Atomic I/O letters column #50Originally published in Atomic: Maximum Power Computing Reprinted here October 2005.
Last modified 03-Dec-2011.
I've had my P4 for about 14 months and until about 8 months ago, it was running sweet. Now all I get is a whole lot of spontaneous restarts when I play intensive games like UT2004, Doom 3, NFS Underground 2 and so on. I have tried 3 different power supplies, changing the RAM from dual channel to single, disabling hyperthreading and recently formatting and reinstalling XP. I have spoken to numerous PC technicians both from TAFE and various computer stores and no-one can give me a straight answer.
In years past, you would have been talking about your computer mysteriously hanging, rather than rebooting. The auto-restart is what XP does by default when, if it were an earlier Windows flavour, it would have bluescreened.
(This is generally a good thing; if nothing else, it means flaky Windows servers restart themselves instead of making some admin poke the reset button when they finally notice the problem. On most desktop systems the auto-restart just saves time, since bluescreen data is seldom helpful, even if you're one of the 1337 few who can understand it.)
Repeated Blue Screens Of Death or BSOD-replacing restarts can indicate hardware or software problems, but in your case it's probably a hardware issue.
Heat's the first possible culprit, of course; dried-up thermal paste between the CPU and its heat sink, sticky-bearinged fans, carpets of dust and so on can cause the symptoms you report. Try my usual testing method - take the case off and point a desk fan into the works, and see if the machine starts behaving itself. If so, start hunting heat.
If it's not heat, then it could be static damage to the motherboard or video card (probably not any other component). Static zaps much too small for you to feel are enough to damage chips, and such damage certainly can make the chip flaky, rather than kill it outright. If you leave the computer plugged in and obsessively touch chassis metalwork all the time while working on it, or use an anti-static wrist strap, then you'll probably avoid this problem - but most people (including some people who work in computer shops...) just roll the dice, and usually get away with it.
Anyway, try to get hold of a loaner video card from somewhere, swap it in (taking proper anti-static precautions!) and see if the problem goes away. If so, then it's RIP your old video card. If not, then it's depressingly likely to be a whole-new-motherboard situation.
Why is it that when I think games I think Nvidia or ATI, but no one mentions 3Dlabs video cards? I would like to know how one of 3Dlabs' cards stands up against Nvidia and ATI for games, and I also wonder why they don't try to break into the gaming industry?
How do they stack up?
Very, very badly indeed.
3DLabs' single-core AGP Wildcat Realizm 200 manages less than 10% of GeForce 6800 Ultra performance for high-res Doom 3, though it approaches 20% of the 6800's speed for high-res HL2. The twin-core PCIe Realizm 800 is better, but not even twice as good - and it still costs $US1700-plus.
For pro 3D rendering, the 3DLabs cards generally streak past all consumer boards of similar vintage, and often beat Nvidia's more expensive Quadro cards handily as well; the Realizm 800 is a monster for 3ds Max and various more esoteric software. But not games.
Actually, it's only been relatively recently that you could count on a 3DLabs card to run most games at all.
More info here.
I know you can clean regular keyboards with water and mild detergent (I do it to mine all the time), but I was wondering if you know if you can do the same to laptop keyboards?
I've just had a guy drop a laptop off that has had beer spilled into the keyboard tray, but luckily it doesn't look like it got anywhere else. But the keys are really sticky now.
As far as I know, you should be fine cleaning a laptop 'board this way. But doing it while the 'board is still part of the laptop is, for the benefit of our slower readers, a Bad Idea.
Stripping the keyboard out of a laptop is seldom easy, but it's usually only a medium-difficulty job, especially if the manufacturer provides a handy-dandy dismantle-and-reassemble guide - IBM are good that way, and Dell have developed a disturbing recent habit of providing acceptable service docs on the Web as well, but I don't know about other manufacturers. You probably know better than me, if you're doing this sort of thing for a job.
The keyboard's a surface component, so you shouldn't have to shift too many Swiss-watch assemblies before you can free its ribbon cable and lift it out, but there is of course ample room for disaster for operators who're less than perfectly coordinated.
Once it's out, I suggest you swamp the thing with clean water (tapwater is fine, a cheap spray bottle may help, de-ionised water can be used if you're paranoid), then leave it in a warm, well-ventilated place (out in the sun will do) to dry, for considerably longer than you think it'll need. Stripping the keyboard completely will help it dry if it's the usual kind with a layered contact assembly - top contacts, holey separator sheet, bottom contacts.
You can speed the drying process if you spritz the wet keyboard with denatured alcohol (diluting and displacing the water with faster-evaporating ethanol). The alcohol shouldn't eat any components or coatings, and it's dirt cheap.
(Usual disclaimers apply if this advice turns out to cause Abrupt Keyboard Death Syndrome.)
I'm a silent PC enthusiast and have managed to passively cool every component in my system except one, the power supply.
I've looked at fanless power supplies (like the Antec Phantom) but from the research I've done they all have failure rates of at least 50% after six (or so) months of use.
Several companies make interface boards and external power bricks to supply power to Mini-ITX boards, and I just happen to have a whole whack of Toshiba 15V@5A power bricks at home (you know where this is going). Is there any way to hook them up together to provide the 250-300 watts I'm looking for my desktop system?
Soon, TARDIS technology will allow PSUs to contain fans that're bigger than the casing.
I'm not sure about the silent PSU failure rate you mention, but it wouldn't astonish me if at least some fanless PSUs had this problem. Mind you, some people no doubt install them in cases with very little ventilation, when they're designed to at least get a reasonable convection breeze.
My first suggestion would be to check out "silent" PSUs with big 120mm temperature-controlled fans, of which there are quite a few on the market these days. They're not really silent, of course, but when the fans are at minimum speed they really are very quiet (and when the fan's running faster, there's probably a good reason).
My next suggestion would be to use a fanless PSU, but make sure it gets good ventilation - install it outside the case, if necessary, propped up on feet that let plenty of air circulate around it.
And then there's water cooling, of course, but water cooled PSUs are something of an advanced project, for reasons you can probably figure out for yourself. And water rigs still need a fan on the radiator, unless the radiator's quite enormous.
Your pile of Toshiba bricks will, I think, have to wait for another project. You could probably run them in parallel and rig some awful arrangement of regulators to give you normal PSU rails, but you'd waste an awesome amount of power as heat.
All such homebuilt systems are, unless you're a pretty expert electronics hacker, going to be some kind of linear PSU, whose efficiency will be lousy and which will need, for the most elegant version, some monstrous transformers to deliver the current a PC needs. You could make one that didn't need a fan, but the transformer hum might drive you nuts anyway.
Stick with pre-built solutions, to preserve your sanity.
I need to build an office computer really cheap, and have settled on an Asus Terminator C3 bare-bones system as the base for it. But the Terminator only has an 800MHz C3, which is gonna be a big fat bottleneck.
I have a Socket 370 PIII 933MHz that's still in a half-cannibalised Dell from four years ago. I could theoretically put that in the Asus box.
Do you think there's a decent chance of the motherboard picking up the PIII without any hitches?
I don't think anyone's ever actually likely to get to the point of being able to answer that question, because I don't think the Terminator has a CPU socket.
Integrated-chip boards like that one, and like Via's own M-series Mini-ITX boards, typically have the CPU soldered in place, because it's cheaper that way (and because they may use CPUs that don't even come in a standard socketed form).