Dan's Data letters #45Publication date: 16-May-2003.
Last modified 03-Dec-2011.
When I bought my first PC, a 486 DX4/100, it cost me $AU1000. At the time (late 1995), that was a second rate PC, all I could afford at the time despite the squillions the government tossed my way to encourage me to complete my tertiary education.
Today, for $1000 I would still buy a second rate PC. The price of CPUs has remained fairly static; the top line Intel chip costs the thick end of $AU1000, maybe slightly more or less, and has as long as I can remember. A good chip a few steps down the ladder was, and is, about $300.
The price of RAM also remains similar, provided you take into account the increasing requirements PCs have for RAM. The 8Mb my 486 came with cost a similar amount to what I would expect to pay for 256MB RAM today.
Peripherals have a similar pattern. Some parts have dropped in price amazingly, like CD burners, while others have remained pretty constant, like sound cards.
And then we have graphics cards. Every new release, be it from ATI, Nvidia, Matrox, or whatever else, has increased the price for both the top end product and the mainstream product.
My first graphics card was a Cirrus Logic 4564 (I think); it cost $AU15. The next, the brilliant Tseng Labs ET4000, was $AU40. ET6000, $AU60. The first 3D card was an astonishingly expensive $AU400! The Voodoo 2 was $AU500, the TNT2 $AU550, the GeForce $AU600, the GeForce 3 $AU700, and this year graphics cards hit the ridiculous $AU1000 mark.
WTF!? How do they get away with it? Why do we accept the price rises in graphic chips, but expect Intel and AMD to keep their latest and greatest at more-or-less the same level? If Intel chips had come up at the same rate, we'd be paying $AU4000 for their 3GHz P4!
There's no Rule of the Universe that says that components of a given type have to stay at about the same price. Some stay about the same in price but increase spectacularly in specifications (RAM), some get slowly cheaper (monitors), some get cheaper faster (hard drives), and some get more expensive (video cards).
Bear in mind that video cards are gaining in complexity faster than every other PC component, including CPUs. The original TNT's eight million transistors made it "huge" by 1998 standards, but the GeForce FX 5800 has more than 125 million (the just-announced GeForce FX 5900 has 130 million). And that's before you start counting all that super-fast RAM.
The 7.5 million transistor P-II was state of the art in '98 (the then-about-to-be-superseded AMD K6 had 8.8 million); today, we've got Barton Athlon XPs at 54.3 million transistors (up from 37.6 million in the Thoroughbred), and Northwood P4s at 55 million.
So we've got about seven times as many transistors in CPUs now as we had in 1998, but "GPUs" now have more than 15 times as many as they did then.
Given how many junctions modern 3D cards have to cram in, it's hardly surprising that they're not cheap.
Hey Dan I got a puzzle that simply screams your name, maybe you know it already. With magnetic stripes on credit/ATM/transport cards etc, if the slide reader fails to read it after a few retries and the dreaded double take, there's one alternative... the plastic bag.
Place the offending card into a store-type thin plastic shopping bag, run the bag-clad card though the reader, and more often than not, it works.
This seems to be common knowledge among store clerks, but not a single one of them seems to know why it works.
This weird trick does indeed work - at least sometimes. And it can be done with things other than a carrier bag. Sticky tape or plastic wrap over the stripe, or a thin thermal-paper receipt slip wrapped around the card will work just as well. It's not infallible, but as you say, it often does make an unreadable card work again.
The plastic or paper or tape or whatever, as far as I know, works because it keeps the heads in the reader a reasonably uniform distance from the magnetic field of the stripe on the card, ironing out the little scratches and dents in the magnetic material.
Card readers have an automatic gain control (AGC) circuit that deals with the differences in stripe condition between different cards. Generally speaking, the older a card gets, the weaker its stripe's magnetism will be, probably just because of mechanical wear - the more stripe material wears away, the fewer magnetic domains remain. The AGC auto-calibrates the reader to compensate for this.
But if an old card's unevenly worn such that the magnetic material's strength varies substantially along its length - as can easily be the case, for a card that's been kept in a wallet that isn't subject to uniform movement along its length - then the AGC won't be able to keep up, and the card won't be readable.
