

MikeSty

If anything, I'd have to say bad RAM. Hook up an FDD and run Memtestx86.

cyrnel

I'm suspicious of the video card. The kernel driver error is thrown because it's become divorced from the card for some reason. If the card has a fan, is it working?

Did the USB 2 card incident immediately precede the problem? Unplug power and check to make sure the video card didn't rock out of the AGP socket slightly. Intermittent contact on a few pins could send everybody to couseling.

Or just try another card or onboard video if you have it.

__________________
There are a vast number of people who are uninformed and heavily propagandized, but fundamentally decent. The propaganda that inundates them is effective when unchallenged, but much of it goes only skin deep. If they can be brought to raise questions and apply their decent instincts and basic intelligence, many people quickly escape the confines of the doctrinal system and are willing to do something to help others who are really suffering and oppressed." -Manufacturing Consent: Noam Chomsky and the Media, p. 195

ratbastid

I'm back from this weekend's travels and running memtest-86. So far it's found one error, at address 148.8 mb, which would be into one of the new 512 sticks. But it's only 21% done. I'll likely need help interpreting its results...

To get the thing burned, I pulled one of the memory sticks at random--figuring if one was bad, I'd have a 50-50 shot at guessing the right one--then downloaded and burned memtest86. Thing seemed pretty solid while I was doing that. And I suspect that the error I just saw was in the stick I pulled--a 512 stick from the middle socket.

Still, though, Windows is clearly reporting an error about the video driver. I just pulled and re-seated the vid card. It seemed pretty firmly seated before, and I'm quite confident of it now. No on-board video on this MB, sadly.

cyrnel

Sure could just be the RAM. I worry something else failed when the same error keeps coming up tied to a device. That can also happen if RAM fails in areas mapped to IO but since those ranges are the minority you have to get "lucky".

Are the new DIMMs identical? Did you change any timings?

There's always a chance you chose memory that didn't pass muster with the MB manufacturer. It's worth checking their approved memory list. They can't approve everything that works but if you have something from their bad list it's a good indication.

Let us know how it goes.

__________________
There are a vast number of people who are uninformed and heavily propagandized, but fundamentally decent. The propaganda that inundates them is effective when unchallenged, but much of it goes only skin deep. If they can be brought to raise questions and apply their decent instincts and basic intelligence, many people quickly escape the confines of the doctrinal system and are willing to do something to help others who are really suffering and oppressed." -Manufacturing Consent: Noam Chomsky and the Media, p. 195

Pragma

Probably a silly question, but have you tried swapping out the video card for another and see if that fixes the problem? It's always possible that the card itself has serious issues.

That said, let us know how the memtest run goes.

__________________
Eat antimatter, Posleen-boy!

ratbastid

Don't have another vid card, unfortunately. All my others are in use.

Here's what's interesting. My ram was (after the upgrade): a 128, a 512, and a 512, in sockets 1, 2, and 3.

I sort of guessed, given how fast the machine was rebooting itself, that it was getting into Socket 2 before finding the error, but not Socket 3 (taking a wild flyer that RAM is addressed numerically up the sockets). I yanked the 512 DIMM in Socket 2. The machine was fairly stable as I downloaded memtest86 and a freeware cd burning package (this is a relatively new install, which didn't have any such utility installed yet). I burned memtest86, rebooted into it, and watched as it tested my 640mb of memory.

Halfway through one pass it was testing clean, but I wanted a control test, so I powered down, reinstalled the 512 in Socket 2, and booted back into memtest86.

Bloodbath. By the time it was done with one pass there were HUNDREDS of errors between 128.1mb and 630.0mb. Hundreds.

I yanked that middle DIMM again and ran memtest86 again (leaving home for a halloween party, leaving it running for several hours). When I came home it was on pass number 6, with zero errors.

Tomorrow morning, that middle 512 DIMM is going back for a refund. 640 should be an ample plenty to run this machine with--in fact, its performance is quite good right as I type on it. I'll look and see if we crash again, but I suspect I've at least narrowed down on my problem.

Pragma

Yea - sounds like you got quite a lemon of a RAM chip on that one. Hopefully replacing that chip for a better one will fix the problem for you. And yes, you're correct about the physical addressing of memory based on which RAM chip is located where (as you already saw from memtest).

I run my machine on 768MB of RAM (had a 256MB stick do nasty things to my computer - the RAM corruption was so bad that Windows told me that my BIOS was not ACPI compatible, removed the RAM stick and suddenly "my BIOS was cured."). 640 should be plenty for your purposes, especially since you won't be doing a lot of heavy gaming on a GF2 and that's probably one of the biggest RAM hogs on a system.

__________________
Eat antimatter, Posleen-boy!

ratbastid

Guess what--much like my baby nephew, the machine stayed up all night. I've got uptime of 12 hours 57 minutes right now. That's like 60 times longer than it was doing before!

liquidlight

It's amazing what happens to video driver data when it's being corrupted in RAM?

With the errors that you found in Memtest regardless of what that error logs are reporting the video card is the symptom, not the problem!

Aren't computers fun?

__________________
Halfway to hell and picking up speed.