Hard technology: Tran *** eta Crusoe of x86 Kamikaze Special Attack Team at the end of 20th century.

Do you remember Tran *** eta, a small x86 processor factory? Do you remember a group of Japanese computer manufacturers actively using their Crusoe to make thin and light laptops?

Kamikaze attack team driver: Tran *** eta Company was established: 1995 Company disappeared: 2009 (acquired by Novafora). Number of employees: 300. The most symbolic product: the turning point of Crusoe from prosperity to decline: the regret that Iverson failed to realize: a pile of future Iverson left in the briefing. 1In the middle and late 1990s, Efficeon developed and sold x86 instruction set compatible processors. Because Windows 95 detonated the shipment of personal computers (the global shipment of personal computers reached the milestone of 1 1 100 million units at the end of 1998), it became a mouth-watering fat market and attracted a large number of people with lofty ideals to participate in the grand event.

Excluding AMD( 1969), a "high-performance x86 processor" factory, according to its establishment date, it was built from NexGen( 1986), Cyrix( 1988), Rise( 1993), Centaur (.

But in the end, everything backfired. Up to now, apart from Intel AMD, the x86 duo are only Centaur of the United States and Elbrus of Russia acquired by VIA in Taiwan Province Province. These two "only exist", I'm afraid they are too weak to distinguish products from knowledge.

However, this does not mean that the products of these kamikaze special forces are worthless. Among them, Tran *** eta, once joined by Linux creator Linus Torvalds, is the most unique in technology and the most topical among team members. The Crusoe processor released by Tranmeta on June 5438+1October 65438+1October 9, 2000 was not only favored by a large number of Japanese computer manufacturers, but also used to build thin and light notebook computers, which was the first public offering (IPO) on June 65438+1October 7 of that year.

Tran *** eta applied for the patent right as early as August 20th, 1996, which revealed its unique "CMS". Through the dynamic binary translation of pure software, the x86 instruction code is converted into the native instruction sequence of the self-developed VLIW processor, which can be optimized repeatedly, reduce the number of actually executed instructions, realize software non-sequential instruction execution and memory loading, make the most commonly executed software run faster and faster, and give consideration to performance and power saving.

Of course, the internal structure of this ultra-long instruction set processor is incredibly simple. This is also fully reflected in the petite wafer area: the same similarity 0. 18? M process, TM5600( 128kB L 1 cache +5 12kB L2 cache) is only 88mm? , significantly smaller than 106mm(32kB l 1 cache +256kB L2 cache) of mobile Pentium III? . When 0. 13? M and 5 12kB L2 cache, even more "55mm? vs 80mm? 」。

In other words, Tran *** eta is different from other competitors who fight transistors with Intel. It tries to subvert Moore's Law with "software thinking in the Internet age". From June 5438+ 10 in 2000 to March 2003, Tran *** eta has no rivals in the field of low-power x86 processors. Crusoe TM5600 (clock 700MHz) integrated with Northbridge chipset and memory controller costs only $200, and its performance is close to that of Intel Mobile Pentium III 500MHz, which costs more than $500 (excluding chipset).

However, the good times did not last long. On March 65438+2, 2003, Intel began to promote its Centrino "mobile computing technology" and its Pentium M processor. Coupled with the overwhelming $300 million marketing activities, the advantages of Tran *** eta have ended. In addition, due to the limitation of x86 software compatibility, Crusoe did not have excellent performance and growth rate, which forced Tran *** eta to adopt a low-price promotion strategy, making it difficult for Tran *** eta to make a big profit, even losing nearly $600 million at the beginning of 2004.

As for Linus Torvalds, the kanban figure of Tran *** eta, he left this company which was once called "the most important company in Silicon Valley" in July 2003. So, can Efficeon, the second generation product, bring them back to life? Of course, it is absolutely impossible. Koko.

Tracing the Hardware World, Archaeological Exploration, Wangying Technology