The Bristol-based company, the brainchild of Transputer-architect Professor David May FRS, Professor of Computer Science at Bristol University, represents a second commercial outing for May’s life-long interest in multi-core processing. The first outing, Inmos’ much-applauded Transputer, failed commercially largely because Inmos decreed that it should be programmed in a new language, Occam. Asking the industry to accept one big new thing, a parallel processing CPU, was a big ask, but asking it to accept two big things, a multi-core and a new language was an ask too far. Asked if XMOS is after the FPGA business, Richard Terrill, vice president of marketing at XMOS, replies, “absolutely we’re after the FPGA business, but the big market is ASSP, that’s where the money is.” The ASIC business is worth about $23bn a year, and the ASSP market about $60bn. Because of cost, high power consumption, and the programmability penalty in terms of transistors-per-die, FPGAs have not taken the expected percentage of those markets, and are still only a $4bn market. Terrill reckons it will all come down to cost. “We can bring our chips to market much cheaper than FPGA”, says Terrill. He equates the performance of a one core XMOS part, delivering 400MIPS and costing $1 in volume, which will be sampled later this year, with a $10 Xilinx Spartan chip. TOMORROW MORNING: TEN BEST SEMICONDUCTOR PRODUCTION TOOL COMPANIES
XMOS Samples First Chip
Today XMOS announces the sampling of its first chip, a four core chip, with each core delivering 400MIPS, selling for $10 in volume.