A Pentium Is Sounded Out

Jerry names his new Pentium system, which he thrusts into the multimedia stream by testing several sound boards

Jerry Pournelle

We're just about recovered from the earthquake. Indeed, now that I've gotten rid of about 50 cubic feet of junk and sent more--including the 8-inch disks on which I wrote what appears to be the first novel ever written using a computer as a word processor--to storage, and have rebuilt my bookcases, the Great Hall is cleaner than it has been for five years; meaning that there's room for new stuff, and that wasn't long in arriving.

It's a lot of fun out here at the bleeding edge of technology, but it's not all roses. I've just spent three very long days getting a new Pentium system to behave. I'm not really complaining: I know there are a lot of young writers out there who 'd love to trade places with me, and if I slow down, one of them may just be able to do it. Anyway, I learned a lot, and now you won't have to make the same mistakes I did.

It all started when Larry Aldridge became vice president of marketing for PC Power & Cooling. He wanted to build me a Pentium system to demonstrate just how well the PC Power & Cooling tower-configuration cases, power supplies, chip-cooling fans, and heat sensors can solve Pentium heat problems. Before it was over, he used a Micronics PCI (Peripheral Component Interconnect) motherboard, a really neat DEC DSP3107 1-GB SCSI hard drive, a Plextor Double Speed Plus CD-ROM drive (formerly the Texel CD-ROM drive), and a Distributed Processing Technology SmartCache SCSI caching hard drive controller that runs both the DEC hard drive and the Plextor CD-ROM drive. SmartCache will shortly run my Palindrome DAT (digital audiotape) drive and one or more optical drives. ATI Technologies contributed their Graphics Ultra Pro Mach 32 PCI video boar d. Larry put the basic system together and tested it before sending it on to me.

Our NEC MultiSync 4FG monitor was damaged a bit in the earthquake. Actually, the only real damage was to the swivel base. The monitor works just fine--we now use the fire ring for a wok as a base--but NEC sent a MultiSync 5FG to replace it. I connected that to the new Pentium system and got out one of my "Pournelle-configuration" Northgate keyboards, in my opinion still the best keyboard around. When I turned on the machine, the video came up, but nothing else happened.

I suppose I had a premonition, because I rolled out one of the heavy-duty worktables and laid the new tower-configuration machine on its side. Opening the computer revealed that the SCSI cable had disconnected from the controller, so it was no wonder the system didn't boot. Connecting the cable fixed that. The machine booted up fine with DOS 5, QEMM 6.03, and Windows 3.1. So far, so good.

There was no network card in the Pentium system, so I u sed Traveling Software's LapLink Remote Access for the initial file transfers. If you use LapLink to connect to a machine that is itself connected to any network, including Windows for Workgroups, you have access to the entire network. Even connected with a parallel cable, it's slow compared to Ethernet, but it does work, and I used it to bring over Norton Commander. I always do that first thing. There are, I suppose, other ways to edit, copy, rename, move, and delete files, but Commander's ability to look into odd-format files is pretty nifty, and the file finder is as good as any I know. Mostly, though, my fingers know Norton Commander down to the cellular level, and I'd be lost without it.

LapLink demands that you name the machine it's running on. I always do that anyway, and the new Pentium system has become Percival, as opposed to Percy, which is the IBM PS/2 Model 77.

Once I got Norton Commander up, it was time to install a sound board. Since everything was working fine, I hadn't bothered t o sit down and read the documents that came with Percival. I just got out the sound boards and started in. That's when my education began.

The first thing I tried was a Creative Labs Sound Blaster 16 SCSI-2 board. Alas, when I turned on the machine, it wouldn't boot. I thought about that. The SmartCache controller has a SCSI connection, and the Sound Blaster board has a SCSI controller built onto it as well. Can two SCSI connections exist on one bus? If they can, can they both be addressed to the same SCSI address?

As it happens, the answer to both questions is yes. I ought to have known that, but for some odd reason I assumed the problem had to do with SCSI interference, and I looked for another sound board. The next one I found was a Sound Blaster Pro; alas, the board didn't say what model it was. The only identification on it was "CT-1600" written in the upper left corner. The machine came up fine with that, but all my Sound Blaster Pro software disks were destroyed in the earthquake. Since I wasn't sure what board I had, I tried the Sound Blaster 16 software. That's very well done stuff, but of course the software couldn't find the board. It offered to look at a variety of addresses with a number of interrupts, but it eventually reported failure.

