[tw]

[from the previous post}

As IBM was self destructing, the PC business suddenly found a dearth of innovation. IBM under Estridge directed each wave of innovation. With Estridge promoted to the penalty box, and an MBA named Cannovino stifling innovation, IBM in 1990 was not introducing innovation standards. Step forward Compaq. Compaq addressed one serious bottleneck in computers - the ISA bus. IBM had introduced MircoChannel as the solution. And then (like AT&T and Unix), chose to stifle the innovation with massive royalties that only made sense on a spread sheet. So Compaq and a large number of PC companies developed the EISA bus. Well, the consortium was unwieldy. It could not get the innovation out fast enough. But from those mistakes, the industry learned how to expand on and advance what was demonstrated previously by DIX.

The bottleneck was eventually replaced by an even better solution - PCI bus. The fundamentals still exist today. A standard by which any manufacturer could contribute innovation to the PC. While Apple remained foolishly attached to Mac's closed architecture and their obsolete processors. Apple was being run into the ground by two business school graduates - Sculley and then Spindler.

Intel had successfully out innovated Zilog, Motorola, and IBM by using product oriented benchmarks. Not using business school (spread sheet and money game) benchmarks. As a result, Intel had amassed quite a few crown jewels. Especially their fundamental research and innovation into semiconductor manufacturing.

The original 1980 Intel processors used NMOS. However in the 1960s, RCA had already developed a CMOS microprocessor (COSMAC). A computer that could run on a battery. But that is when business school graduates started replacing innovators in RCA. Business school graduates would not let innovators in Somerville NJ developed software (ie programming language, compiler, etc) until the product had started making a profit. Which it could not without the necessary software. But again, that is what business school graduates do. Make judgments based upon profits rather than learn anything about the product.

RCA had pioneered CMOS logic (4000 series) which eventually every digital company would convert to a decade later (Texas Instruments, et al would redesign the entire and legendary 7400 family into CMOS). Even Intel abandoned NMOS processors for CMOS. And implemented other innovations (ie power down) by sharing technology with other innovators (Harris Semiconductor). That was in the days when Intel was in competition with Motorola's 68000. Others were given the rights to manufacturer and market Intel products in exchange for innovating new devices. Harris was one. Another was Advanced Micro Devices (AMD).

Well, AMD did no innovation. AMD simply started manufacturing and selling Intel products. Intel and AMD began long, legal battles. AMD's objective was right out of the business schools. Profits. Innovation costs too much. As a result, AMD had no profits and a stifled engineering. Then something major happened in AMD. NexGen was designing an Intel 80686 compatible processor. AMD bought NexGen. Then kept the entire NexGen engineering department intact. And banned any AMD management from even talking to the engineers. The result was a product oriented development known as the K-6. It was equal to Intel's products and was sold at a lower price.

As happens to all companies, business school disease infects the organization. That happened in another of Intel's crown jewel. The processor architectural development group. Processors are first designed at the architectural level. This group, for some reason, started to design the Pentium as if it was a Motorola 68000. They stuffed it with all kinds of new concepts (pipelining). The processor became so large as to be impractical in manufacturing. But the architectural group, for some reason, had no grasp of the monster they were creating. Suddenly Intel had an inferior product. In an emergency redesign, the Pentium 4 was stripped down. To maintain Moore's Law, compilers had to be redesigned to reorder machine instructions. If in the right order, a Pentium 4 was equal to AMD's K-6. But if instructions were not properly ordered, the Pentium 4 would slow to a crawl.

And so again, the only thing keeping Intel alive was Moore's Law. A benchmark to quantify innovation four and more years before those innovations could be seen on a spread sheet. AMD spend quite a few years prospering at Intel's expense because AMD kept management away from their innovators - the team mostly composed of NexGen designers.

AMD also moved production operations to Germany. To take advantage of optics technology and lower labor costs found in Dresden. But that does not address AMD's major problem. AMD did not have the semiconductor innovation abilities that helped Intel to constantly meet Moore's Law.

Read back to a series of earlier posts where I discuss the major problem in semiconductors. A CMOS transistor has a glass surface. A charge on one side of that glass caused the other side to conduct (turn on or off). This glass was reduced so thin as to be only three atoms thick. Made that thin to make the transistor faster. Current leaking through that glass cause processors to consume 60 watts. Wave after wave of innovation (stressed silicon) was increasing the speed. But still not addressing the bottleneck - heat and power consumption.

AMD's solution was to implement technologies from other companies. Hoping others could provide better transistors. A business model that would have significance later to other companies.

Everyone knew where the ultimate solution lay - High K materials. A material with insulating abilities superior to glass. But nobody could make it work.

{one more post follows}