NetNews Usenet Archive 1992 #27/NN_1992_27.iso/spool/comp/benchmar/1761

Internet Message Format | 1992-11-21 | 3.6 KB

Xref: sparky comp.benchmarks:1761 comp.arch:11006
Path: sparky!uunet!ornl!rsg1.er.usgs.gov!darwin.sura.net!zaphod.mps.ohio-state.edu!uwm.edu!rutgers!cbmvax!jesup
From: [e-mail] (Randell Jesup)
Newsgroups: comp.benchmarks,comp.arch
Subject: Re: Who wants faster machines  was  DEC ALPHA Performance Claims
Message-ID: <[e-mail]>
Date: 22 Nov 92 00:42:05 GMT
References: <[e-mail]> <[e-mail]> <[e-mail]> <[e-mail]>
Reply-To: [e-mail] (Randell Jesup)
Organization: Commodore, West Chester, PA
Lines: 53

[e-mail] (Robert Hyatt) writes:
>In article <[e-mail]> [e-mail] (Daniel Packman) writes:
>>In article <[e-mail]> [e-mail] (Peter Gray) writes:
>>>No HP, no DEC, no RS6000. How many of our SUN users want
>>>a faster machine? About 2. For the rest 10 Mips (SS1 type speed)
>>>is plenty....  How many banks need 100 Mips on peoples desktops?

>>As has been said, who cares if the payroll program that runs overnight
>>takes 8 hours or 8 minutes?  Indeed, in certain situations speed is
>>not that important.  For us, however, we find present algorithms still
>>limited by the speed of processors available.  With faster machines,
>>still more complex calculations become possible.  I cannot forsee a
>>faster machine that will be perceived by scientific institutions as
>>irrelevant.

>I agree.  Even if I were to stop doing "number crunching" on my workstation,
>I still care about how quickly it can pop up the mailtool, or pop up a
>new window, or run a document thru LaTeX or compile a program.  Once it can
>do any of these (or similar) functions in "zero" time (which could be defined
>as less than .25 seconds) then anything faster wouldn't affect me.

>In essence, any appreciable
>delay is noticable and anything that can be done to reduce it will have a
>BIG market.  When you start cracking .1 sec intervals, then you really are
>talking to the true number crunchers and not the majority of workstation
>users.  But there are still Cray's for 'em ........

    As I've said before: as soon as you produce more powerful machines,
operations that were considered "too expensive" now become required.  We
software people can absorb just about any likely amount of CPU power and
thirst for more.  There are plenty of past examples (such as windowing systems,
Xwindows in particular, and then ever more layers on top of it); and I
can think of plenty of things that will eat all the CPU power that is likely
to be affordable (multi-media, virtual reality, speech recognition, visual
recognition, adaptive behavior, etc, etc).  Not all of these _will_ become
standard (predicting the future is rough), but certainly if a machine has
more than the required speed for what it's doing, the things it's doing will
change or become more expensive.

    Robert is correct in using response time as an indicator of
performance; it's how people "feel" performance.  A 5 Spec machine running
OpenLook, etc can feel sluggish, and the same machine running a lightweight
OS/windowing system can feel ridiculously snappy.  That doesn't mean the
lightweight OS is "better", it means that people are designing software for
which 5 Spec machines are becoming marginal or unacceptable.  I doubt this
trend will reverse itself.

--
To be or not to be = 0xff
-
Randell Jesup, Jack-of-quite-a-few-trades, Commodore Engineering.
{uunet|rutgers}!cbmvax!jesup, [e-mail]  BIX: rjesup
Disclaimer: Nothing I say is anything other than my personal opinion.