home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!usc!zaphod.mps.ohio-state.edu!rpi!psinntp!psinntp!colorne!dfilip
- From: dfilip@colorne.UUCP
- Newsgroups: comp.os.vms
- Subject: Re: MIPS vs. VUPS
- Keywords: mips vups
- Message-ID: <37@colorne.UUCP>
- Date: 7 Nov 92 13:58:32 GMT
- References: <2929725069.0.p00180@psilink.com>
- Distribution: world
- Summary: MIPS == VUPS ???
- Organization: ColorNet Information Systems
- Lines: 53
-
- George S. Chapman writes :
- >>Ronald Becker Williams writes :
- >>We've been told by DEC Ultrix Personnell that plus or
- >>minus a few, that VUPS = MIPS, and vice versa. Is
- >>this reasonable?
-
- >Absolutely. That was what the original equivalence standard was.
- >I've been using it for years when comparing to other vendors.
-
- The problem with this is that, once you jump machine architectures, this
- number becomes mostly (if not completely) useless. The reason that DEC
- stopped quoting MIPS was that they were beating up IBM a few years back
- (circa 1985 ?) about how their price per MIP beat the pants off of IBM
- by something like a factor of 2 to 1. Then IBM came back with some benchmarks
- and proved that you could get twice as many 'transactions' processed with
- an IBM MIP than you could with a DEC MIP. To which DEC responded that IBM's
- numbers were invalid because they represented batched transactions and not
- online, to which IBM responded that large systems are measured in transactions
- and not interactive access, to which ...
-
- VUPS does make a lot of sense in comparing a VAXen, since the VAX
- architecture is still somewhat the same (although the implementation has
- varied quite a bit from processor line to processor line). However, DEC
- generally gives VUP's in ranges (particularly on their larger systems)
- because a VUP started out as an 11/780 (which they used to call a MIP)
- and there have been enough changes in the VAX implementation since
- then that your rating will vary depend on specifically what you are
- trying to doing. Case in point is that some of the old 8xxx series
- could easily beat some of the earlier 6xxx series machines which had
- a much higher VUP rating, primarily because of some instructions which
- were emulated (the older 6xxx series actually used the uVAX chip set,
- although I am not sure if the 64xx/65xx still do).
-
- An of course, I/O intensive applications will run much better on an XMI
- attached RF-series disk than a Unibus attached RA-series disk, regardless
- of the VUP rating !
-
- IMHO, VUP/MIP ratings between completely different machine architectures
- are relatively useless. VUP/MIP ratings between the same architecture
- machines gives a close ballpark for comparison, but are still not
- concrete, unless you take other differences in the machines implementation
- into effect (i.e., I/O bus, memory bus, emulated instructions, etc.)
-
- Regards,
-
- Dave.
- --
- =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
- = David E. Filip UUCP : dfilip@colornet.com =
- = ColorNet Information Systems CIS : 76430,3111 =
- =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
- = Standards are wonderful 'cause there are so many to choose from ! =
- =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
-