Because the Sound Blaster Pro is last year's stuff anyway, I put in a call to Creative Labs to get a new Sound Blaster 16 board with the latest software. Then I looked through recent arrivals. Sure enough, we had a Logitech SoundMan 16 board. This actually incorporates two chip sets on one board: the 8-bit Creative Labs Sound Blastercompatible chip set, and the 16-bit Media Vision chip set, called SoundMan 16. There are no jumpers on the SoundMan; it's all set up by software.

The installation requires two separate IRQ (interrupt request) lines, one for Sound Blaster and the other for SoundMan. Some older games demand a Sound Blaster at IRQ 5; the Logitech board offers a choice of 2, 3, 5, or 7, as well as three I/O port addresses. The Soun dMan IRQ can be set to just about any value between 2 and 15. There are also choices of DMA channels between 0 and 7; you need two of those--one for Sound Blaster and one for SoundMan--with default values of 1 and 5, respectively.

DMA is a way for devices to send data directly to memory without funneling it through the CPU. DMA channels 03 are 8-bit; channels 47 were introduced with the IBM AT 16-bit bus and are 16-bit.

The SoundMan installation software performs a bunch of tests. Once you've chosen values that work, it offers to write them into your initialization files. I had to fiddle with some of the settings. In particular, DMA channel 1 was not working properly.

I called my friend Rich Heimlich. Rich tests games for a living. He also knows more about using sound boards than anyone I know and quite literally wrote the book on Sound Blaster boards. He said that he had seen this problem before: some motherboards step all over DMA channel 1, apparently through cross talk on traces laid too close together. The remedy is to use a different DMA channel (generally 3), and, alas, some applications, including most older games, expect Sound Blaster on DMA channel 1 and can't use another.

This was worrisome. I called Aldridge, who called Micronics, and we soon had the answer: the Micronics motherboard lets you use either DMA channel 1 or 3 to connect to the parallel port. Doing that converts your port to a fast parallel port (for software that knows how to use it). This is highly desirable. More and more applications know how to make use of fast parallel ports, which are rapidly becoming a kind of poor man's Ethernet.

In assembling Percival, Aldridge had left the DMA channels floating free, and that, as it turns out, can generate noise in channels 1 and 3, which makes them unreliable. The remedy is to use jumpers to strap the parallel port to DMA channel 3. I did that, and Sound Blaster was able to use DMA channel 1 just fine. After that, I had no problems installing the SoundMan soft ware.

Next step was to install an Intel EtherExpress card. Like SoundMan, the EtherExpress card has no jumpers; all settings are done with Intel's SoftSet software. The card requires an I/O port address and an IRQ, and those have to be different from the ones used by the sound board. SoftSet will try to find values not in use by anything else, but you can choose your own settings. Since nothing I know of uses IRQ 10, I set the network card to that. Then I installed Windows for Workgroups 3.11.

The easy way to do that is to create a disk directory called W4WG, copy all nine installation floppy disks to it, and then do a custom installation. This not only saves time, but W4WG can sometimes get so confused during installation that it forgets how to access floppy disks. However, it never forgets how to find a hard disk directory. The installation went smoothly enough, and everything worked just fine until Alex got home.

Alex has a sort of portable "Murphy" field: if something can break, he can break it. In this case, he found problems with the sound. A quick check showed why that wasn't working: the EtherExpress card was set to use IRQ 5, the same one as the Sound Blaster chip-set part of the SoundMan 16 board. I knew I had set that to IRQ 10, but then I remembered: W4WG likes IRQ 5 as the setting for the EtherExpress card, and when you install W4WG, it sets the card to that value no matter what value it finds it at. It doesn't bother to tell you it did that, either. In case you're wondering, this is a bug.

The remedy is to bring up Windows (use WIN /n to bring up W4WG without the network if the IRQ clash is so severe that it kills Windows), go into the Network window, open Network Settings, double-click on the Intel EtherExpress line in the menu, and set the IRQ to 10. You'll have to restart Windows, but this time the card will stay at IRQ 10.

We then started importing some sound effects--and achieved one of the most spectacular disk crashes I've ever seen.