If this unevenness in strength is caused just by the proximity of the head to the stripe on the uneven card, then wrapping the card with something thin and magnetically invisible and then swiping the card so that the card is pressed against the wrapping material should even out the changes and, thus, allow the card to be read. Which, it seems, is why the magic-bag trick works.
This, incidentally, is also the basis of a scheme for ripping off magnetic stripe cards that hold information about an account balance.
Most credit/debit cards just hold identifying information, and a central server of some sort holds the info about the account balance. You can change your card to look like someone else's if you've got a card writer and know their card's stripe contents, but you can't alter your own balance, or anyone else's, by doing stuff to the card.
Cheap magstripe systems for relatively unimportant things, though - like some swipe-card photocopiers, for instance - have the remaining account balance coded right into the card, removing the necessity for the server. A layer or three of tape on such a card will, for some such systems, still allow the read heads in the photocopier (or whatever) to read the account balance.
But the write heads will be unable to change the stripe contents to account for the transaction.
Bingo; free photocopies!
I am looking at getting a Kodak DC3800 digital camera for about $AU240. Is this a reasonable camera for that price?
My wife and I are going to South America, Nepal/India and Tibet. We have nieces and nephews and would like to show them some of the things we see. The photos are not likely to ever see paper - only PC screens - and will mainly be taken of views/mountains.
The only drawback I see is that the DC3800 has no zoom, so if we see a jaguar (fat chance) or a tiger it will only be a spot in the distance.
I have read reviews of the DC3800, and the only problem someone came up with was that they got really lousy photos. Could this be because they didn't wait for the auto focus to work? Five other reviews liked it.
My problem is I can't find it in a shop anywhere to look at, only in one Web store. Is $AU240 reasonable?
I don't know much about the DC3800 in particular, but I do know that this is a good review of it. I think $AU240's a decent price for a new one. It's not at all a new product, of course; it was $US499 when it came out more than two years ago.
For landscapes, this camera ought to be OK, because it's got a decently wide angle lens, and autofocus means you won't get fuzzy shots of distant subjects, which are the bane of fixed-focus lenses.
As you say, it'll be no use for telephoto work. The only way you'll get a decent picture of a big cat is by doing something that'll result in someone later cutting your camera out of the tiger's gullet.
As regards other users' dissatisfaction with the DC3800 - well, anybody can take awful pictures with any camera. The DC3800 is strictly limited in what it can do - only wide angle shots, no manual controls - but it looks to me like a perfectly good little camera for incidental shots while travelling. Small, simple, light, cheap. You just have to learn to not even take it out of your pocket if you're looking at something it won't be able to photograph well, and you also have to live with the fact that if it can't get the exposure right the first time you take a picture of a given subject, it's unlikely to ever do better.
It's not surprising that you can't find a DC3800 in the shops any more, since it's been well and truly superseded, but there are a few of them on sale here and there. An ebay.com.au search, as I write this, finds three on sale here in Australia; they're on sale on eBay in the States as well.
Magnetically controlled fire doors are pretty common in the U.S. now. You'll find them in hospitals, dormitories, even shopping centers. These are doors with a metal foot at the top. On the wall, in position to match that metal foot, is an electromagnet. The door can be pushed open, and the magnet will hold the door open all day, or all month. If the fire alarm goes off, the circuit is automatically cut and the door will swing shut. If someone wants to, a good firm tug will pull the door away from the magnet, so it can be shut manually.
As much as I love the groovy sci-fi quality to doors that swing shut in case of danger, not to mention the clever use of magnetic force, how can this possibly be efficient? A hospital might have a hundred of these doors... and they can't be at risk of losing strength and slamming into some kid on crutches. And these doors are heavy! So, wouldn't the electricity necessary to maintain the magnetic force holding these doors be immense? Even with the magnet being as far away from the fulcrum as possible - which in some cases it is not - that's a pretty strong attraction needed. Sometimes these are steel doors - the kind little kids can't push open.
How do these things work?
The design of the magnetic hold-opens, as they're called, means they need surprisingly little current. They've got a pretty hefty coil, with a field-intensifying pole piece down the middle of it, and they attach to a thick steel plate that's cast to match their profile very accurately.
