How Intel Beat Texas Instruments for IBM's PC Processor

In an oral history recorded by the Computer History Museum, former Texas Instruments executive Walden C. Rhines recalled that IBM's chief executive set one constraint for the Boca Raton PC team: "Don't damage the IBM name, so use IBM qualified parts."

That requirement to rely on IBM-qualified, production-ready components framed the 1981 IBM PC as a project driven by schedule, manufacturing readiness and supplier risk rather than by an open-ended search for the most advanced processor design.

Within that framework, later accounts describe IBM narrowing its 16-bit options to two production microprocessor families that could pass qualification in time: Intel's 8086 and 8088, and Texas Instruments' TMS9900 line.

Inside IBM's Checklist

In a Byte magazine article summarized by EE Journal, IBM engineer David Bradley later outlined four main requirements for the PC's processor choice: overcoming the 64 kilobyte address limit of 8-bit systems, immediate availability of the CPU and its peripherals, use of technology IBM already understood, and access to an existing operating system and applications.

Bradley wrote that the team needed a rich set of support chips, including a DMA controller, an interrupt controller, timers and parallel ports, and that there was no time for new large scale integration development because components had to be available in production quantities.

These criteria meant that IBM prioritized a shippable system with familiar parts and a software base over performance advantages that would have required new silicon or a long learning period.

Why Motorola and Zilog Fell Out

Motorola's MC68000 offered a 24-bit logical address space and a 32-bit internal architecture, which made it attractive on technical grounds, but it was still being sampled rather than shipped in full production when IBM needed to qualify a processor, according to both Rhines and later reporting in EE Journal.

Rhines recalled that the 68000 would have been IBM's preferred architectural choice if it had been ready for production and able to pass IBM qualification on the required schedule.

Zilog's Z8000 was also in development as a 16-bit contender, but IBM did not appear to consider it seriously, in part because Exxon Enterprises owned Zilog and was viewed as a potential competitor, according to Federico Faggin's account cited in EE Journal.

Once those designs were set aside, Rhines described the practical choice as lying between Intel's 8086 or 8088 and TI's TMS9900 or its related parts, all of which were already in production and could pursue IBM qualification.

Intel 8088: Technical Fit at Lower Cost

Intel's 8086 family provided a 20-bit logical address space, which allowed addressing of up to one megabyte of memory, compared with the 16-bit, 64 kilobyte limit of many earlier microprocessors and TI's TMS9900, as described by Walden Rhines in IEEE Spectrum.

The 8088 variant paired that 20-bit address space with an 8-bit external data bus, giving IBM a 16-bit internal architecture while keeping the board design compatible with low-cost 8-bit memory and input or output components.

In his oral history for the Computer History Museum, Intel executive Dave House said Intel already sold a family of inexpensive 8-bit peripheral chips that could be reused alongside the 8088 with minimal changes.

House recounted that these existing peripheral chips had multiple second sources and had become inexpensive through competition, so coupling them with the 8088 helped IBM meet its system cost goals.

TI TMS9900: Minicomputer DNA, PC Problems

Texas Instruments entered the contest with the TMS9900, a 16-bit microprocessor derived from the company's 990 minicomputer line, which shared the minicomputer's architecture and addressing model.

Rhines wrote in IEEE Spectrum that the TMS9900 had only 16 bits of logical address space, which limited it to 64 kilobytes of directly addressable memory and matched the ceiling IBM wanted to exceed.

He also noted in his oral history for the Computer History Museum that when Motorola's 68000 and Zilog's Z8000 were not ready for qualification, IBM's decision effectively came down to Intel's 8086 or 8088 versus TI's 9900 or 9980. The four extra bits of address space on Intel's side were enough to influence the outcome.

At the time, TI lacked a full line of compatible 16-bit peripheral chips, and Rhines wrote that engineers responded by creating the TMS9980, an 8-bit bus version of the 9900 that allowed use of existing 8-bit peripherals.

According to that IEEE Spectrum account, the 9980 needed two bus cycles to move each 16-bit word, cutting effective performance roughly in half and removing much of the practical benefit of a 16-bit CPU in cost-sensitive designs.

Software and Ecosystem Economics

Bradley's criteria also required that both an operating system and application software be available for the chosen microprocessor. This favored architectures that already had development tools and commercial programs in use, such as Intel's 8086 and 8088.

As summarized by EE Journal, this requirement reduced the amount of new software IBM had to commission or build internally for the PC's launch.

House's oral history further emphasized that IBM valued the ability to reuse low-cost 8-bit peripheral chips from multiple vendors, which reduced component costs and diversified supply for key parts beyond the CPU itself.

Price, Sourcing and the Final Negotiation

Price commitments and sourcing terms played a central role in the final negotiations between Intel and IBM.

House recalled in his oral history for the Computer History Museum that Intel pitched IBM on a system that combined 16-bit processing and addressing with 8-bit peripheral chips.

He stated that Intel framed this design as low cost and multi-sourced, in line with IBM's preference for at least two suppliers.

In the same interview, he stated that IBM later required Intel to establish a second source for the 8088 at AMD and that Intel agreed to license the design so AMD could manufacture compatible parts.

House also described making a specific pricing commitment, saying he had to promise to reduce the 8088's price to five dollars in high volume within two years to secure the PC design win.

Decision Day and Immediate Impact

With the field narrowed and the commercial terms in place, IBM's Entry Systems group selected the 8088 as the processor for the IBM PC, pairing it with Intel's low-cost 8-bit peripheral chips.

Accounts in EE Journal note that many of these Intel peripherals, already familiar from earlier designs, were incorporated into the PC's system architecture.

Rhines later wrote that TI's TMS9900 did not achieve the broad market adoption the company had hoped for, particularly in personal computer designs, and that the IBM decision highlighted weaknesses in TI's 16-bit microprocessor strategy.

Long Shadow of a Pragmatic Choice

Across the interviews and articles, IBM's processor choice for the original PC emerges as a pragmatic response to qualification rules, available components, software readiness and supplier diversity rather than a search for the most advanced instruction set.

The combination of a one megabyte address space, reuse of inexpensive 8-bit peripherals, existing software and a clear second source arrangement helped make Intel's 8088 acceptable to IBM despite its narrower external bus and performance compromises.

Those criteria, applied under tight time pressure, helped establish the x86 line as the foundation for mainstream personal computing. They show how procurement constraints and ecosystem maturity can shape the long-term direction of a technology platform.

Sources

• Leibson, Steven. "How the Intel 8088 Got Its Bus." EE Journal, 2022.
• House, Dave. "Oral History of Dave House." Computer History Museum, 2004.
• Rhines, Walden C. "Oral History of Walden C. Rhines." Computer History Museum, 2012.
• Rhines, Walden C. "The Inside Story of Texas Instruments' Biggest Blunder: The TMS9900 Microprocessor." IEEE Spectrum, 2017.