reiter@endor.harvard.edu (Ehud Reiter) (10/27/88)
In article <468@oracle.UUCP> csimmons@oracle.UUCP (Charles Simmons) writes: >John Mashey brings up a point that I've never had a satisfactory >answer to. If we assume that RISC-based manufacturers can build >machines that outperform mainframes, where will companies like Amdahl >make their money? [Chuck points out that mainframes have lots of fast memory, lots of I/O bandwidth, and lots of reliability/servicability features, and that incoporating these into any machine would make it expensive, even if the basic MIPS were cheap] To this, I would add that most mainframe applications require $$$$ of peripherals (10's of GB of disk, 1000's of terminals, etc) and software, and even more $$$$ in support staff (10's of people, each at $100K/yr including overhead). The point, then, is that MIP's are one of the cheapest components of a mainframe computer system *in any case* (even before RISC). Replacing a $1,000,000 IBM with a $10,000 RISC box won't help much if you still need $4,000,000 of memory, peripherals, and software; and have an operations staff budget of $1,000,000/yr to boot! As long as MIPS remain an insignificant portion of total SYSTEM cost, mainframe users will have no incentive to abandon their current machines and switch to RISC. Ehud Reiter reiter@harvard (ARPA,BITNET,UUCP) reiter@harvard.harvard.EDU (new ARPA)