[email protected] (Tony Harminc) writes: > How were either of the 4331 or 4361 scope'able? Surely both were at > about the same level of integration as the TCMs in the 30x0... > > Or had the scope'able requirement quietly disappeared by that point in > favour of redundancy, leaving the other advantages of the service > processor?
re: http://www.garlic.com/~lynn/2017c.html#81 GREAT presentation on the history of the mainframe http://www.garlic.com/~lynn/2017c.html#88 GREAT presentation on the history of the mainframe http://www.garlic.com/~lynn/2017c.html#89 GREAT presentation on the history of the mainframe 3081&3090 had significant denser physical packaging contained within TCM modules (no contacts to scope) ... built probes into the TCM packaging and diagnositic using the probes from the service processor. https://www-03.ibm.com/ibm/history/exhibits/vintage/vintage_4506VV2137.html originally single 4331 and then upgraded to pair of redundant 4361s ... scopable ... less dense. 4300s and DEC VAX saw similar explosion in sales in the low & mid-range market for single or few unit orders ... old post with decade of VAX sales sliced&diced by year, model, US/non-US http://www.garlic.com/~lynn/2002f.html#0 Big differrence between 4300 & DEC VAX was large corporations ordering multiple hundred 4300s at a time for placing out in departmental areas ("leading edge of the distributed computing tsunami"). IBM (& DEC) was expecting continued explosion in low/mid range sales with 4300 followon; 4361 & 4381. However as can be seen in the VAX numbers, by mid-80s, the low/mid range market was starting to move to workstations and large PC servers. Folklore is that 3092 went to pair of redundant 4361s ... because there were large number of unsold 4361s in warehouses. other folklore about 3081 (density/weight). 3081 was originally designed to be multiprocessor only (two processors & channels in single "heavy" frame). Problem came up with ACP/TPS which had loosely-coupled support but no tightly-coupled multiprocessor support. The clone mainframe makers was still making single processor machines and there was concern that the whole ACP/TPS market moves to clone makers. The decision was made to come out with single processor 3083 for the ACP/TPS market (3081 with one processor removed). The easiest would have been to keep processor0 and remove processor1. The problem was that processor0 was at the top of the box and simple removal of processor1 (in the middle, channels at the bottom), could have made the box dangerously top-heavy. https://www-03.ibm.com/ibm/history/exhibits/mainframe/mainframe_PP3081.html this talks about 3033 and 3081 quickly kicked off (including 3081 had enormous excess number of circuits for the level of performance) after failure of FS project http://www.jfsowa.com/computer/memo125.htm past FS posts http://www.garlic.com/~lynn/submain.html#futuresys Note while the above IBM 3081 history page says: Model D is capable of an instruction execution rate of up to 21 times that of a 3033UP running under MVS/SP with identical programs and similar configurations. ... snip ... however there were quite a few benchmarks that had application running on single 3081D processor having lower throughput than on 3033 (single processor). IBM then comes out with 3081K with double the cache size of 3081D (somewhat similar to 168-3 having twice cache size of 168-1), assuming it would improve cache hit rate (and therefor throughput) for some number of applications. past multiprocessor SMP posts http://www.garlic.com/~lynn/subtopic.html#smp -- virtualization experience starting Jan1968, online at home since Mar1970 ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
