mccalpin@vax1.acs.udel.EDU (John D Mccalpin) (03/21/90)
Perhaps it would clarify the discussion of KILLER MICROS and single- versus multiple-user machines if we expanded Eugene Brooks dictum: "There is no defense against the attack of the killer micros" by noting that "micro" here means "microprocessor", _not_ "microcomputer"! As Eugene has often pointed out, in order to make a machine capable of performing large scientific calculations, it is not going to look much like a traditional microcomputer --- with a minimum of 128 MB RAM and a few GB of disk, it is still not going to be cheap enough for each individual to have on the desk, but it is going to be powered by a cpu of the same family as the machines available all the way down to the "personal workstation" range. -- John D. McCalpin mccalpin@vax1.acs.udel.edu Assistant Professor mccalpin@delocn.udel.edu College of Marine Studies, U. Del. mccalpin@scri1.scri.fsu.edu
jg@crltrx.crl.dec.com (Jim Gettys) (03/21/90)
In article <5899@udccvax1.acs.udel.EDU> mccalpin@vax1.acs.udel.EDU (John D Mccalpin) writes: >As Eugene has often pointed out, in order to make a machine capable of >performing large scientific calculations, it is not going to look much >like a traditional microcomputer --- with a minimum of 128 MB RAM and >a few GB of disk, it is still not going to be cheap enough for each >individual to have on the desk, but it is going to be powered by a cpu >of the same family as the machines available all the way down to the >"personal workstation" range. Well, times do change. Lets look at the cost of the machine you are talking about, using current prices. Memory is around $100/megabyte, (third party), so 128 meg is $12800 Disk is around $6000/gigabyte. So I could put together a "killer micro" system for performing scientific computation (according to your definition) for between $30K and $40K this year. This is about the price range that workstations started to go into individual offices. A machine slightly less ambitious is in the $25K price range (1 gig of disk, 64 meg of memory). And these are this year's prices. So some people will be getting such machines in their offices as "personal workstations" later this year. Your point is exactly correct though, compatible $5-10K machines that are much smaller will be ubiquitous, with some number of jucier machines in one's network. This all presumes you don't need what Supercomputers are actually good for (tons of I/O, vectorized code, cost/cycle no object because real time a concern). The attack continues..... - Jim -- Digital Equipment Corporation Cambridge Research Laboratory