DTACK GROUNDED #4
November 1981
DTACK GROUNDED, The Journal of Simple 68000 Systems
Issue # 4 November 1981 Copyright Digital Acoustics, Inc
DTACK GROUNDED, The Journal of Simple 68000 Systems
Issue # 4 November 1981 Copyright Digital Acoustics, Inc
This month we are going to sort of ramble on, rather than try to pursue a coherent sequence as is our usual wont.
We have received our first inquiry from the Better Business Bureau. We received a nice letter stating that an inquiry had been received about DE-TAK Grounded at our address, along with a two sided form to be filled out. We glanced at the second page of the form, saw where the color of the eyes of the third assistant janitor's maternal grandmother was required, and round filed the form. Then we examined the accompanying letter to determine the nature of the complaint. Apparently someone, seeing that there was no phone number in our ad(s), thought we might be running a scam (we are: we are fitting Apple and Pet owners with a 68000 for about a tenth the price of what is available elsewhere).
So, we decided to phone the BBB to discuss the situation. We examined the letter carefully. It listed the complete mailing addresses of four BBB offices in the Southern Calif. area, but NO PHONE NUMBER. We pulled the form out of the round file: NO PHONE NUMBER. We pulled the envelope out of the round file: NO PHONE NUMBER. We have bad news for whoever complained to the BBB: that organization does not provide its phone number with its correspondence! (We did get the phone number from information and determined that there was no complaint, merely an inquiry from out of state.)
Happily, no one has called the BBB's attention to pages 2 thru 5 of the last newsletter. We could be arrested for impersonating a newsletter editor.
This bottom half of the front page is the LAST portion of this issue of the newsletter to be written. This requires some explanations:
Two VERY IMPORTANT EVENTS for persons interested in the 68000 were announced in the press on Oct. 12. We have therefore published a special edition of REDLANDS to cover these events in depth. And we have moved the publication date up about 10 days to get this news to you soonest.
About that promise to publish a picture of the Apple II compatible production prototype in this issue: there isn't going to be a production prototype. Since only three I.C.s on the main board differ from the Pet version, we have ordered a short production run instead.
Once again we repeat our request for input from you, the reader of this newsletter. What material are we covering that is NOT of interest? What are we NOT covering that we should be? We asked for opinions on the half size format of newsletter 3's REDLANDS. Regrettably, we printed the request IN the half-size edition!
EVERYBODY IS OUT OF STEP BUT DTACK GROUNDED AND MOTOROLA! We are no longer the world's only purveyor of simple 68000 systems. Motorola has finally found a use for the warehouses full of 4 MHz 68000s it has on hand. Having discovered that absolutely nobody was interested in paying $105 for the 4 MHz product, they have marked the part down to its true value of 13 cents (for internal accounting purposes only) and are building a small 68000 demonstration board for educational purposes. The emphasis is on learning how the 68000 works, not speed. Communication with the host is via RS232. The board will be priced at about $500 and will be produced about the first quarter of 1982. The part number is MEX68KECB.
When the board becomes available, it will be covered in this newsletter in much greater detail. In the meantime us folks at DTACK GROUNDED are very relieved; at least SOMEBODY else is planning a simple 68000 system that attaches to a host computer. And that makes us just a bit more credible.
By purest coincidence, the technical honcho at the local Motorola office visited our facility in early August. At that time, we demonstrated our prototype board and our (then) crude "hand assembler's helper" which assembled one instruction at a time. The brief press release of the new Motorola board mentions a remarkably similar function: "These functions operate on a line-by-line basis and the source programs are not stored in memory." Great minds think alike?
HAVE YOU SEEN THE LATEST ISSUE OF BYTE MAGAZINE? Aside from the danger of developing a hernia trying to lift it, the second most remarkable thing was the large number of 68000 systems. Now, from the point of view of an end user looking for a 68000 system that will be well supported, the very best thing that could happen is that two, and two only, competently run companies completely dominate the 68000 market.
That would provide the competition that is needed to keep the 68000 vendor on his toes, and yet provide only two target machines for the software writers. As the TRS-80 and Apple II saga has proven, the best way to get lots of software is to have lots of machines available for the software writers to sell their product to (a chicken and egg situation to some extent).
But what seems to be happening is that there is going to be a HUGE number of companies trying to sell 68000 based systems. It is very unlikely that any one (or two) company(s) are going to be able to dominate the market. Hewlett Packard had the best chance, but took themselves out of the running by using a 50 (fifty) column CRT. And by their traditionally high price. This means the software vendors are going to be faced with a bewildering array of machines. Unfortunately, this will turn off a lot of potential software writers. There is, however, a solution to this problem.
Very simply, the software writer should write his programs to be as I/O independent as possible. After all, 68000 code is 68000 code and will run on any 68000 processor. The problem is the 40 column CRTs (Apple and Pet), the 50 column CRT (HP), the 52 column CRT (Osborne)... that is where all the grief is in connecting the software to the target 68000 system. Then he can, with a little extra work, sell his software on a number of different 68000 machines. Some of these machines will be Apple IIs and some will be Pets...