It happened all at once. We were experimenting with MIDI wave sounds, and suddenly the C drive was filled with lost chains and cross-linked files. Pretty soon, one of the cross-linked files was COMMAND.COM, after which the machine would work only when booted from a floppy disk.

The DSP3107 hard drive is divided into two 500-MB logical drives. The D drive was fine, but the C drive was messed up beyond hope of repair. Sometimes you can get out of a bad situation with Norton Utilities, but this time the Norton programs were themselves cross-linked. There was nothing to do but reformat the hard disk, or at least the C drive.

Unfortunately, even that didn't do it. We brought up the machine with a floppy disk that had FORMAT.COM on it, but when formatted with the /S qualifier, the C drive still wouldn't boot. It showed IO.SYS and COMMAND.COM as truncated. Oddly enough, though, using the SYS command took care of the situation. By now it was 4:00 a.m., and I went to bed.

Next morning I removed the sound board and c ompleted the recovery. That didn't take long, because when I copied the W4WG floppy disks, I did it to the D drive. Once W4WG was running, it took almost no time to use the network to restore nearly everything to the C drive.

One thing I didn't restore was the version of QEMM that Larry had used in setting Percival up; that was just plain gone. On the other hand, I have a new copy of QEMM 7.03, so I installed that. It works just fine, yielding 600 KB for DOS and Windows without using QEMM Stealth. When everything is stable and backed up, it will be time enough to experiment with Stealth. For the moment, I don't have any programs that need more than 600 KB, and I'll leave well enough alone.

Actually, things didn't go quite as smoothly as I implied. For about an hour, I had more mysterious problems. Then I found out that my SmartCache controller was addressed to DC00, and if you don't tell QEMM to exclude DC00 to DEFF, the system can't possibly work. That done, though, everything was fine without the sound board.

Then I did what I should have done in the first place: read Percival's documents, particularly those having to do with the controller. It didn't take very long to see what the problem was.

SmartCache uses DMA to speed up disk reads. In particular, it uses DMA channel 5. The SoundMan board also uses DMA channel 5. When SoundMan put a wave sound onto channel 5 just as SmartCache did a write to disk, we wrote "Hail Columbia" all over the FAT (file allocation table). No wonder it crashed.

The simple remedy was to reinstall the SoundMan board and be very careful not to use DMA channel 5. With channel 3 strapped down to the parallel port, channel 1 worked just fine for Sound Blaster, and SoundMan had no problem with channel 7. We played around with enough MIDI wave sounds to be sure everything was working. No problems.

The Logitech SoundMan board works just fine, and it's compatible with both Media Vision's Pro Audio and Sound Blaster, as well as MIDI and various Windows sounds. We had the WindSurfer flopping about, and SoundMan comes with Icon Hear It, a way to associate sounds with different icons and Windows activities. Alas, there are problems, not with SoundMan, which works about as well as anything else, but with the whole multimedia sound business.

There are problems with cables. Every CD-ROM drive seems to have a different pin-out for internal audio output. At this point, I'd better explain for those just getting started.

Your CD-ROM drive is similar to the CD player in your stereo. Both can take recorded sound off a CD and send it to an amplifier or headphones. In addition, a CD-ROM drive can get digital data from a CD and move it to the computer bus; the computer translates that data into sound that you can hear. These two kinds of "recorded sound" are quite different, and they come out in different places.

Digital audio--like all other digital data--comes out of your CD-ROM player through the big, flat SCSI cable. It goes to your CD-ROM control ler (which may be your sound board; in my case, the CD-ROM drive is on the same SCSI cable as my regular hard drive, and the controller is SmartCache). By contrast, standard audio comes out the front of your CD-ROM player; you can hear it by plugging in a stereo headset. In addition, every CD-ROM player has an audio connector on the back of the drive, and, with the proper cable, you can pipe that audio to the sound board so it can play it through the speakers.

Unfortunately, there are no standards for that internal audio, and every CD-ROM drive seems to use a different cable connector and pin-out. Most sound-board companies will sell you a cable if you tell them what model of their sound board you're using and what make of CD-ROM drive you have.

Of course, if you buy your sound board and CD-ROM drive as part of a kit (e.g., Creative Labs' Sound Blaster Digital Edge CD Multimedia Kit), you won't have that problem, because the internal cable will be included. That's one of the main advantages to b uying such kits: everything works together.

