I was just pointing out that multi-CPU machines don't seem to go back as far as the 1950s, much less exploitation of those multiple CPUs in any meaningful ways. Here are the dates I can find on the various machines you list:
Bendix G21: The G20 was introduced in February, 1961, so the G21 had to come after that. (And after the D825, which was delivered in August, 1962, since that machine is widely credited as first.) Bendix sold out to CDC sometime in 1963, so the G21 might have had a CDC badge (more or less) by the time it was delivered. (CDC kept producing the G15s for a few years.) The Carnegie Institute of Technology, one of the predecessors of Carnegie-Mellon University, might have been the sole G21 customer. The G21 had 2 CPUs. This machine architecture does not have any direct (or probably even indirect) descendants. Burroughs B5000: December, 1962 (after their D825), and perhaps later than that for the 2 CPU version. The B5000 is controversial here since in 2 CPU configuration it was more master-slave from what I gather. But it was certainly a highly significant machine in the evolution of computing. The B5000 has direct, lineal descendants in today's Unisys ClearPath MCP mainframes. GE 625: Actually, it was the 635 that was available with multiple processors (usually 2 if ordered that way; one customer ordered 4). The 635 was November, 1964. (Again, MP systems might have shipped later.) This machine, too, has direct, lineal descendants in today's Groupe BULL GECOS-based and NEC ACOS-based machines, with an intermediate journey through Honeywell. UNIVAC 1108: There was an 1108 II model (a.k.a. 1108A) introduced in August, 1965, which was the machine available with multiple CPUs. Wikipedia says 296 of the 1108s (uniprocessor and multiprocessor) were sold, so this was a reasonably popular machine for its time. Again, these machines have direct, lineal descendants in today's Unisys ClearPath Dorado mainframes. But your central point remains, even if the 1950s aren't involved: MP designs appeared long before many people assume. And IBM certainly wasn't the first (though obviously being late to that particular party didn't matter and might have even been a feature, not a bug, a la today's maxi-core-fixated Sun). Somebody else might remember and be able to characterize the vendors' relative uniprocessor performances in order to add a little a color to the picture. Also any information about cross-machine clustering, if any, in that era. - - - - - Timothy Sipples IBM Consulting Enterprise Software Architect Based in Tokyo, Serving IBM Japan / Asia-Pacific E-Mail: timothy.sipp...@us.ibm.com ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html