Atomic I/O letters column #40Originally published in Atomic: Maximum Power Computing Reprinted here December 2004. Last modified 16-Jan-2015.
I was looking over the Atomic issue where you had the review on the 2004 GT Edition 3D Cooler-Ultra, and I began to wonder - if all of the top end heat sinks are made from copper, because copper is very good at dispersing heat, why don't Antec and Lian Li begin making copper cases? They would disperse heat extremely well and ensure that us overclockers have an extremely cool PC!
Nobody makes a copper computer case because it'd be very expensive, and very heavy, and prone to fold up like a concertina if anybody sat on it.
Copper is a very tough metal - whack it with a hammer and it'll dent but not shatter or tear - but it's not a very strong one. This is why copper heat sinks seldom use copper for their structural components - the frame that holds the fan, for instance. That'll be steel, or aluminium, or really hefty copper.
You could make a copper case as rigid as an aluminium one, if you used a lot of metal to make it. But since copper's more than three times as dense as aluminium (it's 14% denser than iron), a case made from mighty girders of the stuff would weigh a very great deal indeed. A few decorative panels and, maybe, drive bays are one thing; a whole copper frame is another.
There are exotic copper alloys that have some properties approaching those of steel alloys. Beryllium copper, for instance, is good for making springs. You could make a case out of that, and it wouldn't weigh much more than an equivalently strong steel case. But a whole case made of beryllium copper would cost a fortune (the raw metal sells for something like $US60 a kilo...), and anybody who cut a hole in it for a side window would be poisoned if they breathed the dust!
In any case, the material a PC case is made from makes little difference to its cooling. Throughflow ventilation is what matters. This is why people can make computers with entirely wooden cases, or metal cases with wood over the top, that work perfectly well. Almost all of a PC's cooling is done by its fans, not by its metal.
A couple of days ago, me and some mates were cleaning out my garage, when we stumbled upon an old pair of Pioneer speakers that I never knew I had. I decided that they would look kinda cool attached to my stereo, so I stripped them back, and made them look almost cool. Then came the moment of truth - seeing whether they actually worked.
I hooked them up, and... nothing. In desperation, I asked my uncle what might be wrong. He said that I needed an amplifier or a pre-amplifier, I forget, I asked what sort, but he didn't know, and so I turn to you. Do you have any idea what sort of amp I would need? The speakers say "Max Input - 18W, Impedance - 8 ohms". Ry
In the olden days, sound cards had titchy built-in power amplifiers with a watt or three of quite badly distorted power per channel. With an adapter cable, you could just plug plain old speakers straight in, and get adequate volume for desktop purposes.
Modern sound cards don't have amps. They can't drive anything more challenging than headphones directly; hook up unpowered speakers and you'll get a whisper out of them, but that's all. So yes, you need an amplifier.
You don't need a fancy amplifier, though. Any old stereo receiver from a garage sale, or wherever, will do. All you ought to need to connect it is a 1/8th inch stereo plug to twin RCA plug adapter to hook up the computer's Line Out to your amp's Aux In. If the speakers have the usual bare wire leads, you ought to be able to connect them to the back of the amp with spring clips or screw terminals, so you shouldn't need an adapter there.
Just remember to connect the speaker wires the same way around on both sides; one wire will have printing on its insulation, or will be a different colour, and that wire should go to the same terminal on both sides. It can be stripe-wire-positive both sides or stripe-wire-negative both sides, but not positive one side and negative the other, or your speakers will be out of phase and sound weird.
Since PCs provide volume-controlled line level output, a simple power amplifier is all you need; you don't have to have an integrated amplifier (with its own pre-amplifier, so there's a volume knob on the amp) or a separate pre-amp. Cheap second hand amps are usually integrated amps, though, or receivers (integrated amps with a radio tuner built in).
Our office is looking to get a second ADSL service to use as backup when our primary (BigPond) ADSL goes troppo. Is it possible to use ADSL modems in tandem, to obtain greater bandwidth when both services are running?
