Computer issues...

[omicron32.livejournal.com]

Something is up with my computer.

Either it's the PSU, CPU or the graphics card. Whatever it is, it is NOT drivers, because I'm having the exact same issues in both Linux and Windows. Whenever my CPU usage jumps up to more than 30%, my monitor clicks and goes blank (like a mode change) for a fraction of a second then comes back on, and sometimes my computer will simply restart.

Just to say again, this is not a driver issue. Problems occur in both Windows and Linux. I've run Memtest on my memory just on the offchance it was that, but my memory is fine.

Now I've been monitoring my voltage coming from my PSU and there doesn't appear to by any major drops there. It could be the CPU, but I don't see how that would be drawing enough power to cut out the graphics card... Shit, it could be my motherboard. It seems to be intermittent, though. Because generally whatever OS I'm booting to will manage to get up and running without any issues, and occasionally I can work with 100% CPU with no trouble, and others it'll just die.

Heating issues.... Well, I've never had this problem before, and I've check that the fans are working without issues. CPU temp never seems to get over 50'C and is generally between 35'C-45'C when idle. Motherboard temp is less and doesn't rise above 40'C for sure. I don't think it's heating because I've had this machine compiling KDE for 16 hours straight (100% CPU usage) without any issues whatsoever quite recently.

I haven't added any new hardware for ages, either.

Specs:

2.53GHz Pentium 4
1024MB RAM
GeForce 4 Ti4600
Primary OS is Gentoo Linux, but I dual boot to WindowsXP too.
One 250G HDD, one 80G HDD.
450W PSU.

Any advice?

X-Posted.

Comments

[torkell]

It does seem rather like a power-related problem. Another possibility is something obscure going on with the BIOS. You might not actually spot it if it's power, as the drop may be too short to appear on whatver you're using to monitor it. Don't forget also that if the POWER_GOOD line drops (one of the ones on that 20-pin monster), then it goes regardless of voltages.

A couple of ideas would be to try pulling out, say, a CD-writer to see if it's power related, or underclock the CPU. If it's a timing issue in the BIOS then you may find it goes away if you *overclock* the CPU slightly.

Good luck - these sort of things are never fun to troubleshoot!

[omicron32.livejournal.com]

Because the CD-writer pulls power off a different rail, will that actually help? Presumably it would just be the 12v rail that was dropping since it seems to affect the motherboard?

Then again, I'm no PSU expert, so I'll try that just in case.

[torkell]

The CD drive pulls from the +5 and +12 rails. Generally, drives pull +5 and +12 (up to a couple amps from each with some ones). The MB gets +/-5, +/-12 and +3.3 to play with, but the stability of those usually depends on the +5 or +12 line.

I'm no PSU expert either, but I know a bit about what they supply and how they work.

[jcaswell.livejournal.com]

When you're in Windows and it restarts, does it write anything to the error log? (It'll probably do it in Linux as well, but my Linux knowledge is limited) That may well help in telling you what the issue is. It certainly sounds power related, but anything's possible.

[torkell]

I think windows will only write useful stuff if it went and had a BSoD.

Another possibility I forgot was some sort of hardware conflict. At my past school the main server was randomly rebooting (as in worked fine for a few days, then went straight to the POST from windows without passing go and without collecting £200). The culprit: apparently the unused but enabled onboard network controller. Disabling that cured it.

[jcaswell.livejournal.com]

Yeah, but if you leave the settings at default in 2K and XP, it'll reboot as soon as it hits a bluescreen, rather than leaving the bluescreen there long enough to read...I was thinking of the reboots rather than just the flickering bit. Forgot to say that though :)

[torkell]

I know that one... it was great fun when I had both that and auto-login enabled. Switch on computer, POST, text-mode, VGA, "Starting Windows 2000", "Logging on", BSoD, reboot, POST, text-mode, VGA, "Starting Windows", "Logging on", BSoD, reboot, POST, ... I can see the resoning behind the suto-reboot, but it's less useful when it's stuck in a loop.