If the above paragraph looks familiar, you're right. We put the same suggestion in newsletter #2. What is really fascinating is that, with the 68000 market as fragmented as it is, DTACK GROUNDED could possibly wind up two years from now as the leader, based on the total number of 68000 systems installed. After all, if the 68000 processor is such hot stuff at $8000, isn't it even hotter at about $1000 (memory size dependent) for the half million owners of Apples and Pets (and the OTHER half million owners of TRS 80s and S-100 systems starting next year)? No, we're not smoking any funny tobacco, but we ARE very optimistic!
FORTH-RIGHTLY SPEAKING: The 68000 is a nearly ideal FORTH engine. There are just two problems. The first is that the 68000 has a 16.7 MEGABYTE address space. According to the true word of St. Moore, only 64 Kbytes may be used since the ONLY method of accessing the CPU's memory is via 16 bit pointers. Now, our basic board has provision for 92 Kbytes of RAM plus an expansion connector for even moore (oops) memory. Fortunately, we can limit our board to 64K by sawing three inches off one end of the board and by sabotaging the memory expansion connector. We will do this for keepers of the true faith.
The second problem, applicable only to heretics with more than 64K, is a peculiarity of the 68000 called "sign extend". The 68000 uses 32 bit address registers, with the upper 8 bits not brought out to the outside world (YET!), Now, FORTH works by fetching a 16 bit word from memory and using that word to point at an address. But when a 16 bit word is loaded into a 68000 address register, the most significant bit is sign extended into the high order 16 bits. That is, MOVE .W $7FFE, A0 results in $00007FFE being placed in A0. MOVE .W $8000, A0 results in $FFFF8000. In a 68000 system with more than 64K, an indirect fetch will NOT fetch the contents of the next word above $7FFE in the 68000's 64K zero page.
we want to do this: but we have to do this instead:
MOVE .W (A1)+, A0 CLR .L D0
MOVE .W (A1)+, D0
MOVE .L D0, A0
This is the price the FORTH user has to pay for having a large memory space available. And since we have two extra, unwanted machine instructions in the innermost part of FORTH, the system will run more slowly than otherwise.
That last paragraph is true for any 68000 system EXCEPT the DTACK GROUNDED boards. Since we think FORTH is a pretty good language to use with the 68000, we decided to provide a hardware fix for this software problem. We do that by decoding the highest 64K page of memory into the same place as the lowest (zero) page. Therefore, address $FF8000 decodes to the same location as $008000. And the FORTH user can forget about the sign extend problem. There is, of course, a penalty: where other 68000 systems can potentially address 16.777 megabytes, our system is limited to 16.711 megabytes.
Small companies like us also occasionally do things correctly.
ARE WE GOING TO SUPPORT FORTH? No. Also, we are not going to support Cobol, Fortran, Lisp, Ada, Pascal, Comal, Sail, Algol and some others. We are not going to support FORTH because about half of the initial purchasers of our board are going to jump on FORTH with both feet. We are not the only ones who see that the 68000 is a natural for FORTH. And we are not going to support these other languages because we are a small company and we don't have the resources.
What we are going to do is support interpretive and compiled BASIC, together with 68000 assembly language and 68000 utilities, such as floating point packages, sorts, FFT applications etc. Although we all know that BASIC has limitations, it is the one language which everyone can use. It is the primary language of our targeted host computers. Since our resources permit support of only ONE high level language, there really is as other choice.
But since others are going to support FORTH, some users will buy our boards to run those FORTH packages. And others, seeing that support exists for interpretive BASIC, compiled BASIC, FORTH and 68000 assembly language and utilities, just might decide to bring up a P-code interpreter. Or maybe implement UNIX. And that is the way that the DTACK GROUNDED board is going to become popular.
Let us place ourselves inside the head of a professional software writer. Last year he wrote a general ledger package for the Apple. This year he wrote a data base management program for the Apple. That being finished, he thinks about writing a text editor for the Apple, and sees that his would be about the 63rd text editor, and might not sell. Maybe an electronic spread sheet? Visicalc got there first.
The fact is that the Apple II is nearing the end of its useful commercial life in terms of new computer sales. New computer sales are the best source of ongoing software sales. And, as we have implied, the software market for the Apple II is getting saturated. So what are we (the professional software writer) going to do?
Simple. We move on to the next machine, one which is going to sell well but which is new and is not saturated with software. The Xerox machine is a vanilla Z80 machine with a slow clock and inadequate disk storage. The IBM machine is an 8088 machine with inadequate disk storage. Since the 8088 is really just an 8080 with a face lift and a new coat of war paint, it is source code compatible with the 8080. When running cross compiled 8080 code, it runs at 8080 speed. And since there is lots of 8080 code already written, the software market for the IBM machine may be pre-saturated.
