Random, frequent BSoD's with no minidump or stop code

[aib]

Hi all,

I have a brand new computer with about 10 programs installed. I'm having random crashes and BSoD's. There is what seems to be a pattern forming:

I do a brand new SP3-bundled installation of Windows XP 32-bit. I get all the latest drivers for my hardware.
I play Team Fortress 2.
The game crashes 2 or 3 times per day, sometimes it's just memory protection ("illegal instruction," etc.) and sometimes it's a BSoD.
Sometimes I get a proper BSoD with a stop code and a memory minidump. The faulting driver is almost always ks.sys (Windows Audio something), rarely nv4_disp.sys (GFX) and rarely others.
But here is the weird part: Sometimes I get a BSoD without a stop code. One without a minidump. At the bottom, it says "Beginning dump of physical memory" but it just gets stuck.
I just assume it's a problem with the game and hope they fix it.
After about a month, my desktop crashes.
It crashes when I'm playing a video with VLC media player and no other programs running; it crashes when I leave uTorrent to download at night (with only an IM program running in the background), my computer welcomes me with a big blue screen in the morning. No minidump is ever generated here.
Oh, and when this happens:
I reboot. Fire firefox up and start looking through troubleshooting forums. Start WinDbg up to look for the minidump that isn't there. As I'm writing a post or e-mail not unlike this one, the computer crashes again. No stop code. No minidump.

I'm now in the second incarnation. The whole thing I wrote above has happened to me before, on my previous Windows installation. It was a lot dirtier in terms of game and software installed, but what happened was more or less exactly this.

I'm convinced this is a hardware problem, but I'm having a hard time pinpointing the source. I installed SpeedFan as per the guide's suggestion, but I'm not sure what I'm looking for.

Any ideas?

CPU: Intel(R) Core(TM)2 Duo CPU E7400 @ 2.80GHz
MB: Gigabyte EP45-DS3L
RAM:

Decoding DIMM #0
Memory type is DDR2 / Module Ranks : 2 / Levels : 1.8V / Parity : NO PARITY / Refresh Rate : 7.8us / Total Size : 2048MB
Decoding DIMM #2
Memory type is DDR2 / Module Ranks : 2 / Levels : 1.8V / Parity : NO PARITY / Refresh Rate : 7.8us / Total Size : 2048MB

PSU: Cooler Master RS-460-PCAP-A3 (460W)
GPU: Axle GeForce 9600 something
Other: 1 SATA, 2 IDE hard drives, 1 SATA DVD-RW

I've just noticed a sticker on my PSU which says "combined output should not exceed 401.5W" - could this be the problem?

SpeedFan reads "0.26V" for the +12V voltage rail.

I ran Windows Memory Diagnostics (boot CD) and Memtest86+ without any errors. (Memtest86+ once locked up right after showing 1 error.)

 

[Route44]

If you have errors on Memtest you have memory corruption. Random crashes are a good sign that this is so. Run Memtest again per stick with a minimum of 7 Passes per stick. If you start receiving errors before 7 Passes you know that stick must be replaced.

 

[aib]

If you have errors on Memtest you have memory corruption. Random crashes are a good sign that this is so. Run Memtest again per stick with a minimum of 7 Passes per stick. If you start receiving errors before 7 Passes you know that stick must be replaced.

I have run Memtest for long times, without any errors.

One of my ram sticks used to have errors; when I first bought the computer, it took either WMD or Memtest some 30 minutes to detect the errors. I used to get much more frequent crashes and BSoD's back then. After a BIOS update suggested by the vendor, the bus speed increased, and I got millions of errors on one stick.

I've since replaced the faulty stick, though. And for all my running memory diagnostics since then (~2 hours but I will do more) I've only got _1_ error. That is, a single bit. One bit of error in whole of 4 GB.

And the weird thing is Memtest crashing after reporting that error. Memtest did not crash when I had a faulty stick which gave me millions of errors per minute, which could not run any program for more than, say, a couple of hours without causing a crash.

If I had to guess, I'd say that the problem that made Memtest crash is the same that makes Windows crash. That is, a hardware problem that manifests itself only once a couple of hours.

I will leave my computer to test the RAM before going to bed tonight. But I'm still open to other methods of diagnostics.

And here's another question: Is there a way to test graphics RAM?

 

[aib]

I left my computer to test the RAM today.

Apparently, on about the 15th minute, it found one bit of error in each RAM and then got stuck.

When I took one of the sticks out after 5 hours, it was very hot.

At least I have a replicable problem now.

 

[Route44]

Again, if you get errors you have corrupted memory. It can only be replaced, not fixed.