home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!zaphod.mps.ohio-state.edu!cs.utexas.edu!uwm.edu!ogicse!das-news.harvard.edu!cantaloupe.srv.cs.cmu.edu!lindsay
- From: lindsay+@cs.cmu.edu (Donald Lindsay)
- Newsgroups: comp.arch
- Subject: Re: Mainframe >> PC, why?
- Message-ID: <BxCvCu.JIn.2@cs.cmu.edu>
- Date: 7 Nov 92 17:02:53 GMT
- Article-I.D.: cs.BxCvCu.JIn.2
- References: <1992Nov7.015649.8958@fwi.uva.nl>
- Sender: news@cs.cmu.edu (Usenet News System)
- Organization: School of Computer Science, Carnegie Mellon
- Lines: 29
- Nntp-Posting-Host: gandalf.cs.cmu.edu
-
- djakman@fwi.uva.nl (Kemal Djakman) writes:
- >But, my question is, what makes the (current) difference in power for the
- >mainframes against the PCs?
-
- There are a number of differences. CPU power used to be one of them,
- but that is ceasing to be true.
-
- The other differences largely remain:
- - multiprocessor - 4,6 or 8 way.
- - big caches: may be megabytes/CPU instead of kilobytes.
- - big memory: plural gigabytes. [And it may be SRAM, not DRAM, at that.]
- - "solid state disk", ie plural gigabytes of block-access RAM.
- - high aggregate memory bandwith: the processors do not compete
- for a limited resource such as a memory bus.
- - high aggregate I/O bandwidth - GB/s not MB/s.
- - high channel counts - e.g. a hundred times as many I/O devices as
- you can imagine attaching to one PC.
- - extensive fault detection and fault tolerance.
- - features for rapid fault analysis and rapid repair.
- - serious power conditioning and power backup - like, diesels.
- - software suitable to keep the above busy.
-
- Not that I'm advocating traditional mainframes. It will be very
- interesting to watch HP and DEC as they build RISC-based mainframes.
- These new machines are for a new technical environment (increasingly
- fast networks) and for an application mix that is increasingly
- distributed.
- --
- Don D.C.Lindsay Carnegie Mellon Computer Science
-