So what is going to be the next popular product for which the market is not saturated? The answer is, 68000 machines. Not one particular 68000 machine, but the collective set of all the 68000 machines being offered: Computhink, HP, Microdasys, Wicat, Q1, Dual Systems, Empirical Research, Forward Technology, Chrometics, Ragen, Fortune Systems, Charles River Data Systems, TSD and Omnibyte. New 68000 systems are being introduced almost daily. If Apple Computer does not get its fanny in gear, they may have the dubious distinction of introducing the 100th 68000 system.
Now all we have to de is choose WHICH of those systems to write software for. Which will dominate the market? Evidently, none! Like we said twice already, the software writer will have to adopt a strategy which permits sales to many different 68000 systems. Now, all of those companies listed above are making stand-alone systems. To use them one needs disk drives and a printer. The disk drives, at least, will be dedicated to that particular 68000 system. And since they are all expensive, the programmer is going to pay through the nose to get his initial system up and running. Unless, that is, he hears about an outfit called DTACK GROUNDED which is not listed up there because it is NOT a stand-alone system.
What if the programmer discovers that for about $1000 he can get a 68000 board with enough memory to develop useful software for those stand-alone 68000 systems because he ALREADY OWNS the needed keyboard, CRT, disk drives and printer? And that the problem of adapting his software to the DTACK GROUNDED board is a non-existing problem because that's the one he DEVELOPED the package on. Now, do you think this guy is going to adapt his software for the HP, Wicat and Dual but NOT market the version that runs with the Apple II/DTACK GROUNDED board?
Now let us switch inside the head of an Apple II owner (you may know one very well, no?). With two disk drives, a printer, and probable accessories such as a modem and an 80 column CRT board, plus all the software, he has an investment of $5000 or more in his system. If he wants to get up and running on the best of the current generation's processors, he has three choices:
Look, it can easily be proven that the computer industry does not exist. The proof goes this way:
Since there are a whole bunch of computers out there, and another whole bunch of programs to run on those computers, there MUST be something wrong with that proof. The fallacy is simple: the proof assumes that people will base their actions TODAY on conditions which exist TODAY. That is patently false. The first guy to buy an Apple II did not buy it because of all the software support that is out there, because NONE of that software existed when the first Apple II was sold. No, the first buyer probably thought the machine had some neat features (which it did, for its day) and that other people would buy, also, and that some of the buyers would start developing a software base. It turns out that the first buyer of the Apple II was right.
On the other hand, the first buyer of a Sphere or Interact was probably wrong.
Let us examine the possible ways a new product in the personal or small business area can crash and burn, and our reasons why we don't think that will happen with DTACK GROUNDED:
LOTS OF NEW COMPANIES ARE STARTED BY HOBBYISTS WITH NO BUSINESS SENSE OR BACKGROUND. We have been in business under the same management since 1973. The necessary business infrastructure already exists.
A LOT OF THE GUYS GETTING INTO THE PERSONAL COMPUTER MARKET WERE MAKING FOUNTAIN PENS LAST YEAR. We have been building microprocessor based acoustic noise monitoring equipment since 1973. We have built and sold systems using the 4004, 4040, 6502 and 68000. This writer took delivery, at home and as a private party, of a BASIC computer with about 30K ROM, 4K RAM and a 64 X 16 line CRT in 1972, and still has the invoice to prove it. The computer was a WANG 2200 and this newsletter is being typed on that computer.
EVEN IF THIS THING DOES TAKE OFF, OR especially IF IT TAKES OFF, YOU ARE GOING TO BE KILLED BY THE CASH FLOW. Our initial sales terms, 10% down, balance C.O.D. were chosen specifically to facilitate rapid cash turnover. WE will adopt more conventional sales terms when sales stabilize and/or we are in a strong enough financial condition.
YOUR BOARD CAN'T BE ANY GOOD. EVERYBODY SAYS THE 6809 IS THE PROCESSOR OF THE FUTURE. Motorola is the designer of both the 6809 AND the 68000, and they state that the 68000 at 8MHz has ten, repeat TEN times the performance of a 1MHz 6809.
YOU SAID YOURSELF THERE ARE A HUGE NUMBER OF COMPANIES OFFERING 68000 SYSTEMS. NOW ARE YOU GOING TO SURVIVE IN THAT RAT RACE? Sir, all those other companies are selling stand-alone systems. We are selling a 68000 'attached processor' which is designed to enhance existing host computers. In marketing terminology, our product is strongly differentiated. In survival-of-the-fittest terms, Charles Darwin would say that we have found our own little ecological niche. If you call one million potential host computers little.
YOUR BOARD IS TOO EXPENSIVE. I'M LOOKING FOR A CHEAPER 68000 SYSTEM. Sir, you have imbibed too heavily. Come back when you are sober and we will show you the error of your ways.
