News & Views / April 1996

Intel: 200-MHz Pentiums to Arrive This Year

Tom R. Halfhill

By the end of the year, the fastest Pentium will run at a blazing 200 MHz and CPU prices will drop so fast that businesses will regard PCs with 120- and 133-MHz Pentiums as entry-level boxes. While low-end systems will range in price from $1200 to $1500, mid-range PCs costing $1500 to $2000 will have 150- and 166-MHz Pentiums.

It's not magic; it's science. Intel says the transition to its next-generation process technology is several months ahead of schedule. By the fourth quarter, Intel's foundries expect to manufacture 90 percent of their microprocessors on the denser 0.35-micron process.

That means higher clock speeds, higher yields, and lower prices. It also spells trouble for Intel's competitors. AMD, still struggling to get its next-generation x86 designs out the door, says the SSA-5 version of its K5 processor was expected to ship in March. But it runs at only 75 MHz and delivers about the same performance as a 75-MHz Pentium. AMD says an improved K5 that matches the performance of a 133- or 150-MHz Pentium won't ship until the fourth quarter.

AMD's K6, designed by NexGen, also is scheduled to ship in the fourth quarter. At 180 MHz, it's supposed to beat the performance of a similarly clocked Pentium Pro.

Cyrix is fighting hard to keep up, too. Its new 6x86 is now in production at 100, 110, 120, and 133 MHz. The 133-MHz 6x86 exceeds the performance of a 166-MHz Pentium, according to the new P-rating system adopted by AMD, Cyrix, and IBM. P-ratings measure application performance relative to a Pentium. For example, Cyrix says the 6x86-P166 is at least as fast as a 166-MHz Pentium, even though the 6x86-P166 actually runs at 133 MHz. To determine the P-ratings, AMD, Cyrix, and IBM submit their CPUs to MDR Labs, an independent testing facility operated by MicroDesign Resources, publisher of the Microprocessor Report. MDR Labs doesn't assign the P-ratings but provides raw data that's interpreted by the vendors.

Copyright 1994-1998 BYTE

Return to Tom's BYTE index page