In my case, SmartCache is faster than Sound Blaster, and the Plextor CD-ROM drive works well with it; and, of course, the controller gives me hardware caching on the controller. Understand, very few applications need that much speed or, indeed, can take advantage of it, so the Sound Blaster Digital Edge CD Multimedia Kit is plenty good enough for almost anyone and is one of the easiest ways to get both sound and a CD-ROM drive.

Anyway, I didn't have an internal cable; so I took a Radio Shack 6-foot stereo audio cable, plugged that into the earphone jack on the CD-ROM drive, and plugged the other end into the input jack (not the microphone jack) on the back of the SoundMan board. Later on, the same thing worked when we changed over to a Sound Blaster 16 board. When everything is stable, I'll get the right internal cable and connect things up, but for now the external cable works, and it's the only audio-cable standard in the business.

Once I was s ure the SoundMan board worked, I wanted to try Creative Labs' Wave Blaster add-on board, because everyone tells me it's neat. My only problem was that when I put the Sound Blaster 16 board in, Percival wouldn't boot. All my Sound Blaster 16 boards have SCSI on them, and if there's any way to disable that, the documents don't tell me. There's also no way to change the SCSI address from whatever default it's set to. There is a way to readdress the SmartCache board from SCSI 7 to SCSI 6.

Fortunately, Alex got home before I did anything to the SmartCache board, because the SCSI ID is irrelevant; you can have two SCSI buses going at once, provided each has its own IRQ and DMA channel. The reason that Percival wouldn't come up with the Sound Blaster 16 board installed is that one function of that board defaults to DMA channel 5. As we've seen, SmartCache uses that, too.

That was when I got out the Distributed Processing Technology documents and studied everything. They're quite complete and clear enou gh. I found out that the controller uses memory address C800 by default, but it's set to DC00 on mine; it needs an I/O port address, and that can be 330h, an address the Sound Blaster 16 board defaults to; another place to be wary. The interrupt is set to IRQ 15, which is unlikely to be used by anything else.

Creative Labs provides complete instructions with diagrams on how to use jumpers to set all the various addresses, IRQs, and DMA channels on their Sound Blaster 16 board. One warning: it's in the thin document called Getting Started and is not repeated in the thicker Reference Manual; you should keep both. We used the manual to set the 16-bit DMA channel to 7 and otherwise used the defaults on the Sound Blaster 16 board. Those include a default to IRQ 5, the same as W4WG, and the MIDI port defaults to 330h, which might be in use by either the controller or the network card. I made sure those didn't interfere.

The machine came up just fine. Alex installed the Sound Blaster 16 software, inclu ding the Windows drivers. We made sure that all worked. Then we shut down and attached the Wave Blaster board. Wave Blaster piggybacks onto the Sound Blaster 16 board, and it does wonders for the sound. It has no settings; just attach it to the Sound Blaster 16 board.

With that in place, Mr. Spock really sounds like Mr. Spock, music has a brilliance it didn't formerly have, and it all sounds really nice, especially if you've got Creative Labs' Yamaha-designed Active Servo Technology speakers. These high-end speakers are about twice the size of the (actually very good) smaller speakers that come in the Sound Blaster Digital Edge CD Multimedia Kit, and if you care a lot about sound quality, they're worth the price. Of course, if you really care about sound quality, you'll feed the sound-board output into a regular stereo mixer, but that's going farther than I care to.

Next thing to look at was ATI Technologies' MediaMerge. This comes with an audio editor, a text editor to make text animations and flying logos, a scene editor, a storyboard editor, and a partridge in a pear tree. We've looked only at the audio editor. Alex used it to tune up the wave forms of various sound icons, add reverberations, clip sounds, and put in silence. He also tried to do some noise removal. This is as good as any wave editor he has seen, but note that he's not especially fond of any we have; none of them lets you change pitch, for example. He was able to put together some sounds so Mr. Spock can respond when the ingenue screams for help. More on that another time.

Alas, we're still getting anomalies. If you leave the system running with Berkeley Systems' After Dark with sound, after an interval that can be as short as 10 minutes and as long as 8 hours, the sound is messed up. Simple sounds work, but wave sounds are truncated or don't play at all. The only remedy is to power down the system. Neither Ctrl-Alt-Del nor hardware reset fixes it, and once I have the sound problem, I can't even cleanly exit W4WG.