THERE IS PRACTICALLY NO SOFTWARE SUPPORT FOR YOUR COMPUTER. We are not selling a computer. We are selling an enhancement to existing host computers which already have strong software support. Our enhancement uses the newest and best modern microprocessor. Because it is the NEWEST, there CAN'T be strong software support for it at this early date. Because it is the BEST, that support WILL be forthcoming.
We have presented all of the objections we can think of to our program and our response to these objections. Our purpose is, hopefully, to improve your opinion of the operation we have undertaken and to increase your "expectation level". The initial sales of ANY new computer or computer enhancement are based on this "expectation level".
We suggest you go back and review the three possible ways an owner of an existing Pet or Apple system can upgrade to a 68000 system. Which makes the most sense?
We concede that the first time computer buyer, especially the small business variety, is going to make a choice between the Xeroxes, IBMs and Q1s. We at DTACK GROUNDED are not trying to sell to those people. We are trying to sell to the person who has one of the older generation machines and is trying to protect his investment by attaching the best modern generation microprocessor.
This newsletter is not about 68000 systems in general, just SIMPLE 68000 systems. Any simple 68000 system will of necessity need to be hooked up to a host computer to be useful. Unless, of course, somebody wants a fast number cruncher to occupy a SMALL corner of an intelligent measuring instrument such as our computerized community noise analyzer. In which case the mere existence of this journal reveals that such small, simple systems CAN be made. And our board makes a nice, cheap development system for such a project.
If you are knowledgeable regarding the architecture of the modern generation of microprocessors, any skepticism you have had was probably based on business factors rather than on the performance of the 68000. We hope that the preceding information about us has dispelled much of that skepticism.
If you are NOT knowledgeable regarding the 68000, and wonder what all the fuss is about, find somebody who IS. Ask a professor of computer science.
University courses on microprocessor architecture these days start out explaining one of the last generation processors (8080, 6800 or 6502) because they are simple and easily understandable. Then they proceed with modern microprocessor architecture and the example they use is the 68000. Not the 8086, or 6809, or Z8000, or 9900, or 8086. They use the 68000 as their example because it is the job of a university to be at the forefront of technology. The 68000 is universally regarded (among knowledgeable people who have no ax to grind, such as university professors) as having the most advanced architecture.
As a divertissement, we now present the wire service report on the 101th 68000 system introduction:
SANTA ANA (dUPe) DTACK UNBOUND has announced its entry into the burgeoning 68000 based small business computer market. The system is utterly unique, having a keyboard, CRT, dual floppy disks and compatibility with several letter quality output writers.
According to Mr. Slick, marketing manager for DU, the new computer is fully supported by every computer language known to mankind, including three which will not be invented until next week. Asked whether the language implementations were standard, Mr. Slick stated, "Our languages ARE the standards. Any differences with other implementations are the result of deficiencies in those other implementations."
DU's new computer is said to be programmable by mental telepathy and to be remarkably user friendly. "We recommend a chaperon for under age coed programmers", stated Mr. Slick.
The wide range of languages supported by the new computer is said to open new vistas in applications programming. "For instance, we have a text editor written in APL, a structural analysis package written in COBOL, a tic tack toe game written in ADA and a general ledger package written in FORTRAN. The inner interpreter of our FORTH is written in BASIC, a truly innovative approach made possible by our brilliant systems programming staff", explained Mr. Slick.
The price of the DU 68000 is said to be "quite reasonable, given its fantastic potential". Delivery is estimated to be "soon".
Now many genuine computer introductions have you seen that could be mistaken for this? Speaking for ourselves, we have seen entirely too many that were remarkably similar. That is why we try to be as factual as possible in reporting our progress. When we tell you that we have done something, past tense, then we have done it! If we tell you we plan to do something, it represents honest intentions at the time of writing. For instance, in our second newsletter we projected availability of a 128K byte expansion board in October. We have not built that board yet for the simple reason that nobody has bought a full house 92K board from us yet, and until that happens we don't need the expansion board.
Which leads us to another subject: Our board DOES have the ability to hold more than 64K. The vast majority of our readers are currently limited to 64K, and many of them are beginning to feel a little squeezed in that space. Well, to go beyond 64K, you have only two choices: get a 68000 system with up to 16.7 megabytes of linearly addressable memory, or join the wonderful world of segmented or bank selected memory. Here is how the 68000 moves 200,000 bytes from location 100,000 to location 1,250,000 (all decimal numbers):
203C 0003 0D40 MOVE .L #200000, D0 set byte count
207C 0001 86A0 MOVE .L #100000, A0 set source address
227C 0013 12D0 MOVE .L #1250000, A1 set destination address
MOVEN 12D8 MOVE .B (A0)+, (A1)+ move a byte
5380 SUBQ .L #1, D0 decrement byte counter
66FA BNE MOVEN repeat until count done
Three instructions to set up, three to execute (if the byte number were always even or divisible by 4, word or long word moves could be used). Now, try THAT using a segmented or bank switched memory system!
