Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!ukma!xanth!mcnc!ecsvax!dukeac!wolves!ggw From: ggw@wolves.UUCP (Gregory G. Woodbury) Newsgroups: comp.arch Subject: Re: 88K vs. MIPS vs. SPARC vs. 386 Summary: Opus Systems 88K box Message-ID: Date: 9 May 89 02:16:32 GMT References: <13089@paris.ics.uci.edu> <1916@se-sd.sandiego.ncr.com> <162@dg.dg.com> Organization: Red Wolfe Software Lines: 60 In <162@dg.dg.com> rec@dg.UUCP (Robert Cousins) wrote: > In article <1916@se-sd.sandiego.ncr.com> Dave Lord(SSP) writes: > >In article <13089@paris.ics.uci.edu> John M.A. Roy writes: > >>In article , grunwald@flute writes: > >>|Does anyone, other than Data General, make an 88K box *available now* that > >>|runs a reasonable UNIX (i.e. should be able to run Berkelyish code). Has > > The product was announced in February. . . . > > >By the way, you didn't hear this from me. :-) I haven't seen anyone yet mention the Opus/Everex box. Opus makes UNIX co- processors for AT-bus machines that are pretty good. They are now shipping a 20MHz 88000 card w/20MB ram (model 417-20) and if you want an Everex 386 clone workstation underpinning it, they have the whole thing as the model 8040 "personal mainframe". (Enough quotes from their marketing hype) I happen to have one of these wonderful boxes sitting on my desk. 1st impression: very nice! Once I figured out what I had to do to get everthing and X windows and TCP/IP up and running (actually - the 8040s come with everything pre-loaded and configured) I transferred in my main numerical analysis application and ran my regression test. In the otherwise unloaded machine, the GOM benchmarked at 8.39 minutes, versus 21.4 minutes for a 30MHz Intergraph Clipper. This includes a mix of double precision floating point (a lot), periodic disk i/o (checkpointing) and the usual mix of character and integer calculations supporting the main d.p. loops. 2nd impression: still very nice! The X11 version software that I received was beta test and still has some problems, but its better than waiting for UNIX to access the "ultragraphics controller" for normal ascii i/o (really! clearing the screen takes a good 4-6 seconds!) Additionally, I managed to find a bug in the pcc compiler (with a work around) and a bug in the ftpd. Since I managed to snag the 1st 8040 that they shipped commercially, I guess that I deserve getting the "low serial number" problems - but that seems to be normal with us. Opus Tech Support is very responsive and says I'll get a new kernal in the next week or so. For a price (very steep) source support is available, but we are operating in a binary-only situation. 3rd impression: still very nice! More information as I glean it. BTW - the introductory price for this beast was around 28,000. 20MHz '386, 88000, 330 MB disk, ethernet, x-windows and Berkely Environment (over SysVr3.2). What a deal. The Usual disclaimers apply, I am not connected with Opus other than as a rather satisfied customer (6 machines and one more coming). -- Greg Woodbury, system programmer and admin (UNIX) Duke University Center for Demographic Studies UUCP: ...mcnc!duke!dukcds!wolves!ggw or ...mcnc!ecsgate!dukeac!wolves!ggw ^ this is my home machine Internet: ggw%dukcds@cs.duke.edu or ggw@dukeac.ac.duke.edu Voice: 919-493-1998 (home) 919-684-6126 (work) USMail: 902 A1 Park Ridge Rd. Durham NC 27713