home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!hela.iti.org!cs.widener.edu!cs.widener.edu!usenet
- From: YOUNG@tattoo.cs.widener.edu (Rob Young)
- Newsgroups: comp.sys.sun.hardware,widener.misc
- Subject: Re: New Product Announcements: WHERE???
- Date: 12 Nov 1992 06:14:29 GMT
- Organization: Widener University Department of Computer Science, Chester PA
- Lines: 56
- Distribution: world
- Message-ID: <1dsso5INNmo9@cs.widener.edu>
- References: <19921110215132CSP1DWD@MVS.OAC.UCLA.EDU> <lg2i1dINNj7k@appserv.Eng.Sun.COM>
- NNTP-Posting-Host: tattoo.cs.widener.edu
- X-News-Reader: VMS NEWS 1.20
- In-Reply-To: dwf@doug.Eng.Sun.COM's message of 11 Nov 92 17:51:41 GMT
-
- In <lg2i1dINNj7k@appserv.Eng.Sun.COM> dwf@doug.Eng.Sun.COM writes:
-
-
- > These new Sun(TM) workstations deliver 59.1 MIPS, with the
- > SPARCclassic offering industry-leading value at $63 per MIPS.
-
- BZZZZZTTTTTTTTT!!!!!!!!!
- wrong
-
- First the math is wrong... 63 * 59.1 is $3723.30. This is less than 3,995. Am
- I missing something here? Is there a discount for a gross (144) that works
- out to 3700 dollars per workstation that hasn't been talked about yet? ;) ;)
- ;)
-
- But even to be talking about dollars per MIPS in November of 1992 is a
- bit puzzling. To quote John Mashey of Silicon Graphics from a post in May:
-
-
- BEGIN QUOTE
-
- .... So far, overall, at least the following are multiple-issue superscalar /
- LIW chips: i860, IBM RS/6000, SuperSPARC, and all of the results so far tend to
- show fairly wide variations in the number of actual instructions per cycle
- achieved; also, nobody so far seems to have gotten close to achieving the peak
- theoretical instruction rates that you'd expect by multiplying the clock rate X
- by the number of issues N to get X*N peak "mips".
-
- ANYONE WHO LABELS A CPU WITH X*N "mips" AS MEANING SOMETHING USEFUL IS:
- EITHER: a) Completely unaware of the research on this topic, as well as
- the numerous published measurements on real machines.
- OR b) Trying to fool you...
-
- This silliness very similar to that of "peak MFLOPS" from supercomputer-land;
- it's even worse than using Dhrystone-mips as being anything useful.
-
- [section deleted]
-
- 5) Anyway, as a botton line, if somebody tries to feed you peak mips,
- please tell them politely to stop insulting your intelligence...
-
- END QUOTE
-
-
- Besides it isn't industry leading. Using DEC's AXP 3000 400 @133 MHz
- 266 MIPS (dual-issue) comes out to $56.39 per "MIPS".
-
- I have a feeling someone at Digital had engineering read over the market blurbs
- before posting, whereas at Sun, well . . . Better would be a dollar per
- Specmark ratio, no?
-
- Rob
-
- young@tattoo.cs.widener.edu
- "The reasonable man adapts himself to the world; the unreasonable one persists
- in trying to adapt the world to himself. Therefore all progress depends
- on the unreasonable man." --George Bernard Shaw
-