roberts@nbs-vms.ARPA@caip.RUTGERS.EDU (02/25/86)
From: "ROBERTS, JOHN" <roberts@nbs-vms.ARPA> > Subject: Expand at your own risk! > Date: Sun, 23 Feb 86 07:00:49 -0800 > I advise everyone to not buy either fixed hard disks or memory for their > Amiga yet. We are about to see the next quantum leap in the capacity of > memory chips, to 1 megabit per chip. A company called TheSys has > announced a single IBM PC expansion board containing 8 megabytes of 1 > megabit chips. Actually the board contains 12 MB; the extra capacity is > used to implement error correction on board. The memory is CMOS, meaning > it can be backed up with batteries. Cost: $800. ... > I hereby predict that fixed hard disks will be obsolete within a year. > Steve Walton > Caltech Solar Astronomy From the February 20, 1986 issue of EDN, page 136: "Currently the 256k-bit [DRAM] parts go for $4 to $8 (1000) based on process, organization, addressing options, and speed. ... Most industry observers believe that the cost per bit of 1M-bit parts will reach that of 256k-bit chips sometime between mid-1987 and early 1988. ... Generally, customers in 1985 paid $60 to $100 for 1M-bit dynamic-RAM samples. Mason expects the price to hit the $35 to $40 level early in 1986. By the end of 1986, Mason believes the price will drop to about $20. ... All potential 1M-bit dynamic-RAM vendors plan a 256kX4-bit version and in general will introduce a 256kX4-bit IC within half a year of their 1MX1-bit introduction...." - 12MB of megabit DRAM would represent 96 chips. At the highly optimistic price of $20 per chip (near the crossover point with 256k bit chips) this would result in a chip price alone of $1920. You could buy the boards for $800, pull the chips off, and sell them at a fantastic profit. Seriously, I would be surprised if any modern, independently viable memory board were offered for less than several times the price of the memory chips. - If a board were designed to use the 265kX4 chips, availability and drop in price could be expected to lag noticeably behind the 1MX1 curves. - I am as enthusiastic about low memory prices as the next person. However, it seems a shame to drive the current memory system manufacturers out of business by waiting for a product/price that will probably not be available at least for several years. - In reference to hard disks: Semiconductor memory has always had trouble in its price war with magnetic storage. As semiconductor prices per bit drop, so do magnetic memory prices. Hard disk technology for small systems today is vastly better and cheaper than that of only 3 or 4 years ago, and there does not seem to be any obvious limitation to the trend. I suspect that magnetic storage is still more reliable than long-term semiconductor storage. For example, semiconductor storage is far more susceptible to radiation-induced errors than magnetic storage, and it is possible to bungle the design of a semiconductor system such that all the data can be lost when you change batteries. (An on-board short-circuit is pretty much guaranteed to erase the data.) With any type of fixed memory system, any sensible person will periodically back up vital files to removable media, which thus far has meant magnetic media. - In reference to optical disks: CD-ROMs (read-only) are in the process of becoming available for a number of systems, and are excellent for storage of large amounts of pre-written (i.e. reference) data, with fairly high throughput but pretty long access times. "Write once" and "write many times" technologies, however, are much less well-developed and vastly more expensive. Barring unforeseen changes in research emphasis, it could well be 5-10 years before such systems are available at a price that personal computer users can afford. In addition, the properties that make optical disks so useful for storing large amounts of data make it virtually inevitable that track access times will be relatively long, though doubtless current performance levels can be improved. - Personal preferences: I would like two of each, please. (Standard disclaimers.) John Roberts roberts@nbs-vms.ARPA ------