How CompaQ views Intel roadmap

It is inconceivable that the roadmaps from Alpha Processor Inc (API) we published yesterday could have reached the light of day without Compaq's imprimatur.

After all, not only does Compaq have a minority share in API, with Samsung the main shareholder, but the Houston company is also irretrievably married to the Alpha platform itself. A close examination of API's competitive microprocessor roadmap reveals a lot not only about the Alpha roadmap but about Intel's plans too. Rather significantly, API only compares the Alpha 21264 with Merced, Xeon and CuMine, and ignores the K7 and other microprocessors.

Despite the fact that CuMine is now delayed, the API roadmap shows that the 750MHz Alpha Slot B processor will best the Xeon 666 and CuMine 600 in SPECint95 performance in Q3 of this year. The 21264-750 reaches around 43 SPECint95s or so, the Xeon 666 under 30, and CuMine 22, according to API's estimations. Further, when the 21264-833 is released in Q4 of this year, it will reach 50 SPECint95s, and even though API estimates that the Xeon platform will have got to 32 by then, the Alpha processor is still thrashing it.

By the time API/Compaq release the 21264-933 in Q1 of next year, SPECint95 performance will be around 55, while the Xeon 733 will be around 33 and CuMine 700 a mere 30. The estimate for the Alpha 1GHz 21264 is that it will reach 60 SPECint95 by Q2 of next year. Merced will (may?) enter the picture at 46 SPECint95, while the Xeon 8XX will tip in at 39 and the CuMine 8xx at 32.

If these figures are to be believed, and if API is able to pull the Alpha processor down into the commodity desktop and server markets, it seems that Intel will have a real fight on its hands. One wild card is Willamette, the kind of Intel's equivalent of a mass chip of destruction.

Two other questions. Does Compaq operate Chinese Walls between its Alpha and Intel operations? And does Intel, which fabs some Alpha chips, also employ the same infamous Chinese Walls? ®