INTERESTING PUBLICATIONS DEPT: If you are interested in methods of calculating transcendental functions (LOG, SINE etc.) you absolutely MUST purchase "Computer Approximations", Wiley, 1968. This book is a cookbook of formulas for computing transcendental functions.
If you are interested in the Intel 8087 math processor (which is not yet commercially available), get Intel Application Note 113. If the 8087 is too slow (!!) for you, subscribe to TRW's "Digital Signal Processor" newsletter. P.O. Box 2472, La Jolla, CA 92038. Call (714) 578-5990.
SOFTWARE STATUS REPORT: The "search for line number" problem in long programs mentioned in the last issue has been fixed. As mentioned on page 7, the problem has nothing to do with our board and neither does the fix. The fix is a simple 6502 software routine which permits the Pet to GOSUB or GOTO any line number instantly, with no searching being done. We have offered to publish the details in one of the 6502 magazines to help get publicity for DTACK GROUNDED, so we can't publish the details here - magazines like to have exclusive rights, not reprint rights. But if they don't publish the details, we will. We have no intention of trying to preserve proprietary rights (we did make it clear that you don't need our 68000 board to use this fix, didn't we?).
The greatest benefit of this fix is that writers of long BASIC programs can put the body of their program at the start and the subroutines at the end, where they belong.
We have done several day's work on our 68000 "hand assembler's helper". There are a lot fewer bugs and a lot fewer instructions that are not yet implemented. We estimate about 4 more days work are needed to complete the "hand assembly" aspect of that program. The next task will then be to hook a file structure onto the program to turn it into a true cross assembler. If we are sensible, we will put a substantial part of the code up into the 68000 board so that the assembler can't be used unless our board is purchased. But no other protection method, other than the standard copyright notice, will be used.
HARDWARE STATUS REPORT: We have just begun shipping production Pet/CBM compatible boards, with solder mask and silk screen. It works, and works very well, and at 8MHz with a 16 bit data bus and a 24 bit address bus. What more can we say except, if you own a Pet/CBM, how come you haven't bought one?
For you Apple II types, we have just ordered a short production run of Apple II compatible boards. Our next newsletter will carry a message for Apple owners pretty similar to the one above for Pet owners.
ACKNOWLEDGEMENTS: Our spies tell us that Pet and CBM are still trademarks of Commodore and that Apple and Apple II are trademarks of the Apple Computer Company. Apple III is also a trademark of the Apple Computer Company but they don't brag about it much lately.
SUBSCRIPTIONS: $15/6 issues U.S. and Canada. $25 U.K. Payment should be made to DTACK GROUNDED. The subscription will start with the first issue unless otherwise specified. The address is:
DTACK GROUNDED
1415 E. McFADDEN, St. F
SANTA ANA CA 92705
This month we have a SPECIAL EDITION of REDLANDS, printed on yellow paper. Next month we will continue discussion of our floating point package covering the rounding and adjustment of the exponents for the 68000 multiply and divide routines, and begin the discussion of the floating point add routine.
WELCOME TO (SPECIAL EDITION) REDLANDS!
As you can see, redlands is yellow this month. For those who are interested in both 16 bit processors in general and the DTACK GROUNDED (or other) simple 68000 boards in particular, two VERY IMPORTANT EVENTS have just occurred. For one, the overcrowded 16 bit microprocessor market has shaken out with just two winners emerging. More on this later, with lots of background info.
But the most important and exciting event for us folks at good old D.T. is that Apple is re-introducing the Apple III (please to not throw up on our rug, this is GOOD news!), this time with a miniature Winchester replacing the floppy disk drive. In case you don't know WHY that's good news, stick around and we will reveal all!
An Apple III with a minnie winnie is sure to be marketed in the $4000 - $6000 price range. This means that the forthcoming Apple 68000 entry (reportedly codenamed Lisa) will be priced at $8000 or more. This is good for US for two reasons: Lisa will be just another (yawn) entrant in the race to build $8000 and up 68000 systems and will not dominate that market. Secondly, our 68000 board is going to look VERY attractive to Apple II owners who do not have $8000 in spare change.
Let us confess that a recurring nightmare we have had involves the following scenario: Apple writes off the Apple III, freeing the "more than Apple II" market for a totally new product. Then Apple introduces Lisa with an 80 X 25 CRT and 64K (initial) RAM to go with the 68000. If that product were priced at $2500 (sans disk), it would be, shall we say, VERY competitive with certain other recent entrants into the small computer systems market.
At that price, our board would not look that attractive to the Apple II owner. He would probably figure to run his II as far as possible and then buy a Lisa when he really HAD to have 68000 performance.
But we can forget that nightmare at last. From the bottom of our heart, we give thanks and blessings to the executives at Apple Computer for re-introducing the Apple III (look, fella, we already warned you once about barfing on our rug).
WHY does Apple continue to support the III instead of writing it off? We don't know, of course. But we DO have a guess we would like to share with you. For Apple legal eagles, we emphasize that this is a GUESS, based on rumor and such.
