Atomic I/O letters column #39Originally published in Atomic: Maximum Power Computing Reprinted here November 2004. Last modified 16-Jan-2015.
When I was trying to find out why an old swap-meet Socket 370 machine wouldn't power down when I told Windows to shut down, I found it was because ACPI wasn't enabled.
I stumbled over how to enable ACPI when initially installing Windows XP here, but I have yet to try that.
In that same thread, I discovered that ACPI in Windows 2000 and XP slows your system down. It was commented that home systems should have all forms of ACPI disabled for optimal performance. The writers rave about stability and smooth performance increases without ACPI.
Is there any truth to this? Is the performance improvement significant enough to warrant me disabling ACPI on my Athlon XP 1800+ system?
First up - you don't have to do a from-nothing reinstall to enable (or disable) ACPI on a Win2000 or XP system.
You can do that if it thrills you, but you can also disable ACPI in the BIOS setup program and then reinstall Windows over the top of itself (a "repair" install). That won't wipe all the stuff you've set up in the non-ACPI install.
The thread you link to is not exactly a mine of accurate information. Disabling ACPI certainly doesn't make more IRQs available, for instance; on the subject of what actually happens, you may find the last couple of letters in this column interesting.
ACPI also has a trivial impact on system performance, unless something's wrong.
The usual way in which something goes wrong is that ACPI interrupts do start taking up a big chunk of your CPU time, because far more of them are being done than is necessary (as they were on Paulo's machine in this column). A BIOS update can fix this problem on some motherboards; on others there's nothing for it but to disable ACPI.
But if the Processes tab of Task Manager doesn't show something like a pair of "Time Placeholder" tasks taking up a big block of CPU time, you don't have this problem.
I was looking at buying a new case. I know that most aluminium cases will be lighter than steel ones, but is it true that aluminium cases will be cooler than a steel? If there were two identical cases with the same components/cooling, except one was aluminium and the other steel, would the aluminium one be cooler?
Yes, the aluminium case would be cooler, but not by enough to matter. Aluminium is a much better thermal conductor than steel, which can help keep hard drives a bit cooler (because they often don't get a lot of airflow) and will allow a small amount more heat to be disposed of to the outside air. But compared with the cooling from the ventilation fans, case conductivity is unimportant. A steel case with the same number of fans and vents as an aluminium one will be, for all practical purposes, just as well cooled.
How can I stop Windows from switching focus to new windows? Like whenever an ICQ message is sent it switches to that window, even if I'm typing in another. Or when something pops up. I don't know how I've avoided accidentally installing GAIN.
You want the General -> Focus options. They're also helpful for preventing great big Photoshop lasso operations being terminated half-way through.
I have an idea about how to solve a problem all LANners (well, probably all of them) get, but I don't know how to implement it.
When I play a network game, and somebody else leeches off me (er, just all my really cool open source music, of course), terrible lag occurs as a result.
Instead of just disabling file sharing on that network card, could it be fixed if I used two network adapters, and bound one to gaming and one to file sharing?
You're talking about load balancing via link aggregation and/or trunking, which won't work right without a special switch that understands that single nodes on the network can receive data at more than one MAC address. Typical cheap LAN party switches are not trunking-aware - and even if you're on a big old corporate network, trunking probably still won't work, if people have been hooking up to it in the usual ad hoc LAN party stylee, when the LAN is officially declared complete as soon as everyone can ping everyone else.
Gigabit Ethernet can solve this problem, provided you've got a gigabit connection to a switch (possible, but not very likely, at the average LAN party today) and there's only a regular hundred megabit (Fast Ethernet, 100BaseT) connection to the leeching machine (or the machine you're leeching from). As long as that other machine isn't running the game server you're playing on as well, and as long as nobody on that other machine is trying to play a game, nobody will be discommoded if you suck data down the Fast Ethernet pipe into your wider gigabit pipe as quickly as you can. Your pipe to the switch won't be saturated unless you decide to leech from a few people at once.
Actually, most PCs with gigabit Ethernet can't saturate a gigabit pipe, because the plain PCI bus most gigabit adapters are plugged into can't shift as much data per second as the gigabit link can. Disk speed can be a limiting factor as well; this can be a good thing for your problem, if it turns out that disk speed at both ends of the leeching connection bottlenecks the transfer enough that there's spare bandwidth for your games.
A better solution, if you can get people to use it, is Nullsoft Copy, from the recently moribund (or not) makers of Winamp. NSCopy is a directory-leeching program (it doesn't do individual files) that can be throttled to the speed of the user's choice. NSCopy lets you prevent your own leeching from stuffing up your gaming experience, but if everybody doesn't use it (or if they wind it up to its maximum Spaceballs gone-to-plaid speed) then you can still end up with links being saturated here and there.
In case you're wondering, the Quality of Service packet scheduler in NT-descended Windows versions won't help you much, unless your games are QoS-aware (which I doubt). QoS reserves a certain amount of connection bandwidth (by default 20%) that'll always be available to, and shared between, QoS-aware programs, if any are running. Since I don't think games or Windows file transfers will know anything about QoS, though, and since lots of WinXP users, definitely including many tweak-obsessed LAN-partyers, disable the standard QoS Scheduler, it's a waste of time. All it does is make the pipe your activities can use thinner.
Why is it that CRT screens fuse dust onto the wall? I recently purchased a 21 incher and after a month of PlayStation, it seems I have textured wallpaper.
CRTs produce a lot of positive ions (so do various other high voltage electric devices, like laser printers and photocopiers, for instance). Dust near highly charged things gets charged itself (usually by contact with the faint ion breeze from the device, not with the device itself), and is then prone to stick to anything that's got a relatively opposite charge. This includes the walls near the CRT, which are neutral, but negative compared with the positively charged dust.
The way this happens causes dust to cake up and look as if it's been "ground in", producing a dirtier look than you'd get from ordinary dust accumulation - especially on vertical surfaces, which of course normally have next to no dust on them at all.
The plastic casing of a CRT monitor (or TV, but they're not usually beige and so don't show the dirt as much) will also accumulate crud this way. A CRT that's been turned on seldom or never will, all other things being equal, be much cleaner than one that's been running 24/7.
The dirt isn't really ground in, though; it should be fairly easy to remove from painted and plastic surfaces with a damp cloth.
I was wondering if a box exists so that I can plug in a legacy PCI card (a special TV-out card), and then connect it to my laptop via USB. I've have a look on the Net, but I can only find PCI cards that provide additional USB ports, not the other way around.
Sorry, but there's no such thing. Even USB 2.0 doesn't have enough bandwidth to do the job - 480 megabits per second for USB 2, versus something like 1056 megabits per second for PCI. That would make the creation of such a device very difficult.
I've got a motherboard that I would like to stick in a one unit rack chassis. The problem is that it has a vertical audio block on it, which won't fit. Would cutting this vertical audio block off of this (or any?) motherboard add any instability to the system?
No, it should be fine, provided you don't damage the board in the process. You'd probably be fine even if you ended up shorting the audio traces on the board.
If you want to do this properly, though, you should desolder the thing. If it's some invisible-glue surface mount jobbie then that might be a problem, but connector blocks are all still traditionally mounted with (relatively) fat pins through holes in the board, to my knowledge, so you shouldn't need anything special to remove them. Melt solder with medium-heat iron, remove solder with desoldering braid or a solder sucker, and you should be able to lift the connectors right off.