home *** CD-ROM | disk | FTP | other *** search
- Xref: sparky comp.os.vms:13579 comp.sys.dec:4491
- Path: sparky!uunet!haven.umd.edu!darwin.sura.net!mips!sdd.hp.com!ux1.cso.uiuc.edu!mp.cs.niu.edu!fnnews!fndcd.fnal.gov!nagy
- From: nagy@nagy.fnal.gov (Frank J. Nagy:VAX Wizard&Loose Cannon)
- Newsgroups: comp.os.vms,comp.sys.dec
- Subject: Re: New VAXes/VAXstations: can these be realized?
- Message-ID: <1992Aug12.140056.1@nagy.fnal.gov>
- Date: 12 Aug 92 20:00:56 GMT
- References: <1992Aug11.212818.19498@thijssen.nl> <1992Aug12.092737.674@calmasd.prime.com>
- Sender: news@fnnews.fnal.gov
- Followup-To: comp.os.vms
- Organization: Fermilab Computing Division
- Lines: 49
- Nntp-Posting-Host: nagy.fnal.gov
-
- >> What would make it impossible for DEC to make a multi(VAX)processor
- >> VAXstation in the same box as a 4000 model 60 or 90, for instance a four
- >> 32VUPprocessor VAXstation 4000 model 94? (imagine 120VUP on the desktop
- >> :) VMS can't be the problem: it supports SMP since V5.0.
- >
- > Technologically speaking, there's nothing keeping DEC from developing a
- > multiprocessor VAXstation based on the 32 VUP chipset.
- >
-
- Digital actually had a multiprocessor VAXstation. Now what was that
- name... The VAXstation 3520 (2 CPUs) and 3540 (4 CPUs) as I remember.
- These were based on the same chip used in the 3100 on a board with
- two processors which fitted into an M-BUS backplane (which then had
- an adapter to a Qbus).
-
- > Ah, but 4 32VUP processors does not equal one 120 VUP box. The only time you
- > are ever going to take full advantage of having more than a single processor is
- > if there are multiple processes (or threads) in the compute queue ready for a
- > processor.
-
- Actually, a multiprocessor VAXstation makes even more sense today.
- It was noted that the VS 3520 ran pretty well - what happened was that
- the X server code ended up running on one CPU and the client application
- on the other. Using X automatically guarantees that any display application
- is going to have >= 2 processes. Also, the old VAX Workstation Software
- (VWS) is pretty much dead; this was a kernel-based graphics windowing system
- and was not useable on a multiprocessor.
-
- > The development costs of such a system are probably prohibitive as well based
- > on the returns.
-
- I believe that the software costs to VMS should not be any more difficult
- than those needed for any new VAX since VMS supports SMP already. The
- hardware might be just a tad more difficult (more board real estate for
- the second CPU, cache synchronization and the like) but its not clear it
- is worth it today when a 2x or 3x faster chip is just 4-6 months away
- anyway.
-
- I do believe that 2nd or maybe 3rd generation of the Alpha machines will
- begin to see multiprocessor workstations reappear as a means of getting
- higher performance (remember that one of the means of getting the 1000x
- performance improvement in the Alpha lifetime is multiprocessing and this
- likely means across the entire line of Alpha systems).
-
- = Dr. Frank J. Nagy "VAX Guru & Wizard, Loose Cannon" {{Group Leader!}}
- = Fermilab Computing Division/Distributed Computing Dept/Special Projects Grp
- = HEPnet/SPAN: FNDCD::NAGY (43123::NAGY) or FNAL::NAGY (43009::NAGY)
- = Internet: NAGY@FNAL.FNAL.GOV = BitNet: NAGY@FNAL
- = USnail: Fermilab POB 500 MS/234 Batavia, IL 60510
-