Apple "went public" with a stock offering last November, with about a 100 - 1 price/earnings ratio and also an exorbitant ratio of paper value to book value. These ratios were so high that the New Jersey commissioner of Corporations forbade the sale of Apple stock in his state. The stock offering was beautifully timed, just AFTER the announcement of the Apple III and just BEFORE the technical problems (and very non-technical lack of sales) became evident.
The only way you can sell anybody stock at 100 - 1 price/earnings is if you can convince people that the management of the corporation can walk on water without getting the bottoms of their feet wet. In less than a year, Apple executives have descended to the point that they show up at work in scuba diving gear. If Apple wrote off the Apple III, which is the only product of PUBLIC Apple, the price of the stock might not stop dropping until it got into negative numbers. As it is, Apple stock was traded at $27 the first day it went public, and has been down to $14 in the last month.
Apple has therefore been pretending that all is well with the III, more software is on the way and LOOK OUT FOR THE COMING EXPLOSION IN APPLE III SALES! Kids call this "whistling past the graveyard". As a privately held company Apple could get away with this indefinitely. But for PUBLICLY held companies, there is that anti-fraud device known as an audited annual financial report.
Now, that was a very bad word we used at the bottom of the last page (fraud). But if a publicly traded corporation reported that it had a value greater than was factual, that would be fraud. And publicly traded companies have every incentive to report a value as great as is possible. The potential for fraud (there's that bad word again) is so great that our much-maligned politicians long ago passed laws requiring that publicly traded companies release an INDEPENDENTLY AUDITED annual financial report. The FIRST such report is due from Apple Computer in November, 1981.
When Apple went public a year ago, it received an ENORMOUS infusion of cash. Corporations do not hide cash in the mattress or bury it in the back yard. If the corporate executives were confident that they could walk on water without getting their feet wet, they just might invest a portion of that money in themselves, so to speak, by stockpiling completed computers in numbers that could not be justified by (then) CURRENT sales on the theory that the sales were, of course, going to come.
Now let us suppose that there are 30,000 completed Apple III computers sitting in warehouses. That would represent a book value of about fifty million dollars. Once again, we emphasize that this is a supposition.
When preparing audited annual reports, it is up to the corporation (short of obvious criminal fraud) to decide whether any particular inventory should be carried at full book value, partially written off, or completely written off. On the other hand, the auditors must place an opinion at the end of the financial report. Since corporate officials are understandably reluctant to be stuck with an unfavorable opinion, the preparation of the financial report for a large company usually involves some last minute negotiations with the auditors. This is all correct and proper, and results in much more realistic financial reports than would otherwise be the case.
Let us see how our hypothetical fifty million dollars' worth of unsold Apple IIIs would affect the audit process. Although auditors are not technological experts, they know that computers, like tomatoes, depreciate rapidly. For an inventory of completed computers to have value, the sales rate must be high compared to the inventory. To this situation, this is patently NOT the case.
Apple computer would be confronted by three choices:
According to our FICTITIOUS scenario, this is how the Apple III came to be re-introduced (easy there, fella). And by an ABSOLUTELY REMARKABLE coincidence, the re-introduction takes place only a couple of weeks before the first real-life auditor's opinion is to be placed on Apple Computer's first annual financial report as a publicly held corporation.
We figure Apple can find about 1300 people dumb enough to buy a computer with no realistic backup for the data on disk. This means we can write another purely hypothetical scenario a year from now over the continuing battle between the auditors and 28,700 unsold Apple IIIs with miniature Winchesters.
Now that we have assured that DTACK GROUNDED will not appear on Apple Computer's Christmas card list, we can talk about the OTHER recent important happening. Up until last week, the 16 bit microprocessor race was between contenders 8086/8, Z8000 and 68000 with some others on the outside looking in. What happened last week was that the Z8000 crashed and burned.
The choice of a 16 bit processor is now very clear, the 8086/8 or the 68000. Since the two devices are vastly different, the choice should be easy. And since both devices have application areas in which it is the logical choice, this means that BOTH the Intel and the Motorola device have won the race for 16 bit sockets!
When Intel decided to build a 16 bit microprocessor, it made the deliberate decision to build the device so that EVERY 8080 instruction could be executed on the new device (after having the source code cross-compiled). This meant that the new device had to have a calculator-like set of registers like the 8080 rather than a more modern architecture similar to the famous PDP-11 series. And this obviously limited the performance that could be obtained.
On the other hand, the simpler 16 bit device would have a smaller area as a result of having a simpler internal structure. Intel was therefore able to bring its 8088 to the marketplace two years prior to the 68000 with its much more complex internal structure and much larger chip area. The 8086 SHOULD therefore be a full two years ahead of the 68000 in terms of software and hardware development. In fact, the lead is ONE year at most.