I only found one place quoting actual current consumption specs for these things, here; it says it's only 30mA, though it says that's from 12 to 24 volts, which is a bit odd. Anyway, it's not much; even 100mA at 24 volts, times however many doors there are, is trivially little power compared with a building's whole electricity budget.
This is, I think, supported by the way these latches behave; they're specifically designed to disengage easily if you give the door a push. There may well be specialist high power models made for doors that people aren't meant to be able to manually close, but the regular kind of magnetic hold-open devices aren't very strong.
This page has no electrical specs, but its installation and troubleshooting PDF files are quite interesting.
Note that the weight of the doors is immaterial; it's the strength of the closer mechanism that matters. Since the closers on general purpose auto-closing doors don't pull terribly hard - otherwise the doors could be quite dangerous, or at least difficult to open manually - you don't actually need a terribly strong magnet to hold them open.
And now, a letter that I couldn't answer. I had some ideas, but they were all wrong. Read it, have a think, see if you can figure out what the problem is - then click on through to the page with the answer on it.
If... you dare.
One that got away
My two home-built systems, stuffed full of Athlon and Mb and Gb and raw power, work flawlessly in almost all respects.
Except for the floppies.
They don't work. Well, they don't work in all cases. More to the point, they don't work when screwed into the cases, as normally you'd want them, mounted in their respective 3 1/2" slots.
Through trial and error, I've found that they do work when hanging away from the cases, not touching (grounded) to the frame, dangling only from their floppy cables and power connector. I'm not making this up. In my two systems, the floppy drives, which pass power-on tests (the systems complain if I disconnect them that the floppy drives failed), report "Drive not ready" whenever I try to actually use them when the systems are up, and the drives are secured in the cases. Moving the drives outside the cases, isolating them from physical contact with the metal frame, brings them to life (after a reboot).
I'm running Win2000 on both PCs (one is Server, one Pro), and they have few similarities beyond the case manufacturer. One has an Epox EP-8K3+ motherboard, the other an Iwill KK266. I've tried several different floppy drives in each, new floppy cables (including rounded cables - I figure at floppy transfer rates cable crosstalk isn't a deep concern), and BIOS reconfigurations that might apply, with no success.
I suspect the cases, and their native power supplies, are culprits in this all. I also don't believe for a second that the cases can or are contributing to this, because that's just crazy. But that's what I'm down to.
Both are unmodified "Performance Series" from Antec, one an SX830, and the other an SX635. Their power supplies are stock with the cases, both "Smart Power," with the SX830 housing their 300W supply, and the SX635 using the more potent 350W supply. I'm suspecting the cases because it's the only common link between the two systems, and because, while details are fuzzy, I never had such a problem before I upgraded both systems cases at around the same time, a few months back.
That said, it doesn't make sense. Two days ago I built a system for my mom, using a $40 sure-to-cut-your-hands case and spare parts to assemble a suitable K6-III+ system she'll use for e-mail. It's an upgrade from her even lesser P200MMX system that fell to age and virus attack. Both systems, of course, had working floppies.
In fact, I've built quite a number of systems, and never had such a problem.
Now, both computers are plagued with non-responding floppy drives when mounted, and every single attempt I've made to fix it or simply upgrade the systems (the KK266 mobo used to be in the SX830, but I moved it to the smaller case, and upgraded the old Asus mobo that used to be in the SX830 to the DDR-supporting Epox mobo, including an upgrade to a T-Bird 1900+ from the previous Athlon 1.2GHz) doesn't resolve the problem.
I've been able to live with it since floppy drives are rarely needed today. But on those occasions when I do need them (BIOS upgrades, running DriveCopy to clone my laptop drive using those systems) my routine is to disassemble the systems, pull the floppy out, isolate it from physical (electrical) contact from the case, ensure everything's plugged in right, and boot. Works fine. Even then, with the systems up & running, floppies hanging in the breeze but working just fine, I've experimented by grounding the floppy to the case via alligator clips to see the effect. Nothing. Keeps working fine. It's only if I mount the floppy in the cases, and boot, that they fail to respond.
There you have it, folks. I was thinking broken cable conductors, some kind of eldritch electromagnetic interference issue, cosmic ray strike...
I wasn't close.
The real answer's here.