It does however sit on the BSoD *just* long enough for you to go "what the...shit!"

[jcaswell.livejournal.com]

*nods* Just long enough to get the idea something's wrong, but not long enough to work out what...I'm sure they do it deliberately.

I once left a machine at work rebooting as I left for the weekend. Got in on the monday, and it had been doing the bluescreen / reboot thing all weekend. Whoops.

[torkell]

Well, if you set it to do a kernel or full memory dump as opposed to the minidump, it'll actually sit there for about a minute (on the kernel one) dumping a chunk of memory and swapfile to disk. Tired that once when I was having some more unusual problems (which I think boiled down to the eternal disagreement between my GeForce4, Win2k, and nVidia's later (as in 5x.xx) drivers).

*randomly friends you*

[jcaswell.livejournal.com]

That all sounds far too sensible for me :)

*randomly adds you back* :)

[dmsalem00.livejournal.com]

sodding nvidia drivers...one of the reasons i FINALLY stopped using WinME(shut up i explained that to all of you already ;-p).

I'd switch to ATI's cards, if I hadn't been using Nvidia's ever since 3DFX stopped making Voodoo cards.

[torkell]

I might switch at the next graphics card upgrade, but for the moment I rather like my one. GeForce4 Ti4600, dual-DVI, VIVO, golden sample, it's basically about the best one of its generation.

[omicron32.livejournal.com]

I always turn off automatic rebooting on any machine I build (at work and for friends and family) because I'd much rather have diagnostic info right away.

[jcaswell.livejournal.com]

I'll shut up then :)

[methedras.livejournal.com]

My first step would be to pick up a new PSU and try that. Otherwise it could be your mobo being funky.

I've had computers do weird power related shit like that too due to motherboads or PSUs fucking out.

Popped caps?

[coyoteden.livejournal.com]

Look for capacitors (the cylindrical components on the motherboard around the CPU and slots) with bulging tops or even leakage of what looks like battery acid.

You have a newer system, so I doubt you'll find any, but a couple of years ago a ton of motherboards were made with bootleg capacitors that built up gas and popped.

Your problem definately sounds like a power-related issue. I'd agree with trying a different PSU.

Re: Popped caps?

[dmsalem00.livejournal.com]

Oh, god I heard about that capacitor snafu. Saw a few of them. Looked like ittle-bitty soda cans that had been dropped. Kinda funny looking if you ask me.

[ozdeadmeat.livejournal.com]

Lock me in for RAM or RAM socket being the problem.

[omicron32.livejournal.com]

I ran memtest already.

hmmmm.....

[jecook]

I'm to call it either a PSU issue, or a Mainboard issue, or a GPU issue.

I've seen GPU (aka video cards) cause spurious freezes and reboots. Start there, especially with the spurious "shuts off video and comes back or resets" issue acting as the major pointer here. Try swapping it for a known good card.

There was discussion earlier about BSODs. Typically, if it's a hard ware re-set issue there won't be an entry in the system log, only one that says that the previous shutdown was unexpected. Linux will usually do an fsck after this, which will be reflected in the syslog after it comes back up. rarely, it won't come back easily from a fsck, which means either a single user mode startup or other entertaining issues. Usually, unless you pulled down the debugger for dump analysis, it's pretty hard to diagnose a BSOD issue based on the stop error alone some times (the descriptions can get quite vague, and at one point I had to resort to using the dump analysis tools to tell me that it was a peice of software behaving badly))

I don't think it's a CPU issue, because according to what you've stated, a 16 hour compile session runs ok.

I'd borrow (or get) a thrmometer, and start checking the temp of most of the logic chips on the video card and mainboard, and start from there. for the GPU, I'd also look at the support logic. Sometimes if it goes flakey, it can cause problems.

Re: hmmmm.....

[torkell]

There was discussion earlier about BSODs. Typically, if it's a hard ware re-set issue there won't be an entry in the system log, only one that says that the previous shutdown was unexpected.

