Digital Equipment Corp told the Microprocessor Forum last month that its Alpha chip is in production at 150MHz, with usable yields up to 182MHz. Despite boasts months ago, it has of course been unable to get volumes at 200MHz and probably won’t until third quarter 1993. The forecast 1993 volume pricing is $1,100 in quantities […]
Digital Equipment Corp told the Microprocessor Forum last month that its Alpha chip is in production at 150MHz, with usable yields up to 182MHz. Despite boasts months ago, it has of course been unable to get volumes at 200MHz and probably won’t until third quarter 1993. The forecast 1993 volume pricing is $1,100 in quantities of 1,000 and $800 in quantities of 50,000. DEC expects its next iteration BEV-45 design to produce 300MHz chips that could start shipping in early 1994. Their performance is pegged at 150 SPECint92 and 250 SPECfp92. Meanwhile another design team is working on a four-way superscalar design called EV-5 specified to produce a minimum 300MHz chip whose performance could exceed 200 SPECint92 and 350 SPECfp92. DEC hopes to get this EV-5 part into production in late 1994. On top of these efforts, which also include an EV-6 eventually, DEC is also working on a low-cost LCA 21066 chip to rival Sun Microsystems Inc’s Tsunami and Hewlett-Packard Co’s Hummingbird, and apparently intended for high-volume systems running Microsoft Corp’s Windows NT. The CPU, floating point, instruction and data caches, memory controller and input-output interface will all go on to a single chip. It is expected to have 8Kb caches, larger than Tsunami’s 4Kb-plus-2Kb, and will interface to Peripheral Component Interconnect, Intel Corp’s next-generation bus. Due to ship in the last quarter of next year, it should clock in around 100MHz to 150MHz but the Microprocessor Report reckons its integer performance is unlikely to exceed Pentium’s significantly. The Report’s observations on the overall development programme are worth repeating. It’s been hyped beyond what the company can deliver, it says, pointing to the failure to produce 200MHz Alphas in volume. The plan would strain the resources of any large computer company. Digital, in the throes of major staff reductions caused by billion-dollar losses, will be hard-pressed to keep up with it. Yet it thinks that despite the aggressiveness of the programme, Alpha may be too little, too late. Much of DEC’s future success, it estimates, rests with how well NT does in the market and whether NT independent software vendors will be willing to write to Alpha machines as well as iAPX-86 and R-series RISC ones.