I d on't know why. My guess is drivers interacting with the Pentium, but whether it's Berkeley Systems' software or the Sound Blaster 16 driver isn't clear. It's certainly not a defective sound board; I've tried several Sound Blaster 16 and Wave Blaster boards, and I always get the same result. I'll keep looking, and I expect the problems will be fixed by the time you read this.

We've also had difficulty getting Word for Windows to print across the network. The symptoms are precisely the same as when we tried to get an OS/2 machine running OS/2 LAN Server to print through a W4WG network (see my April column). Because we were able to get a W4WG machine to print files through a printer attached to that OS/2 machine, I have to conclude that the guilty party is Print Manager, not OS/2.

It's clearly the case that if I attach the printer to the OS/2 system, I have no problems printing from any machine on the network, while even without any OS/2 machines on the network, the best way to print through W4WG is to send the file across to the machine that has the printer, read it into Word, and print it locally. Another anomaly to look into. I'm sure it will be fixed when I learn the proper settings for the W4WG machines, but the fact is that OS/2 LAN Server seems to be more reliable when printing across a network than W4WG.

This mostly shows that we need a good Windows debugger. Windows is cool if you get everything set right, but no one really knows what's going on inside the .INI files. It's a pain to get Print Manager set up properly. Then, too, every now and then when Windows comes up, it reports that a group file has been clobbered. When that happens, you can replace the dead group file with a copy you've saved--and if you don't save copies of your group files in their own subdirectory, you'll one day wish you had--or you can painstakingly rebuild the group file by hand.

Actually, you'll need to do both. That is, you must go into the Windows subdirectory and delete the aberrant group file, come b ack to Program Manager and create a new group file, copy your saved group file into the Windows subdirectory, and restart Windows. With luck, that will solve the problem of a clobbered group file.

You have to do all this because Windows has no debugger and group files can't be edited by human beings. No one knows why. If someone knows of a good Windows debugger, please let me know.

By contrast, System Notebook (formerly ConfigWiz) from VacNat Software is a neat system-configuration editor for OS/2. OS/2 settings are done by means of a pop-up notebook, but, alas, IBM didn't provide one for figuring out the AUTOEXEC.BAT and CONFIG.SYS files.

System Notebook brings up multiple pages of a notebook and explains what changes you can make. When it's done, it saves it all as new OS/2 CONFIG.SYS and AUTOEXEC.BAT files. If you run OS/2, you need this. I wish the Windows people would come up with something as neat. Recommended.

We expect better Windows video drivers for the Pentium PCI bus. Th e best overall Win Tachometer performance of the Pentium PCI with ATI Technologies' Mach 32 PCI board has been 52, and we can nearly get that on the 486DX2/25 with a VL-Bus Mach 32 board. The fastest system in the house is still SuperCow, the Gateway 2000 486DX2/33 with a Hercules Dynamite Pro video board. Of course, a Win Tachometer rating over 50 is nothing to sneeze at; it wasn't all that long ago I was satisfied with machines that had ratings of 16 to 20. Now, anything under 30 seems intolerably slow. So it goes.

I have reports that Number Nine Computer has already solved the driver problem, and Number Nine PCI boards in Pentium machines are getting truly astonishing Win Tachometer results. We'll have one here soon, and I'll let you know.

While we're waiting for new drivers to take advantage of the Pentium's architecture, we shouldn't forget that the PowerPC chip is rushing upon us. Informants tell me that some software already runs faster on the PowerPC than on the Pentium; some of it much faster. All of which means that hardware has again leaped ahead of software, and, as the software people catch up, we'll see more spectacular improvements in what these machines can do. But by the time we get jaded with those new improvements, we'll have even-better hardware.

I can remember when a 10-MB hard drive cost several thousand dollars. CD-ROM drives used to be old and slow and cost a thousand dollars. The street price of my 1-GB DEC DSP3107 hard drive will be well under a grand when you read this, and CD-ROM drives cost less than 8-inch floppy drives used to. Ain't it wonderful?

Despite residual glitches, I don't doubt that you can get a Pentium, ATI Technologies' Graphics Ultra Pro Mach 32 PCI board, and Sound Blaster/Wave Blaster sound boards working together; mine does most of the time, the only problems being with wave sounds in W4WG. The Pentium should be a real killer of a flight simulator and a truly awesome games machine; which means that as the right software is developed, it w ill also be excellent for desktop conferencing, multimedia, and that sort of thing. At least that's what you can tell your boss when you say you have to have one.

