MIPS Technologies Inc recently staged a briefing to explain to the world just why users of Windows NT should run the operating system on machines using MIPS processors. During the conference, Chris Rowen, a director of MIPS Technologies Inc Europe, flashed up a host of slides to show that MIPS processors are better-designed to run NT than chips from Intel Corp and Digital Equipment Corp. Using a Dell Computer Corp 80486DX2/66 personal computer as its base level of measurement, the company ran an NEC Corp RISCStation 4400 processor using a 150MHz R4400SC chip against a DEC Alpha AXP/150 machine using a 150MHz DEC21064 processor, and a Compaq Computer Corp Deskpro M/60 model, with a 60MHz Pentium chip under the bonnet. The firm tried to keep it all above board by shopping for each machine with roughly the same budget. NT was used to run a number of applications on the machine, including image processing and visualisation tools, a Hamilton C-Shell, the Byte benchmark and software compilation. The MIPS R-series-driven machine came out three times as fast as the Dell machine on average, twice as fast as the Pentium and roughly a one and a third times as fast again as DEC’s Alpha machine.

Bad light

Predictable results – after all, MIPS isn’t going to show benchmarks which throw its processors into a bad light. The diatribe continued. According to Rowen, $350 can buy you part of a Pentium chip, a little bit of an Alpha chip or – and here’s the hook – a complete MIPS RISC personal computer base system bar the memory. The company’s low-cost R4200 CPU costs you $70, while secondary cache, 80486 base motherboard, logic chips, slots and other electronic goodies will fit nicely into your $350 budget, which should please the manufacturers. From a commercial and technological standpoint, these two points seem to clear the way for MIPS, although it should be pointed out that the benchmarks were performed just on processing power and made no allowance for the different bus possibilities involved. Certain applications could benefit from a Peripheral Components Interconnect implementation of the Pentium chip that the MIPS chip couldn’t offer, and this could change the results a little. For the uninitiated, the background story on MIPS is that the firm designs its own chips but doesn’t fabricate them; like a number of other firms including Cyrix Corp and Advanced RISC Machines Ltd, it sells the designs to other companies who make and sell the chips. These firms include NEC, LSI Logic Corp and Siemens AG.

By Danny Bradbury

Focusing only on chip design helps to keep costs down for the company, which says it could spend $2m on architecture definition and $10m to $100m on product development, but will leave the hundreds of millions of dollars spent on product design and manufacturing to its industry partners. Several systems builders have adopted the MIPS architecture: NEC uses the company’s processors, along with Acer Group Inc, Siemens, Sumitomo Electronics Corp, Sony Corp, Deskstation Inc and DEC but not for much longer – also use the company’s chips in their machines. It is quick to distance itself a little from Silicon Graphics, its parent, by pointing out that Silicon uses only 3.5% of the total production of MIPS processors. This is a statement obviously designed to damp down concern over the lack of independance inherent in a chip designer that is owned by a system manufacturer, Silicon Graphics Inc. MIPS went through a bit of a rough patch around the turn of 1991-92, when the Advanced Computing Environment initiative effectively went west. ACE was initially a RISC personal computer standard that had been co-ordinated by Compaq and MIPS and backed by roughly 250 industry players, most of which pulled out for various reasons. When Compaq pulled out to concentrate on core business and Santa Cruz Operation Inc withdrew Open Desktop development for the MIPS processor, to wait and see how ACE was shaping up, the initiative took a distinct dive, and although it still hasn’t officially died, the body’s been cold for a long time now. Rowe

n admits that The ACE initiative is effectively dead as a marketing programme, but says that some of the technical aspects are still there, such as MIPS’ Advanced RISC Computing chip set, around which the ACE architecture was based. Other benefits are Unix on MIPS and NT on MIPS. The worrying thing is the importance that MIPS is placing on NT, which after all still has to prove itself in the market. NT represents a real significant volume opportunity over the next couple of years, says Rowen. It has the same characteristics as Unix – people want low cost and high performance. They want a starting point for the emergence of multimedia, he adds. This is all very well, assuming that NT makes its mark: if it doesn’t then the company is going to have to rely on the Unix market for the medium term. Admittedly, this wouldn’t be such a great hardship but will it offer the growth potential that the company is hoping for from NT? Whatever happens, the firm, obviously chastised by the ACE fiasco, is looking to expand its activities outside the desktop computer market in the long term to take advantage of the increasing opportunities in the embedded processor market. Rowen estimates that 90% of the units shipped currently go into the embedded market, although the value would be lower than that because the embedded market tends to take older chips with a lower price. To this end, the firm has been busy signing with Nintendo Co to design the engine for a virtual reality video game.

Outlandish

It also scooped a deal with Time Warner Inc to provide its R-series RISC chip for use with the company’s interactive television technology. Using this sort of deal to tap into the vast consumer market, MIPS hopes to increase the number of units shipped from 1m in 1993 to hundreds of millions. This year’s figure is three times that of 1992, so this ambition may not be as outlandish as it sounds. In the meantime, the company is preparing the next generation of workstation chips, which should help to speed the current generation towards the embedded sector. The TFP superscalar R4000 derivative chip will provide Cray Research Inc Y-MP performance, while the T5 single superscalar chip will run at over 500 MIPS. The R4200 chip is a low cost R4000 derivative running at under 2W, aimed at the portable and low cost personal computer markets. These were all discussed in CI No 2,192. Assuming these hit their schedules and targets, they should see the company attacking new markets and keeping up appearances in the short to medium term – no matter what happens to the NT market.