Computer Shopping Update
I was recently reminded that I haven’t written about computers in these pages for some time.
I wrote an article about shopping for computers in the spring of 1983, and announced a couple updates to that article in later issues. The last re-write of the shopping article has grown to 40 full-size pages, and was done in the spring of 1986. That is available to any subscriber who wants it for $2.
Since I wrote that, the field has continued to change at its usual pace.
I recently timed our chart program on several different machines, and got a graphic demonstration of the increase in computing power available on a home budget. The following list is based on computers we have bought in the last four years.
In the spring of 1983, a new Osborne Executive cost $2495. That bought a machine using a 4 MHz Z-80 (8-bit) CPU, 128K RAM, two 183K 5.25 inch floppy disk drives, a small monochrome text screen, and a lot of included software (CP/M+, WordStar, CBASIC, MBASIC, SuperCalc, Personal Pearl, UCSD p-system). The only upgrade we have gotten for this machine was a larger separate screen. Our program, running under interpreted MBASIC, takes 78 seconds to calculate and store one natal chart on this machine.
In the spring of 1984, a new Apple Macintosh cost $2495. That bought a machine using a 7.83 MHz 68000 (32/16-bit) CPU, 128K RAM, one 400K 3.5 inch disk drive, a 9-inch monochrome graphics screen, and limited software (Finder, MacWrite, MacPaint). We have upgraded this machine twice because the original configuration was severely limited: first to 512K, then to 1M (1024K). It started with one 400K disk drive. We added a second 400K drive, then upgraded one drive to 800K, than added a 20M hard disk. Our program, running under interpreted Microsoft BASIC, takes 24 seconds to calculate and store one chart on the hard disk, 28 seconds on an 800K disk, or 29 seconds on a 400K disk.
In the fall of 1985, a new Amiga cost $2255. That bought a machine using a 7.16 MHz 68000 CPU with special support chips for graphics and sound, 512K of RAM, two 880K 3.5 inch disk drives, a 13-inch color graphics screen, and limited software (AmigaDOS/Workbench, Amiga BASIC). For charts, it comes out a little slower than the Macintosh (33 seconds). The area where this machine excels is in fancy sound and graphics, because of the special support chips. The Amiga has also had a multi-tasking operating system from the day it was shipped, something that is still in the future for most machines on the market.
In the spring of 1986, a new Kaypro PC listed for $1495, but we got it for $1293. That bought a machine using a 4.77 MHz 8088 (16/8-bit) CPU, 256K RAM, two 354K 5.25 inch floppy disk drives, an 11-inch monochrome text screen, and a fair amount of included software (MS-DOS, WordStar, Mite, PolyWindows). We upgraded this machine immediately after purchase by adding RAM to fill it to its 768K limit, a 20M hard disk, and an 8087 math chip. Now the same machine offers a higher speed and 768K RAM for a lower price. Running interpreted GW-BASIC on floppy disks, a chart takes 67 seconds. However, on that same machine, I can now get a chart calculated in 4 seconds! The most recent BASIC compiler from Microsoft (QuickBASIC 3.0) can use an 8087 math chip, which I have installed on the K-PC. Running on a hard disk and using the compiled version of the program which uses the 8087, one chart only takes 4 seconds.
In the spring of this year (1987), a Z-NIX 286 (an AT clone) cost $1945. That bought a machine using a 6 or 10 MHz 80286 (16-bit) CPU, 1024K RAM, two 1.2M 5.25 inch floppy disk drives, a video controller which automatically switches between several video modes used on PCs, a 13-inch color graphics (NEC) multisync monitor, and only the software we chose to buy (MS-DOS with GW-BASIC). (The video controller and monitor were bought separately, but are included in the system price above.) The machine also got an immediate upgrade of an 80M fast hard disk, which was another $1000 not included in that price. On this machine, on the hard disk, a chart takes 1 1/2 seconds to calculate & store. On this machine an 80287 math chip gives no noticeable speed improvement for charts.
The next step in the ongoing hardware progression will be machines based on the 80386 and 68020 CPUs. Both of these are true 32-bit CPUs with high clock rates (the slowest run at 16 MHz). These machines have not come down to the home price level, but they are moving in that direction. Some are already in the $2000–$3000 price range.
All of those times are for the same set of programs, doing the same thing. We can now calculate charts 50 times faster than four years ago! Some of that increased speed is from improvements in hardware, but a large amount of it is also from improvements in software. Compiled programs run several times faster than interpreted, and newer versions of both compilers and interpreters are faster than older versions.
To show this distinction more clearly, here are some examples: 78 seconds on the Executive, 67 seconds on the Kaypro PC, and 19 seconds on the Z-NIX 286 are all times using interpreted BASIC on floppy disks. Moving to hard disks on the K-PC and Z-286 cuts the times to 56 seconds and 11 seconds. The software is comparable, so these speed increases represent hardware improvements. Switching from interpreted BASIC to compiled (QuickBASIC 3.0), the K-PC and Z-286 times for floppy disk operation go down to 13 seconds and 9 seconds, while the hard disk times go down to 8 seconds and 1.5 seconds. These speed increases represent software improvements. Switching to a version of the compiler that can use the 8087 math chip on the K-PC, we get floppy and hard disk times of 10 seconds and 4 seconds. This represents a hardware improvement (using a special math chip), but this hardware doesn’t do any good unless the software can use it! I had the 8087 math chip in the K-PC for a year before an affordable BASIC compiler that could use it came out. The interpreter still doesn’t use the 8087 chip.
