info-vax@ucbvax.ARPA (11/01/84)
From: Richard Garland <OC.GARLAND%CU20B@COLUMBIA.ARPA> DEC announcement: Oct. 31, 1984 I was at the official DEC announcement today of the "VAX 8600" code named "VENUS". This was announced simultaneously in NY (where I was) Marlboro Mass. (where it is being built) and I think 180 other cities world-wide. For completeness, DEC also announced the VAXstation-1. This is a VAXstation plugged into a microVAX which was actually announced in the trade press 2 weeks ago. I was dissapointed the low end announcement wasn't the microVAX-2 so I skipped listening to most of the VS-1 information. The VAX 8600 - "More than a mainframe" (That catch phrase was spoken about a dozen times.) This is the long awaited VENUS. The basic details as I jotted them down and from the glossy handout: Basic cycle time: 80 nsec Control store: 8k x 86 bits Data path width: 32 bits 4 stage instruction pipeline Memory Cache: 16k write-back Memory bandwidth: 28 MegByte/sec (private bus) IO bandwith (aggregate) 20 MegByte/sec (with 2 SBI's) FPA 4.43 MegWhetstones/sec single Memory limit 32 MegByte in basic box (256 k chips) quoted speed 4.2 x VAX 11-780 Console subsystem: PDP-11 RL02 diagnostics, environmental monitoring, etc. while main processor is running. Size: Same as basic 780. Quieter, slightly less power Availability: First customer ship April 1985 "Widely available" Mid-summer 1985 Packages: 2 were announced: #1 - Cluster Starter: basic box Star coupler HSC cost $500k B.M.C. $1659 (you add tape drive and disk drive: $578k) # 2 - Cluster upgrade: basic box cost $450k B.M.C. $1492 Basic box in above packages: 8600 CPU FPA 12 MegByte memory SBI with: CI UBA with: 1 BA11 box with: 1 DMF32 1 DEUNA Console subsystem with RL02 VMS and DECnet licenses bundled in I believe most figures are accurate but I may have missed some. The basic monthly maintenance charge seems particularly low, either indicating a real bargain in operating cost, or indicating I got the numbers wrong. Check with your salesman obviously. DEC indicated that other packages would be forthcomming but that these initial packages were to satisfy large end customers first. Much emphasis was placed on clusters as a means to obtain "More than a mainframe" total system performance. They kept speaking of 4, 8, 12 etc Mips systems (meaning clusters with 1, 2, 3 etc. 8600's). They stated clusters had advantages over tightly coupled systems such as the 782 and various competitor's machines. No explanation was given for the choice of "8600" except to indicate it was a new class of systems. It does not stand for "twice a 4300". Rg (I have no connection with DEC except as a customer.) -------