What happened is that Motorola, either accidentally or intentionally, developed the art of lying about delivery dates to previously unheard of heights. In the time period leading up to the introduction of the 8086, Motorola was promising delivery of the 68000 in a mere three months. And since the most ignorant and naive of microprocessor users can easily see that the 68000 has a tremendous performance advantage vs. the 8086, most of us (yes, us) waited for the 68000. And waited. And waited.
(The situation was complicated by the fact that another contender, the Zilog Z8000, was coming down the track and wound up actually being delivered a year after the 8086 and a year before the 68000.)
This left the 8086 stuck in the starting blocks. But nobody ever said Intel gives up easily. They continued to develop the chip, coming out with ever-faster clocks, an 8 bit bus version (the 8088), and they announced (and actually began development work on) the most advanced math processor ever dreamed of, for use as a co-processor.
Just about the time most of us were getting terminally fed up with Motorola's ever-sliding 3 month promised delivery, along came the Z8000 in an "engineering sample" version that ALMOST worked properly, and was in fact close enough for some commercial products. And the Z8000 ALSO proved to have significantly better performance than the 8086.
So these of us who are performance-oriented (and the sixteen bit market is a performance-oriented market) stood around kicking the tires on the Z8000 while wondering if those lying S.O.B.'s at Motorola were EVER going to deliver.
Right about this time, some bright marketing type with a little technical knowledge made an interesting discovery: although 8080 code cross compiled into 8086 code ran at 8080 speed (whoopee), there was a lot of 8080 code around. And it DID run. And if you built an 8088 system, maybe S-100 based, for this slow code to run on you could advertise "HOTCHA!! STEP RIGHT UP AND GET YOUR SIXTEEN BIT PROCESSOR! GENUINE NEXT-GENERATION PERFORMANCE, FOLKS!!" even though the unit was just idling along at 8080 speed.
This seems to have attracted the attention of an outfit which has a small amount of marketing expertise, called IBM. So talks opened between IBM and Intel. One of the subjects raised by IBM was, "Who is your domestic second source?". Intel at that time had a THEORETICAL "dss", one that was signed up but was not producing chips. Outfits like IBM and DOD are not particularly thrilled with dss's which are not producing working parts. Just about the time the discussion over this was beginning to heat up, Intel even lost its theoretical dss and was stranded on top of the mountain dressed only in BVDs and with a cold, cold wind blowing.
How all this came about, and what followed, constitutes:
Intel has never been philosophically attuned to the idea of a second source. They figure "we spent the money to develop it, we should get the profit when we sell it". But they also recognize the way the marketplace works, so they looked around for a dss when the 8086 was introduced.
Originally, the most logical second source for the 8086 was Advanced Micro Devices, a competent organization that had second sourced the 8080. When discussions began between Intel and AMD, Intel apparently demanded cash payment for the 8086 mask set in an amount equal to the gross annual product of Norway. Whether the result was intended by Intel or not (AMD is a GOOD second source and thereby BAD for Intel) AMD rather angrily refused and wound up as a second source for the Zilog Z8000. The parting left bad blood on both sides, particularly on the side of AMD. Bridges were burned.
Having bombed with AMD, Intel and Mostek opened discussions. Mostek also did not want to fork out the enormous amount of cash wanted by Intel for the mask set. But Intel wanted a lot less money for the logic design data. The Mostek management decided that their brilliant design engineers could reconstruct the 8086 from the logic design data. One can assume that the brilliant design engineers were NOT consulted. This left Intel happy because it had a domestic second source who was likely a long, long way from producing actual AND COMPETING microprocessors.
There now comes an interlude in which Intel and its customers were both satisfied, for different reasons, with the second source arrangement with Mostek. During this interlude both the Z8000 and, finally, the 68000 became real working products which could be purchased over the counter.
And the folks at Motorola developed a very strange, from Intel's vantage, attitude toward second sources. If you walk into Motorola with $1.50 in cash, proof that you own an oven, and are willing to sign a contract to develop a specific 68000 peripheral chip with a modest penalty clause, Motorola will hand you both the specs on the 68000 and the mask set. As a result, everybody who was not already developing their own 16 bit processor or signed on as second sources for other processors joined up with Motorola. This trend accelerated noticeably when the 68000 became a real part.
It was at about this point that Intels' larger customers began to notice that Mostek still hadn't produced working parts and didn't seem close to doing so. They started placing pressure on Intel to do something about it. Now, YOU may believe that what happened at this specific point in time was accidental, but WE don't believe it. We think that a very smart and devious Motorola executive spotted Intel's DSS VULNERABILITY and yanked the rug, the dss and everything else but their BVDs out from under Intel.
What happened was that Mostek decided it didn't want to continue designing its own mask set for the 8086. They instead dropped that project and sighed up with Motorola and the 68000. One imagines that the terms were made especially favorable for Mostek by our hypothetical devious Motorola executive.
The largest customers for the 8086/8 snapped to attention and pointedly explained to Intel, using short pithy sentences constructed of Anglo-Saxon words, that a GENUINE domestic second source was an absolute necessity, NOW!