Yes and no. Mainly no.
"Aggregating" two DSL connections - making them look like one connection, so that you can download a single file as fast as you'd expect by adding the speed of the two links - is something that your ISP has to support. Multilink bonding is not going to happen with any consumer DSL product, probably not going to happen even with current Australian commercial DSL products, and definitely not going to happen when one line is from one ISP and the other is from another.
If all you want is load balancing, though - where you use both connections at once, but not for the same thing simultaneously, so one client can be downloading two files and get each down a different pipe - then that may be doable. It's still not straightforward, though; if you're trying to use two connections simultaneously, you're showing two different IP addresses to the Internet, and if your LAN's Internet access is shared evenly between those two connections then you'll have various client machines that seem to be flickering from one IP address to the other, which'll screw things up royally. But it can be done, by someone with adequate networking chops (i.e. Not Me).
A dedicated appliance (like this one, perhaps) is certain to be the most straightforward solution, but you'd want to check carefully to make sure that it'll work in your situation before you buy. GNU Zebra running on a Linux or FreeBSD box can also apparently get this job done.
It should also be noted that since Telstra is, at base, responsible for all DSL connectivity in Australia - other companies are just reselling Telstra wholesale broadband services, and the data passes over Telstra wires - it's likely that another DSL service from another ISP will go down whenever your Telstra DSL does. It's not as if you can get a second DSL service that connects to a different telephone exchange, after all. You might do better to make your second connection some other broadband option - wireless, cable, satellite...
After reading the Supercomputer article in Atomic Issue 40, and in particular the insert about Garry Kasparov and IBM's "Deep Blue", I was left wondering why it took a supercomputer to beat a human at chess. It's chess - there can't be that many instructions involved to not only calculate the computer's own moves, and all of the opponent's counter-moves. Why does chess seemingly overwhelm the desktop processor?
The number of possible positions of all of the pieces on a chessboard is staggeringly high. Even if you only look at all possible chess games, and filter out repeated move scenarios, you've still got more of them to think about than there are atoms in the universe.
Yes, more or less "brute force" chess computers can already beat all but the very finest human players with no trouble (most of us, of course, lose to chess computers that run on AA batteries...), and will be generally superior to all of us meatbags not very long from now. They'll still be a long way from playing truly optimal chess, though.
A lot has been written about this; check out this for starters.
A while ago I heard somewhere that people (government officials...) may have been able to see data sent over modems just by recording the lights on the front.
I have recently brought this up and have been laughed at, and I was just wanting to know if this is really that far-fetched?
I would have thought it was possible, but I just can't find any info on it.
Yes, it is possible, in some situations at least. It could also be done, in theory, with network switches, routers and anything else that blinks an LED to indicate data traffic. LEDs respond very quickly - in the tens of nanoseconds. Even 50ns response time means response way up into the megahertz range.
This'll only be an issue, though, if the LED blinking is directly driven by the stream of bits - on for a one, off for a zero (or vice versa, or blinking only when a 0-to-1 or 1-to-0 transition happens, or whatever). If the LED indicates anything with a bigger granularity than that (like, for instance, blinking once when each multi-bit baud unit of a modern dial-up modem goes by), you can't read the data stream from it. Network card activity lights, for instance, seem to generally just latch on constantly when the card's sending or receiving.
You can't tell just by looking whether an LED is blinking zillions of times a second, or glowing constantly; you have to point a similarly speedy photosensor at it and plug the output into an oscilloscope or frequency counter or something.
The first article about this was linked and discussed on Slashdot some time ago. It's an extension of existing "TEMPEST" data security ideas. There've been a few other related developments since; differential power analysis, for instance, and acoustic analysis (which got no respect in this RISKS Digest), and even keyboard click analysis. Address bus snooping to crack security features in certain sold-at-a-loss consumer computing hardware is also somewhat relevant.