One multimedia package we looked at was part of Microsoft's newest Home (as opposed to Office) series, Multimedia Schubert 1.0. This product group also includes Mozart and Stravinsky. It's built around the Quintet in A Major (The Trout). Like the rest of this series, it's really well done, with background information on the composer and his times, and a music discussion that starts simple and gets as technical as I want. I find these really do add to musical enjoyment, and I recommend the series for anyone who is or might want to become interested in classical music.

I have about a zillion other multimedia programs, including Grolier's and Compton's encyclopedias. I suppose I should come up with some way to choose one multimedia program over another, but the fact is they're all good and getting better. If you don't have one of them, you should; there's no point in waiting for better, not that better multimedia packages won't be here soon enough. Really good whizbang sound boards are coming this year, but what we already have is darned good, too.

My advice is to get what's out there. For instance the combination of the Sound Blaster Digital Edge CD Multimedia Kit, a Sound Blaster 16 board with a Wave Blaster add-on board, and a good double- or triple-speed CD-ROM drive is already good enough and will last you the year or so that it takes to perfect the new stuff that's coming later this year. If you keep waiting for the best, you'll miss out on everything.

The Macintosh version of Roberta Pournelle's reading program, The Literacy Connection, is done. She'll be doing the grand introduction at her session of the International Conference on Technology in Education in London later this month.

This version was done in SuperCard. The programmer, Chris Innanen, used the QuickBasic code I did in the mid-1980s and began to convert that to HyperCard. Roberta then decided she wanted a color version, which required that they switch to Silicon Beach's SuperCard. SuperCard was then bought by Aldus and now has been sold to Allegiant Technologies. SuperCard is HyperCard on steroids. The basics are fairly easily learned.

As Roberta's program shows, you can do some pretty complex programs in SuperCard, but that requires actual programming. HyperCard and SuperCard are still the best ways to get involved in Mac programming. Chris reports that if the newest version of HyperCard had been available, he would probably have done the program in that, but the advantages of HyperCard II over SuperCard aren't so great that he's going to convert.

The result is a script-driven engine that lets Roberta change the lesson plans without extensive programming. I've talked about The Literacy Connection before. Like the PC version, the Mac version requires a literate person to read the screens to the student, but the "instructor" doesn't have to be a teacher because all the instructions are at fourth-grade level. In one school, a 10-year-old uses the program on an Osborne luggable to teach reading to first-graders. The next version of The Literacy Connection should be self-pronouncing and thus do away with the need for an instructor.

We don't know of anyone who has gone through all 70 lessons and emerged unable to read English; we know it has worked with many students diagnosed as dyslexic or learning-disabled, as well as with normal and gifted students. It unabashedly teaches systematic phonics. In our judgment, to be able to read implies the ability to read any English word, whether or not you have seen it before, thus instantly making your reading and speaking vocabulary identical.

The Mac version is much changed from the PC version I programmed several years ago. It has color and animations. Unlike much software that tries to be educational, The Literacy Connection is not a game. An upcoming version will incorporate a game as a reward for getting the lesson right. Kids who've seen the test version of the game go bonkers for it, but it's still the case that games are more to impress publishers and parents than kids; children find that actually learning to read is a pretty rewarding experience all by itself.

We've had a fairly hefty demand for an Apple II version of The Literacy Connection. Now that schools are converting those Apple IIs to Macs, the pressure has been on to get a Mac version done, and Roberta regrets that it took a lot longer than she thought it would. She's polishing up the manual as I write this.

I know I write about her program a lot, but it is my considered opinion that she's got the best reading instruction program in existence; and if kids can't read, they aren't going to learn much else in school. Anyway, this is available, it works, and I'm proud to be able to recommend it.

I was hoping the game of the month would be SimCity 2000 by Maxis, but I can't seem to install the DOS version. This is twice now, but I'll keep trying. Meanwhile, I've gone back to an ancient game called Sword of Aragon from SSI. It still plays well. Another old-timer I like is QQP's Battles of Destiny.

Maxis also publishes SimHealth, which is a model of health care. Like all models, many of the conclusions are built into the model's assumptions. This is an excellent tool to stimulate discussion; the danger lies in using such simulations as real models of the real world. With that caveat, I can recommend the program to those interested in health-care policy.

