The PowerPC 620 chip, expected to be IBM Corp and Motorola Inc’s contender for the 64-bit systems market, is struggling to keep its place on the PowerPC road-map following problems with the design schedule. The pressure has been increased by IBM’s realisation that the performance of the PowerPC 604 CMOS process could be tweaked up […]
The PowerPC 620 chip, expected to be IBM Corp and Motorola Inc’s contender for the 64-bit systems market, is struggling to keep its place on the PowerPC road-map following problems with the design schedule. The pressure has been increased by IBM’s realisation that the performance of the PowerPC 604 CMOS process could be tweaked up higher than design points laid down a couple of years ago, in so-called 604 Plus and Plus Plus iterations. Indeed it looked so good in performance that it drove past the 620, was how one IBMer described the way 604 has come along. James Thomas, director of RISC microprocessor development at the Systems, Technology & Architecture Division in Austin, Texas, insists the 620 is still being actively and heavily debugged, but admitted that there had not been as much progress as was hoped for. We wanted some limited availability. Although we have had silicon back, it is a tedious, time-consuming debug over 64-bit and symmetric multiprocessing aspects. He said that performance projections were still valid, but that we will be looking at the schedule very hard and that other ways of satisfying these requirements other than rolling out the 620 are possible. Randy Groves, vice-president for RS/6000 systems development, showed a chart missing out the 620 altogether, and heralding the 630 (the Power3 instruction set CPU, not a PowerPC unit) as the next major step on. Although originally designed and optimised for the technical market, the 630 could equally serve for commercial systems, he said. Problems with the 620 – and an apparent re-focus on the Power instruction set – reflect a wider malaise in the PowerPC camp thought to be the result of the nature of the triumvirate itself, the difficulty Motorola Inc is said to have had working to such short time-scales and the fact that the partners were wrong-footed by what Intel Corp has been able to achieve with Pentium. A re-vamped PowerPC road-map was supposedly on the runway as we went to press which IBM had planned to delineate at this week’s Microprocessor Forum. When we spoke to Motorola it said it certainly hadn’t blessed any road-map, indeed it said the thing had been pulled while a slew of outstanding business issues are resolved with IBM. Moreover, Phil Pompa, director marketing, RISC microprocessor division in Motorola’s Microprocessor & Memories Technology Group, said that as far as Motorola’s concerned the 620 is a long way from dead. IBM admitted the PowerPC road-map would more likely show up later in the month than this week, but insisted it would be in play by Halloween as customer meetings had already been arranged where IBM vice-president and general manager of Systems Technology & Architecture Division Phil Hester and his troops are to disclose the plans. Meanwhile, the PowerPC 604 has its own problems, related to the complexities of testing for use in symmetric multiprocessing systems. The 604 suffered from cache coherency problems when working in teams of more than two. Those problems are now pretty much solved, according to Thomas, though systems vendors using the 604 still have to go through their own verification and testing efforts. Machines are now expected during the first half of 1996.