Why does my new computer crash?

June 12, 2006 7:48 AM

Why does my new computer crash?

I recently built a new system (my second build):

Athlon 64 X2 3800+ (Artic silver, factory HS/fan)
Abit A8N-SLI
2GB Corsair XMS DDR400
XFX Geforce 7600GT
SB Audigy 2 Platinum

All clocks are factory except video, which is auto-overclocked using Coolbits. Newest AMD drivers are installed, as well as the XP dual core fix.

The system smokes and I'm having fun with it, but I keep getting BSODs when the processor is under load, and it never gets above 40 degrees C. System temps and video temps are also in normal range. This happens when playing Oblivion, HL2 (which usually crashes to the desktop but occasionallyl BSODs), and once while burning a DVD. Last night I attempted to encode video using Divx 6.2 (multi-threaded) and got BSOD both times.

I am fairly experienced, but lost when it comes to troubleshooting this. Where should I begin?

posted by bradn to computers & internet (18 comments total)

your video card is overclocked? did you try not overclocking it? the only thing(s) that you listed there that don't really use the video card is burning a DVD. (and possibly the DivX encode - not sure what resources it uses.)
posted by mrg at 8:04 AM on June 12, 2006

It could be your RAM, try removing one of the sticks, then test each one individually. If the computer is stable with both of the sticks being used individually, then you likely have a faulty RAM socket.
posted by oddman at 8:10 AM on June 12, 2006

The temps on the vid card never get over 50C, so I assumed that isn't the problem. Also there are no artifacts or video glitches, it's just straight to blue and then reboot.
posted by bradn at 8:12 AM on June 12, 2006

[40,000th question]
posted by cloeburner at 8:18 AM on June 12, 2006

[I searched and couldn't find an answer to my question. So I asked.]
posted by bradn at 8:31 AM on June 12, 2006

it's almost certainly bad RAM.

You probably want this: http://www.memtest86.com/
posted by jaded at 8:32 AM on June 12, 2006

No, I was pointing out that this was the 40,000th question asked on ask.me, it was not a criticism of your post.

Please ignore my derail.
posted by cloeburner at 8:37 AM on June 12, 2006

[Sorry cloeburner. too much coffee.]
posted by bradn at 8:40 AM on June 12, 2006

Thanks jaded, i'll do so when i get home. I had meant to run Memtest when I first built the machine a month or so ago, but never got around to it. I had heard of others having bad sticks of this RAM. I hope I can still RMA it.
posted by bradn at 8:42 AM on June 12, 2006

second memtest86, but this is a more recent version (fork), as far as I understand.

some general things that come to mind:

• if possible, try swapping ram (and test via memtest)
• stress test individual components - ram via memtest, cpu via things like cpu burn
• check for bad capacitors, they have been the cause of random "mistery" BSODs in the past

Let us (and Google) know if you've found the solution! Good luck!
posted by lodev at 8:43 AM on June 12, 2006

Try memtest first, but do physically swap the RAM around too even if memtest doesn't seem to turn up anything. My mom's s computer had a similar problem (beginning after over a year of running perfectly) which I ultimately traced to the 2nd DIMM socket on the motherboard itself.

Memtest (pretty much the first thing I tried) had run fine for 12+ hours without errors. I ended up spending a lot of time exploring other areas fruitlessly afterwards before I started swapping the memory around and then using known good sticks from another computer.
posted by Pryde at 9:24 AM on June 12, 2006

If it actually blue screens, then you should have error codes that might be useful in diagnosing the problem.

Video card temp isn't necessarily going to indicate video card stability.

That said, I had a similar setup that had similar sounding problems and the blue screen codes weren't much help. Memtest also checked out just fine.

It took me a while to find something that would crash it reliably, but on a hunch, I tracked down a 1024p WMV sample from microsoft, figuring that it would stress the CPU, memory and PCIe bus in relatively similar measures. As luck would have it, I was right. I could cause a crash within 15s or so of starting playback of the clip.

Once I had a quick and reliable way to crash the machine it became much easier to track down the problem.

First I found that running with a single DIMM was much more stable, but it didn't matter which of the two DIMMs it was.

I was using the MB detected memory settings, but found that it would run stably if I switched to more conservative settings. I ended up downloading a new BIOS which allowed it to autodetect the right memory timings.

Along the way I also realized that I'd misunderstood the motherboard manual and I actually had both my DIMMs in a single memory channel, with the other channel totally empty, which isn't optimal from a performance point of view.

Bottom line, time to get methodical and do a process of elimination. That means starting with obvious possibilities, like setting your vid card back to stock settings. Sounds like the DIVX encode you tried to do might be a good stability test to use along the way.
posted by Good Brain at 9:40 AM on June 12, 2006

Thanks for the help, I'll unoverclock the video card, and start memtest on lunch. Incidentally, I have the WMV HD trailer for A Scanner Darkly on the machine, and have watched it several times without a crash.
I did have to manually set the timings for the RAM, as they were "recognized" incorrectly. According to the NewEgg comments for the RAM, it's a common problem. I will double check to make sure the timings are correct as well.
posted by bradn at 9:48 AM on June 12, 2006

If you have the dump file (or even just a minidump) then you can try loading it in windbg and getting a backtrace.

General STOP message interpreting.

But I caution that if it is indeed bad RAM (likely) that this will manifest as random bit errors, which means you might not be able to trust what the STOP message says. For example, if a random bit error caused some essential kernel data structure to become corrupted, then whatever driver happens to read it next might cause the crash, but that doesn't mean it's actually that driver's fault.
posted by Rhomboid at 1:36 PM on June 12, 2006

You didn't specify what power supply you purchased for use with this system. You went all out on the other components, but did you skimp on your power?

When you run the system under load, you run the PSU under load. If it's not a quality PS, then voltage may sag or start waving, and that can cause instability as well.

Keep this in mind as you test your memory with memtest86. Also, let that go overnight, not just a few hours- errors may not show until later passes.
posted by id at 2:27 PM on June 12, 2006

Thanks id. I forgot to list the power supply, it's an Ultra X-finity 500W. Not the best but certainly not a cheapo.
posted by bradn at 2:46 PM on June 12, 2006

I have no answers for you, but congrats on being the 40,000th AskMefi poster!
posted by limeonaire at 4:40 PM on June 12, 2006

Or rather, posting the 40,000th question. You get what I mean, though.
posted by limeonaire at 4:41 PM on June 12, 2006