The Inside Story of Texas Instruments’ Biggest Blunder: The TMS9900 Microprocessor
If you operate a pc or computer, the possibilities are it has a microprocessor from the Intel 808x line, regardless of whether or not it’s a Windows gadget or a Mac. The utter dominance of those Intel microprocessors is going back to 1978, while IBM selected the 8088 for its first non-public laptop. Yet that choice changed into a ways from apparent. Indeed, a few realize the records assert that the Intel 8088 was the worst amongst numerous feasible sixteen-bit microprocessors of the day.
It was now not. There was a extreme alternative that became worse. I recognize this because I was in charge of the company within Texas Instruments that developed it: the TMS9900. Although this dog of a chip went on to be used in the global’s first 16-bit domestic laptop, you’ve probably never heard of it. As they say, records are written using the winners. This particular bankruptcy of history is thrilling not only for TI’s chip but also for any other also-ran, the Motorola 68000, which turned technologically advanced to the Intel 8088 and the TMS9900. And but the 68000 did now not turn out to be inside the IBM PC. Here is the inside tale of the way IBM came to choose an inferior chip, TI birthed a loser, and Motorola’s seeming winner misplaced, too.
Wally Rhines Photo: Texas Instruments
Chip Pitch: In 1978, the author gave a presentation on TI’s TMS9900 sixteen-bit chip, which became considered for the IBM personal pc. I did no longer get the agreement. I joined Texas Instruments in 1972, clean out of graduate college, and about two years later, I discovered myself giving a presentation to Jack Carsten, the manager of TI’s MOS Division in Houston, in which the employer’s work on metal-oxide-semiconductor chips become primarily based. As a young engineer, I became really intimidated by using Jack, who sat along with his feet at the conference room table for the duration of my presentation, smoking a cigar and muttering “bull****” whilst he disagreed with something stated.
At the time, the “Big three” semiconductor organizations—Fairchild, Motorola, and TI—were suffering to make the transition from bipolar incorporated circuits to metallic-oxide-semiconductor circuits. MOS chips required a substantial one-of-a-kind layout and procedure technology from bipolar chips, and chip startups like Intel moved an awful lot quicker than the established agencies. Of the Big 3, TI had finished the pleasant job of making the transition, thank you in huge part to human beings like L.J. Sevin [PDF], who left TI in 1969 to shape Mostek and later have become a task capitalist. Carsten, who had formerly served as preferred manager of TI’s worthwhile transistor-transistor logic (TTL) circle of merchandise relatives, turned into a key participant in making a move to MOS.
TI’s MOS Division had accomplished its maximum first-rate fulfillment with good judgment chips for the emerging handheld calculator marketplace. Although the company had competed with and eventually beaten Intel to increase the primary fashionable-purpose microprocessor, TI engineers didn’t actually pay a good deal of interest to the Intel four-bit 4004 or 8-bit 8008 microprocessors. I did pay attention to Intel’s 8080 and subsequent 8080A 8-bit microprocessors, which confirmed a great deal greater promise than the 4004. The MOS department becomes given the task of catching as much as Intel in each microprocessor and DRAM (or dynamic random-access reminiscence, that can cram in greater reminiscence cells in line with chip than can static RAM, however, has to be continuously refreshed to save you data loss).
And as a consequence, a method for fashionable-motive microprocessor development emerged at TI. From this method emerged the plan for the TMS9900. The key assumptions at the back of the method had been that utility software program might force the evolution of these chips and that with a successful line of MOS ICs, TI would be in a function to expand an enterprise general for minicomputers, defense systems, and purchaser products, all of which were unexpectedly growing groups for the employer. But to do that, TI might need to leapfrog the cutting-edge 8-bit kingdom of the artwork, represented through Intel’s 8080, and be first to market with a 16-bit structure.
Intel 8088 die Image: Intel
The Winner: Though the Intel 8088 microprocessor changed a long way from perfect, IBM chose it for its private pc, which launched in 1981. I had already established its pc prowess inside the supercomputer race of the late 1960s. Driving that race had been oil businesses in search of a aggressive benefit in 3-D seismic evaluation for oil exploration, which turned into TI’s founding commercial enterprise. IBM, Control Data Corporation, and others competed in this race, but TI turned into the primary marketplace with its Advanced Scientific Computer [PDF].
So for TI, selecting a chip structure for the sixteen-bit microprocessor become straightforward. I had a approach of “one enterprise, one computer architecture,” which aimed to take advantage of any synergies among the employer’s disparate divisions. TI’s Data Systems Division had already launched a circle of relatives of TTL-based minicomputers to be used in Ramada Inns throughout the USA. So the TMS9900 might use a chip structure very much like that of the TI minicomputer.
Carsten’s team knew that the improvement of the TMS9900 and a bipolar version for the military marketplace called the SBP9900 could require time and that the chips probably wouldn’t be equipped till 1975 or ’ seventy-six. In the interim, the MOS Division needed to act. The plan is to start via copying the Intel 8080A to get something into the market, then increase a TI-authentic 8-bit microprocessor structure (called the TMS5500) and finally circulate to the sixteen-bit TMS9900. (National Semiconductor had already launched a sixteen-bit widespread-purpose logic chipset known as the IMP-16, but it in no way did lots of popularity due to its more than one chip.)