Intel gave in and began to look around. And boy, were they surprised! Fairchild, National, T.I. and Zilog were pursuing their own devices and everybody else was signed on with Motorola and the 68000. Everybody but one.
The name of the ONE company not in the above categories was AMD.
We are sure that the reader can guess what happened. AMD abandoned Zilog and signed up as the dss for the 8086, mask set and all. Now we find that VERY HARD TO BELIEVE given the past history of the relationship between the two companies. But Electronic News devoted about 50 column inches to the story in their Oct. 12 issue, so it must be true.
The real question is, how did the deal come about? We see the services of a very important third party (or two) travelling between the principals of the two companies smoothing ruffled feathers. This used to be called "shuttle diplomacy". An additional fact is that both companies needed to make a deal.
Intel was obviously hurting very badly for a viable second source. But AMD had finally discovered that there was room for only TWO 16 bit processors, the slow simple type compatible with last generation software (via source code cross-compilation) and the highest performance part. And the Z8000 which AMD had been second sourcing was neither of these. So, the deal was made.
But we still can't believe it.
The reason we said earlier that the Z8000 "crashed and burned" was A) they have lost their second source, B) Zilog as a company has consistently lost HUGE sums of money staying in business, and even parent Exxon is going to get tired of that one of these days and C) the marketplace has decided that there is only room for TWO 16 bit micros.
It is a shame, because the Z8000 was clearly the SECOND best available 16 bit microprocessor. But that is apparently not good enough. And nobody will want to buy second best microprocessors from a financially unstable company with no second source. The Z8000 thus becomes the first 16 bit entrant to drop out of the race. It will NOT be the last.
About twelve years ago, T.I. was looking around for some products it could profitably build using its own integrated circuits. It was noticed that T.I. purchased lots of minicomputers, and that minicomputers used lots of integrated circuits. So, having itself as its own guaranteed captive customer, T.I. did a survey to find out what the machines were being used for, and discovered almost all of the minicomputers in house were being used as process controllers.
So T.I. came out with a very fine programmable process controller which it called the 900 series minicomputer. They sold a lot of these, almost all to themselves.
Process controllers need to be very good at "bit twiddling" so they can turn relays on and off. They also need to respond quickly to interrupts. The bottleneck in responding to interrupts is saving the internal status of the CPU, which in a 16 bit CPU means, primarily, the registers. T.I. solved this problem by placing the registers in the general memory, so that only the pointer to the registers needed to be saved. So the 900 series was an excellent programmable controller.
Having the registers in general memory proved a severe handicap in the use of the device as a general purpose computer. Most instructions became "two-address" instructions, and the system proved memory inefficient and also slow when used as a general purpose computer. Despite this, T.I. did develop, over the years, an extensive amount of software for the machine.
So when the time came that T.I. decided to build a 16 bit microprocessor, they naturally decided to duplicate the 900 series in silicon because of all the software they already had developed. And that's why the 9900 microprocessor is slow and memory inefficient. It is really a programmable controller which T.I. has tried, with little success, to market as a microprocessor.
Seeing that the 9900 wasn't selling, T.I. had to decide whether to swallow its pride and second source the 68000 or to build an improved version of the 9900. Naturally, they decided to build the improved controller, which they will try convince you is REALLY a microprocessor.
National Semi is ON AVERAGE a VERY well run, competent company but its microprocessor department drags the average down. Remember the proverbial guy who is always a day late and a dollar short? National in the past has brought us the famous IMP-16, the PACE, the SCAMP (and maybe a SUPER PACE somewhere in there) and NOW (would you believe one day maybe soon?) the 16032 which will naturally add to the glory of National along with those other world-beaters.
In fairness to National, the 16032 project was begun when it was generally believed that there would be room for as many entrants in the 16 bit market as there were in the 8 bit market. It has just recently become apparent that this is not the case.
The only chance for the 16032 to survive is to knock off either the 8086 or the 68000. We think that it is too little and way, way too late for any possibility of that happening. Like we said, a day late and a dollar short.
The 9445 is the successor to the 9440 "micro-flame" which proved slower than even the T.I. 9900 and thus "flamed out". The 9440 and 9445 are attempts by Fairchild to capitalize on the very large amount of software available for the Data General line of minicomputers. They seem serious about bringing out the 9445. Since the 9445 is a bipolar part and they are talking about 24 MHz (!!) clock rates it should be capable of good throughput provided you can find an affordable memory chip fast enough to keep up.
It should be mentioned that to be compatible with the Data General series minis, the device is stuck with the D.G. instruction set and internal register configuration (the 8086 problem all over again). The D.G. instruction set is less than wonderful and there is only 25% as much internal register space as in the 68000. This will not be a problem to the probable target customers for this device: existing Data General minicomputer users who want to upgrade their old computers to more modern parts without giving up their existing software.
And that's 30, folks.