rb@ccird1.UUCP (Rex Ballard) (06/17/86)
AT&T compares a "washtub" computer with a desktop PC and gets flamed for not supporting the same number of users as the washtub. It is quite possible that comparing their UNIX-PC with a 780 with 512K of memory, and one RM-80 (80MB) disk drive, and one user, both running SYS-V, using graphics intensive applications, raw mode I/O, and bit-mapped graphics device, that the AT&T would be about as fast. Of course, when you get the 780 up to full power, put the tty hardware in front, the CDC 300 MB drives, full caching, read-ahead, and put 4.2 or 4.3 on it, and put thirty users running "shared text" programs, the VAX would be quite a bit more powerful. There are still people who think that a PDP-11/23 is more powerful than a Stride 68K or a 5/32. The PC might appear slower, but that is because it takes .25 MIPS per user to maintain the graphics display. The VAX lets the VT-100 take this load. Most of this does not affect CPU speed, but does affect Bus Bandwidth. When refreshing the display, that bus time cannot be used for CPU transfer. True, the AT&T cannot be loaded with 300 MB drives. I don't know that that would be an advantage anyway. Maybe AT&T will come up with an SCSI interface that will support that kind of load, but right now, that is not the goal. Is the ad misleading? Yes. Since only minimum configurations are being compared, people who have been exposed to fully expanded VAXEN are going to assume that the only comparison is the CPU benchmark. Is AT&T lying? No. AT&T made certain trade-offs on how that CPU and Bus Bandwidth speed would be distributed. A more valid comparison might have been to compare 20 AT&T PC's to a VAX and 20 VT-240's. But if you start including the MIPS of the 240's, and the cost, the price/user and the performance/user make the VAX look more attractive (Though the AT&T would still look better). I have run performance tests between a Series 1 and 8 Z-80's running CP/M in a network. In one specific application (data base update and retrieval) the Z-80's won, but not by the pure percentage of CPU power.