aledlhug...@aol.com (Aled Hughes) writes: > Back in the early '80s, I was told that IBM's Model Range for the 3083 > - E, B and J - used the initial letter of the Product Managers' last > names for the models. Anyone know if this was true?
some 3083 topic drift this account has 3081 (& 3033) using warmed over FS technology ... both started off Q&D efforts to get stuff back in the 370 product pipelines (after demise of FS ... FS was completely different than 370 and was going to completely replace it ... 370 efforts were being killed off during the FS period ... and the lack of 370 products during the period is credited with clone processors getting market foothold). http://www.jfsowa.com/computer/memo125.hm posts mentioning FS http://www.garlic.com/~lynn/submain.html#futuresys The initial 3081D was supposedly two 5mip processors but some benchmarks had it slower than 3033 (at around 4-4.5mips). Doubling cache for 3081K then supposedly was two 7mip processors ... and some benchmarks had it about same as 3033. 3081 was initially going to be multiprocessor only ... but TPF (renamed airline congrol program) didn't have multiprocessor support and there was concern that the whole TPF customer base would migrate to clone vendors (which continued to ship newer, faster uniprocessors). Initial response was some very unnatural things done to vm370 for customized TPF running in virtual machine on 3081 (but significantly impacting vm370 throughput for all other customers). They eventually decide to come out with single processor 3083 ... part of the problem was simply removing the 2nd 3081 processor ... was it was in the middle of the box ... which would have made the box dangerously top heavy ... so they had to remove the top processor and rewire the box for the only processor in the middle of the box. Also the latency and throughput of the I/O microcode in the 3081 was really poor ... and TPF environments tended to be very I/O intensive ... as a result there were also customized I/O microcode loads for 3083 TPF environments (that attempted to compensate/mask its otherwise poor performance characteristics). The other issue is long time POK 370s had 10-15% hardware multiprocessor penalty ... processor clock was slowed down 10-15% (compared to single processor machine) to allow for cross-cache synchronization in two-processor system. In theory initial 3083 processor should have gotten a 10-15% processor boost (over 3081), but it continued to run the processor clock at 3081 speed. Lots of difficulty going to 3084 because each processor cache had to deal with three other caches instead of only one other cache. For 3084, MVS & VM370 got a lot of storage allocation work, kernel storage was change to multiple of cache-line size and aligned on cache boundaries (so didn't have two different pieces of storage occupying same cache line) ... which is claimed to increased overall performance by 5-6% (for 4-way operation). some past 3083 posts http://www.garlic.com/~lynn/2005s.html#38 MVCIN instruction http://www.garlic.com/~lynn/2009l.html#65 ACP, One of the Oldest Open Source Apps http://www.garlic.com/~lynn/2010n.html#16 Sabre Talk Information? -- virtualization experience starting Jan1968, online at home since Mar1970 ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
