Path: utzoo!attcan!uunet!aplcen!uakari.primate.wisc.edu!zaphod.mps.ohio-state.edu!usc!ucsd!ucbvax!STC06.CTD.ORNL.GOV!de5 From: de5@STC06.CTD.ORNL.GOV (SILL D E) Newsgroups: comp.misc Subject: Re: Acronyms. Is is `1 MIP' or `1 MIPS'? Message-ID: <9007031705.AA04537@stc06.CTD.ORNL.GOV> Date: 3 Jul 90 17:05:10 GMT References: <+I84:_@xds13.ferranti.com> <1990Jun27.014943.27951@sq.sq.com> <312@spam.ua.oz> Sender: daemon@ucbvax.BERKELEY.EDU Reply-To: de5@ornl.gov (Dave Sill) Organization: Oak Ridge National Lab Lines: 42 This is a repost through UCB's mail-news gateway. Can anyone confirm or deny receipt of message-id: <1990Jun27.154217.22556@cs.utk.edu>? I'm trying to verify that my postings are leaving UT. In article <312@spam.ua.oz> wvenable@spam.ua.oz (Bill Venables) writes: > >Anyway enough of this. Let me ask a more serious question. Since the >original machine used to define the 1 mips standard was a CISC >machine, (the VAX 11/780 I believe), does the unit really have much >meaning for the now more common RISC machines? What is 1 million >instructions, anyway? How should one realistically compare computing >power on radically different architectures? Ooh, good questions. First, the reference MIPS standard is indeed the 11/780. Someone, probably DEC, claimed 1 MIPS for the 780. Whether or not it actually runs at 1 MIPS I have no idea. What's interesting is the way MIPS ratings, particularly for the new RISC machines is being calculated. It's based on dhrystones. When Sun claims 12.5 MIPS for the SPARCstation 1 (or SLC), that's derived using the formula 1858 dhrystones/sec = 1 MIPS, i.e, 23050 (measured SS1 dhrystones/sec) divided by 1858 d/s = 12.5 MIPS. So basically when vendors claim X MIPS, think in terms of dhrystone performance-- which is not generally a valid measure of overall system performance. The "DECstation 2100 and 3100 Workstations Technical Overview" even says that the published MIPS figures for these machines are dhrystone based. The only valid measure of performance is to run your actual applications under "field" conditions on a system configured exactly as it would/will be after purchase. Everything else is just a prediction. Stamp out MIPS! Don't use the term at all, especially in list of requirements for a procurement spec. If you want dhrystones, say *dhrystones*. -- Dave Sill (de5@ornl.gov) These are my opinions. Martin Marietta Energy Systems Workstation Support