The TMS9900 had its fair share of development demanding situations and delays, but it changed into subsequently prepared in 1976. Even then, it faced numerous massive troubles. First, there had been no compatible sixteen-bit peripheral chips. Without peripheral chips to deal with communications and garages, the microprocessor would be worthless for device designs. The 2nd hassle was that the 9900 architecture, the same as that utilized in TI’s minicomputers, had the most effective 16 bits of logical cope with space, which changed into the same as the 8-bit microprocessors of the day. This hassle couldn’t be solved without growing a whole new architecture. The very last problem turned into that while TI should use a single microprocessor era for its minicomputer, defense, and semiconductor companies, competition in the one’s agencies might be at a downside if they adopted the TI microprocessor architecture of their merchandise.
TI9900 Die Photo: Computer History Museum
The Loser: Among the essential troubles plaguing TI’s, TMS9900 turned into a lack of 16-bit peripheral chips, which rendered it vain for device designs. To attack the shortage of sixteen-bit peripherals for the TMS9900, TI engineers landed on an innovation. Why no longer put an eight-bit port on the TMS9900 so that the large quantity of existing peripheral chips designed for eight-bit microprocessors could make paintings with it? I’m sure the concept sounded affordable at the time. The result becomes the TMS9980, which emerged in 1977. The attachment of an 8-bit peripheral to a sixteen-bit microprocessor negated the best real advantage of the 16-bit structure: its performance. The 9980 took training cycles to executive coaching for an 8-bit peripheral, for that reason slicing the effective performance in half and making it no higher than existing 8-bit microprocessors. Before TI’s grand plan changed into found-out, Carsten left to end up VP of sales and advertising at Intel, undoubtedly sensing that Intel becomes a tough competitor to conquer inside the microprocessor market.
Intel turned into, of course, developing its own 16-bit microprocessor, the 8086, which became sooner or later brought in April 1978. The enterprise addressed the shortage of compatible 16-bit peripheral chips in exactly the equal manner TI had, with the aid of including an 8-bit port to its microprocessor, which yielded the Intel 8088. Like the TI 9980, the Intel 8088 also became a dog, showing decreased performance compared to the 8086 in any real gadget layout. The Intel chip did have one essential benefit over the TI chip: It had 20 bits of logical deal with space instead of sixteen. That translates into the ability to address one megabyte of reminiscence compared to 64K bytes for TI’s 9900. In addition, the off-chip registers for the TMS9900 and 9980 bogged down their overall performance even greater.
And at the same time, as Intel had efficiently developed change production sources for the 8086, TI struggled to shut similar deals. At the time, maximum clients wanted a minimum of two competing suppliers for any new circle of semiconductor additives to ensure product availability and maintain down fees.
Meanwhile, a few competition had announced plans for their own 16-bit trendy-motive microprocessors. Motorola’s 68000 become the maximum ambition. Although it had 16 external pins, it without a doubt had a 32-bit architecture internally, with the capacity to cope with 24 bits of logical address area externally. A follow-on product could probably be able to deal with 32 bits. Zilog, the writer of the popular eight-bit Z80 microprocessor, announced it’d introduce the 16-bit Z8000, which had a segmented memory, in the past due 1978 or early 1979. Unlike the 68000, even though the Z8000 had a sincere 16-bit architecture.
Motorola MC68000 die Image: Pauli Rautakorpi
The Also-Ran: Motorola’s 16-bit 68000 microprocessor had a 32-bit internal architecture, but the chip wasn’t geared up in time to be taken into consideration for the IBM PC. In October 1978, six months after the declaration of the Intel 8086, I moved to TI’s MOS Division and have become the manager for microprocessors. By this time, each person at the company and many outdoor people knew that TI’s 16-bit microprocessor strategy wasn’t working. Compounding that trouble turned into the division’s in large part unsuccessful attempt to increase a compatible 16-bit microcontroller, referred to as the TMS9940, which become in its fifth or 6th re-spin by the time I arrived. I knew I changed into inheriting a tough scenario. So why would I surrender an excellent job as engineering department supervisor of the consumer merchandise organization? The answer is area, vicinity, region. The microprocessor enterprise became primarily based in Houston, while TI had moved the consumer merchandise group to Lubbock, Texas. Lubbock is a town wherein the ideal answer to the query, “How do you want living right here?” is “The human beings are super.” U. S . A . Track singer Mac Davis, who grew up there, once wrote music whose chorus went, “I idea happiness changed into Lubbock, Texas, in my rearview replicate.”
Shortly when I arrived in Houston, I become advised that I might need to provide a presentation at the TMS9900 to a group from IBM that changed into operating on a very secret assignment that required a 16-bit microprocessor. The institution got here from as an alternative uncommon place for IBM: Boca Raton, Florida. I spent plenty of time getting ready, gave what I thought changed into a properly-polished presentation, and diligently accompanied up. But the IBM group displayed restrained enthusiasm. We wouldn’t realize till 1981 just what we had lost.
John Opel, president and then CEO of IBM, had finished something alternatively modern when he formed the Boca Raton organization, which later have become referred to as the Entry Systems Division. He found out that personal computers from Apple, Commodore, Radio Shack, TI, and others might sooner or later pose a hazard to IBM’s dominance of the pc commercial enterprise. So he gave the Boca Raton organization, which mentioned to Philip (Don) Estridge, carte blanche at the product it was developing—which become the IBM personal pc, of the path. They could use 0.33 parties for something they selected, along with the running machine and alertness software. This latitude made the machine pretty “open” through IBM’s requirements, and it’d possibly accelerate the time to marketplace. Opel imposed one restrict, however. The product could deliver the IBM call not to harm the company’s reputation for satisfaction and reliability. To that stop, IBM’s large first-rate-assurance organization had to log off at the product earlier than it can be released.