home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.arch
- Path: sparky!uunet!stanford.edu!rock!doppler!ncsc.org!gbyrd
- From: gbyrd@ncsc.org (Gregory T. Byrd)
- Subject: Re: Could I compete the low-end CM5 with multiple SPARCstations?
- Message-ID: <Brt8En.1zo@doppler.ncsc.org>
- Keywords: CM5, vector units
- Sender: news@doppler.ncsc.org
- Nntp-Posting-Host: duck.ncsc.org
- Reply-To: gbyrd@ncsc.org (Gregory T. Byrd)
- Organization: North Carolina Supercomputing Center
- References: <9207221613.AA05637@x1sun6.ccl.itri.org.tw> <1992Jul22.191250.8476@msuinfo.cl.msu.edu>
- Date: Wed, 22 Jul 1992 21:44:47 GMT
- Lines: 24
-
- In article <1992Jul22.191250.8476@msuinfo.cl.msu.edu>,
- enbody@ss33.cps.msu.edu (Richard Enbody) writes:
- >In article <9207221613.AA05637@x1sun6.ccl.itri.org.tw>,
- lycmit@X1SUN6.CCL.ITRI.ORG.TW (Yin-Chih Lin) writes:
- >|>
- >|> I am curious, if I connect 32 SPARCstation 10 model 30 (33MHz, SuperSPARC
- >|> module with 86.1 MIPS - Sun Micro adverted!?) computers (loosely-coupled
- >|> MIMD multicomputers?) either by Ethernet, ISDN or FDDI with the multi-
- >|> thread OS and might use facilities from Linda-like distributed data
- >|> structures. Is it possible for me to obtain the reasonable performance
- >|> when compared with the 32-node TMC CM5 (SIMD machine)?
- >
- >Greatly simplified response:
- >
- >The difference is communication latency. For many, but certainly not all,
- >applications that could kill you.
-
- Another difference -- most folks who order CM5's will get the
- optional vector units, which provide a peak 128 MFlops
- per node. (And they also do integer ops.)
-
- ...Greg Byrd North Carolina Supercomputing Center
- gbyrd@ncsc.org P.O. Box 12889
- (919)248-1439 Research Triangle Park, NC 27709-2889
-