The book of the month is by Richard E. Cytowic, M.D., The Man Who Tasted Shapes (Tarcher/Putnam, 1993). You may never have heard of synesthesia, a sort of scrambling of senses that might cause you to smell in colors or, as the title suggests, taste shapes; but it's very real, and this entertaining book tells of some experiences the author has had with such people, using their experiences to inquire into the nature of sensation and perception. I guess that sounds more like a book reviewer than me, which means I'm getting tired. Anyway, you'll like the book. The most useful computer book this month was Brian Livingston's Windows 3.1 Secrets (InfoWorld Technical Books, 1992).

Next month: more on Percival. With luck, we'll get him running NextStep, OS/2, and Windows NT so we can do some comparisons.

For More Information

Available now for the Mac, The Literacy Connection ($100 postpaid, California residents add $8.25 state tax) is the best reading instruction program in existence. Contact The Literacy Connection, 12190 1/2 Ventura Blvd., Suite 372, Studio City, CA 91604.

ATI Technologies' MediaMerge ($299) multimedia editor for sound, text, and graphics is as good as any editor we've seen. Contact ATI Technologies, Inc., 33 Commerce Valley Dr. E, Thornhill, Ontario, Canada L3T 7N6, (905) 882-2600; fax (905) 882-2620.

Microsoft's Multimedia Schubert 1.0 ($79.95) really does add to musical enjoyment. Contact Microsoft Corp., 1 Microsoft Way, Redmond, WA 98052, (800) 426-9400 or (206) 882-8080; fax (206) 883-8101.

The Plextor Double Speed Plus ($329) CD-ROM drive works well with my multimedia system configuration. Contact Plextor, 4255 Burton Dr., Santa Clara, CA 95054, (800) 886-3935 or (408) 980-1838; fax (408) 986-1010.

SimHealth ($39.95), a model of health care, is an excellent tool to stimulate discussion for those interested in health-care policy. Contact Maxis, 2 Theater Sq., Orinda, CA 94563, (800) 336-2947 or (510) 254-9700; fax (510) 253-3736.

SmartCache III SCSI host adapters (from $285 to $595) are a family of solid, trouble-free SCSI caching hard drive controllers. Contact Distributed Processing Technology, 140 Cadence Dr., Maitland, FL 32751, (800) 322-4378 or (407) 830-5522; fax (407) 260-5366.

The Creative Labs multimedia sound boards tested on my Pentium system include the Sound Blaster 16 SCSI-2 ($279.95), the Sound Blaster Pro Deluxe ($179.95), and the Wave Blaster ($249.95). Contact Creative Labs, 1 901 McCarthy Blvd., Milpitas, CA 95035, (800) 998-5227 or (408) 428-6600; fax (408) 428-6611.

Logitech's SoundMan 16 ($199) sound board works just fine on my Pentium system. Contact Logitech, 6505 Kaiser Dr., Fremont, CA 94555, (800) 231-7717 or (510) 795-8500; fax (800) 732-3102.

With SuperCard ($299), you can do some pretty complex programs on the Mac. Contact Allegiant Technologies, 6496 Weathers Place, Suite 100, San Diego, CA 92121, (619) 587-0500; fax (619) 587-1314.

System Notebook ($24 postpaid) is a neat system-configuration editor for OS/2. If you run OS/2, you need this. Contact VacNat Software, 1370 White Oak St., Harrisonville, MO 64701, (816) 887-2928.

PC Power & Cooling demonstrated their high-end PC cooling technology by building a Pentium system, which included a tower case ($349), a line-conditioning power supply ($229), the Pentacool ($39), and the 110 Alert Plus ($49). Contact PC Power & Cooling, Inc., 5995 Avenida Encinas, Carlsbad, CA 92008, (800) 722-6555 or (619) 931-5700; fax (619) 931-6988.

---

Jerry Pournelle holds a doctorate in psychology and is a science fiction writer who also earns a comfortable living writing about computers present and future. Jerry welcomes readers' comments and opinions. Send a self-addressed, stamped envelope to Jerry Pournelle, c/o BYTE, One Phoenix Mill Lane, Peterborough, NH 03458. Please put your address on the letter as well as on the envelope. Due to the high volume of letters, Jerry cannot guarantee a personal reply. You can also contact him on the Internet or BIX at jerryp@bix.com .