IBM 801/RISC
IBM 801/RISC
I would claim that a motivation for 801/risc was to go the opposite of the complex Future System effort from the 1st half of the 70s, completely different than 370 and was going to completely replace 370s (during FS, internal politics was killing off 370 efforts, claims that the lack of new 370 products during FS is credited with giving the clone 370 makers, their market foothold). When FS implodes, there is a mad rush to get stuff into the 370 product pipelines, including kicking off quick&dirty 3033&3081 efforts.
Then late 70s, there was several 801/RISC activities to replace a wide variety of internal CISC microprocessors, controllers, mid-range 370s follow-on to 4331/4341 (4361/4383), follow-on to S38 (as/400). For a variety of reasons these efforts floundered and things returned to CISC (and some number of RISC engineers left IBM for other vendors). I contributed to white paper that instead of 370 microcode on CISC or RISC microprocessor, it was possible to implement nearly the whole 370 directly in circuits. IBM Boeblingen lab did a 3-chip "ROMAN" 370, directly in circuits with the performance of 370/168.
The ROMP 801/RISC (running PL.8/CP.r) was for OPD to do the follow-on to the Displaywriter. When that was canceled (in part because market moving to PCs), it was decided to pivot to the UNIX workstation market and the company that had done the AT&T Unix port to IBM/PC for PC/IX, is hired to do one for ROMP ... which becomes PC/RT and AIX.
IBM Palo Alto that was working on BSD unix port to 370, then redirected to the PC/RT, which becomes "AOS" (for the PC/RT).
Then work begins on the 6-chip RIOS for POWER and RS/6000. My wife and I get the HA/6000 project, initially for the NYTimes to port their newspaper system (ATEX) from DEC VAXCluster to RS/6000. I rename it HA/CMP when we start doing technical/scientific cluster scale-up with the national labs and commercial cluster scale-up with the RDBMS vendors (Oracle, Sybase, Informix, Ingres). Then the executive we report to moves over to head up the AIM
Somerset effort to do a single-chip power/pc (including adopting the morotola RISC M81K cache and cache consistency for supporting shared memory, tightly-coupled, multiprocessor)
领英推荐
Early Jan92, there is HA/CMP meeting with Oracle CEO where AWD/Hester tells Ellison that we would have 16processor clusters by mid92 and 128processor clusters by ye92. Then by late Jan92, cluster scaleup is transferred for announce as IBM supercomputer (for technical/scientific ONLY) and we are told we can't work with anything that has more than four processors (we leave IBM a few months later).
Comparison of high-end mainframe and RIOS/POWER cluster, which possibly contributed to performance kneecapping (commercial) HA/CMP systems (industry MIPS benchmark, not actual instruction count, but number of benchmark program iterations compared to reference platform):
1993: eight processor ES/9000-982 : 408MIPS, 51MIPS/processor
1993: RS6000/990 : 126MIPS; 16-way: 2016MIPs, 128-way: 16,128MIPS
By late 90s, the i86 chip vendors were developing technology with hardware layer that translates i86 instructions into RISC micro-ops for execution, largely negating the performance difference between 801/RISC and i86.
1999 single IBM PowerPC 440 hits 1,000MIPS
1999 single Pentium3 (translation to RISC micro-ops for execution) hits 2,054MIPS (twice PowerPC 440)
2003 single Pentium4 processor 9.7BIPS (9700MIPS)
2010 E5-2600 XEON server blade, two chip, 16 processor aggregate 500BIPS (31BIPS/processor)
trivia: some Stanford people had approached IBM Palo Alto Scientific Center about IBM producing a workstation they had developed. PASC invites a few IBM labs/centers to a review, all the reviewers claim what they were doing was much better than the Stanford workstation (and IBM declines). The Stanford people then form their own company to produce SUN workstations.
trivia2: folklore is the large scale naval carrier war games after turn of century had (silent) diesel/electric submarines taking out the carrier.
Senior Research Scientist, Nexus
6 个月I was a member of the 801 group from 1978-1983. There were (at least) three alumni of the FS project in the group as well as three members of the original Fortran group. And you should mention John Cocke who was the godfather of the group and later won the Turing award for invention of RISC.