Under windows you don't even get that - all you get is a "eventlog started" without a "eventlog stopped" just before it. (A BSoD will log a "Save dump" entry with the details of the bugcheck). And if you're running NTFS, then you don't even get the obligatory whining about lost clusters and crosslinked files. AFAIK, you only get any sense out of the BSoD if you save at least a kernel dump, or have a debugger talking to the computer (and then you probably need to run a 'checked' version or something fancy). Then again, you actually get something, which is more than you get with the hardware deciding to go back to square one.

Re: hmmmm.....

[jecook]

My mistake, my info was from NT4.0, on the "unexpected" shurtdown. (the entry would read like "the last shutdown on [DATE/TIME] was unexpected") 2000 does not have this. XP does have the Save dump turned on (it's turned off by default in 2000, but can be turned on), and occasionally NTFS WILL do a filesystem check, if the "clean" shutdown flag was not marked.

The only time you can make any real sense out of a BSOD is to run the debugging tools on the dump file. There are checked versions and retail versions of the symbols availible on the Microsoft web site (you *will* have to dig around for them a bit to find them, unfortuantely, one has to wonder why they don't want you looking at why it crashed? ;) The debugging tools are also there as well)

I know of very few techs that would even bother with the serial console and debugging. One would assume that it would be for serious development work (and I'm not sure about that even).

Between the BSOD and the debugger tools, I can usually figure out what program at least caused the show stopper, and take actions from there. That's how I ultimately found out that the Jetsuite software for my LaserJet 3100 does not like XP, even though HP says it ought to work. It also provided me with a clue that the A/V software on a few systems at work caused STOP errors.

As far as the reset issue goes, That's almost always caused by hardware faults. Very seldom is it software caused.

Re: hmmmm.....

[torkell]

(you *will* have to dig around for them a bit to find them,...
You *will* also curse when you realise that the files run into 3-digit numbers of megabytes and you're on dialup.

I imagine you'd use a serial console when developing a device driver, especially one that loads early on in the boot process (say a SCSI driver). Otherwise you can't really step thru it or see what messages it throws up.

BTW, Win2k3 might have the shutdown message back, as it certainly asks you just why you wanted to shut the computer down in the first place.

Re: hmmmm.....

[jecook]

Point. I consider that 'Deep Magic', and I have a tendancy to run away screaming if I'm asked to work on a system like that. :)

BTW, Win2k3 might have the shutdown message back, as it certainly asks you just why you wanted to shut the computer down in the first place.

"But I don't *want* to shut down! I'm a server, I can't shut do-" BEEP!

I find humorous irony (especially when I'm dealing with a machine that is behaving _really badly_) that the shut down option is located in the Start menu.

And I'm sure that server '03 had it's teething problems before it went gold. I know ours has been performing like a champ, but then, we don't have much running on it. Come to think about it, all our servers have been performing pretty decently, except the mail server... hmm...

Re: hmmmm.....

[dmsalem00.livejournal.com]

You guys have a point. 2k3 *really* doesn't like to shut down. My dad runs it in his law office, and he hasn't(or had to) reboot his servers in nigh on a year.

Then again, my dad's even more of a maniacal computer genius than I am, and I'm the one running modern games on a 733 celeron with pc100 ram and a PCI video card. Doom 3 anyone??? >=D

[taiirei.livejournal.com]

I have the same problem as you except it's NOT related to CPU usage! Occasionally, the screen will turn off by itself, and either come back on or restart. I think it was my sub $100 video card... but I'm not sure. Anyway, I hope you get this fixed...

[taleya.livejournal.com]

I had the same problem with my Radeon VE7000. I sat there and poked it and poked it, then eventually hit it with a stick. Knew it wasn't the card, i'd tested it in another machine. Knew it wasn't anything ELSE, because it only had problems with the damned card. Drove me batshit because it just started out of the blue.

I ended up doing a BIOS flash. Fixed it up wonderfully :)

[taleya.livejournal.com]

Was just reminded by a friend - he had something similar when installing a new vid card....the culprit, believe it or not, was the USB drivers for the mainboard. They were fritzing the AGP