We can also handle much larger numbers of charts much more easily because of increasing disk capacity. On the 183K Osborne disks, we could store 292 charts. On the 1.2M AT disks, we can store 1896 charts. On a hard disk, we can store many thousands (our 80M disk would hold 130,000 charts if we didn’t put anything else on it).
More dramatic than single charts, primary directions take 4 hours to calculate on the Osborne Executive running under interpreted MBASIC, 26 minutes on the K-PC running compiled QuickBASIC without using the 8087, and only 2 1/2 minutes on the Z-286 running compiled QuickBASIC using the 80287 math chip. Our best time here is 89 times faster than our worst, changing primary directions from something unmanageable to a technique worth trying.
One trend in the computer industry is pretty obvious: the mice are moving in!
On most computers until recently, you told the computer what to do by typing commands, usually with no prompting available about the proper spelling and syntax of the commands.
The Macintosh popularized an operating environment using a mouse (a small box you move around on your desk to move a pointer on the computer screen) and icons (pictorial images on the screen instead of text). You tell the computer to do things by using the mouse to move the pointer to an icon or to a choice on a menu, then clicking a button on the mouse. Unfortunately, the Macintosh feels slow because the processor (CPU) is overburdened.
The Amiga also comes with a mouse-based operating environment called the Workbench, but you have the option of using a text command based environment if you prefer that. Even with its special support chips for sound and graphics, the Amiga also feels slow (mostly in disk use rather than screen displays).
PCs (the MS-DOS world of IBM PCs and clones) are evolving in the direction of a mouse-based environment, but slowly. They haven’t had enough power till recently. Microsoft Windows and Digital Research GEM are two currently available icon environments for PCs.
Mouse and icon based environments are slow on most current processors because the icon environment usually requires a full screen of bit-mapped graphics, which requires a lot of memory and processing power. Special chips designed for fast graphics can help a lot, but there is still more work for the CPU than with a text command environment.
An icon based operating environment is a little easier to learn, though I think many people overstate how easy it is. The area where there is a real improvement is that it is easier to remember once learned. When programs are written to adhere to the standards set by the operating environment, it is also easier to learn new programs because they are more similar to each others’ behavior than programs in text command environments.
The Macintosh has been harder to program on than CP/M or MS-DOS machines, but it has gradually gotten a good set of working languages for programmers. (I sometimes say that it is an Apple tradition to be hard to program, since that was also true of the Apple II.) Aside from the lack of language support on the Macintosh, the other reason it is harder to write for is that it is supposed to be easier to use. Writing programs that are easy to use takes MUCH more work than writing programs that work. Also, adhering to the standards imposed by the operating environment can feel like working in a straitjacket.
The PC world has had a much greater variety of languages and other programmer’s tools available for years, and has also been a much larger market in terms of the number of customers available, so much more programming has been done for PCs than for the Macintosh. Since the PC world is moving in the direction of mouse-based environments and the Macintosh market is now large enough for more people to write for it, there will be increasing crossover and overlap between Macintosh and PC programs.
Another change in the small computer market since I last wrote is in the area of printers. You are seeing a result of this trend right now. The camera-ready copy of this issue of The Mutable Dilemma was printed on a Hewlett Packard LaserJet series II. We have been using a Brother daisy wheel for several years, taking several hours to print and several more hours to paste the proper pages opposite each other for printing. The LaserJet cuts the print time from hours to minutes, and completely eliminates the paste-up time by allowing us to print sideways on a page to get the correct pages printed facing each other. The LaserJet is rated to print 8 pages per minute, and actually does print plain text that fast. It slows down a little when there are a lot of formatting commands, and slows more for graphics, but it is still 5 or 10 times as fast as any other printer we have. We have printed hundreds of pages of data for the ISAR database project in a fraction of the time it would have taken before. It is also QUIET! You have to listen closely to hear it from across the room, which is a dramatic contrast to the Toshiba we keep in a sound-deadening enclosure.
The LaserJet lists for $2500, and is available locally for about $1700. This is less than we paid for the Toshiba 1350 when it was first available about 5 years ago. The LaserJet does not include PostScript, a page description language used in the Apple LaserWriter which makes graphics programming much easier, but we will probably add PostScript capability later if the price comes down enough. If your use of a computer produces a lot of pages of text, you should definitely think about a laser printer. (Right now I don’t know if any astrology programs support laser printers, but I will support the LaserJet in the CCRS program as soon as I have a little time to learn its way of doing graphics.) I chose the LaserJet because it has more expansion capability than some of the others on the market and HP codes are a standard, but there are others on the market even cheaper than the LaserJet.
Hank Friedman is working on an update to his book Astrology on Your Personal Computer. If you have information about small companies in the astrological software field, feedback based on personal use of astrological software, or ideas for things you would like that are not being done in current astrological software, write to:
1236-B Gilman St.
Berkeley, CA 94706