NetNews Usenet Archive 1992 #19/NN_1992_19.iso/spool/comp/graphics/9379

Text File | 1992-09-02 | 3.6 KB | 74 lines

Newsgroups: comp.graphics
Path: sparky!uunet!orca!mesa!rthomson
From: [e-mail] (Rich Thomson)
Subject: Re: NCGA Picture Level Benchmark (was: Time to put up or shut up)
Message-ID: <[e-mail]>
Keywords: benchmarks NCGA GPC
Sender: [e-mail]
Nntp-Posting-Host: 130.187.85.21
Reply-To: [e-mail] (Rich Thomson)
Organization: Design Systems Division, Evans & Sutherland, SLC, UT
References: <[e-mail]> <[e-mail]> <[e-mail]>
Date: Wed, 2 Sep 92 18:21:20 GMT
Lines: 59

In article <[e-mail]>
    [e-mail] (Allen Akin) writes:
>Gee, Rich, that's not very imaginative.  Surely you can draw more than
>one conclusion from that data. :-)

I'm basing my conclusion on what the local rep. reported ("we do very
well on them") and the lack of any data that backs up that assertion.
If the numbers on the PLB are great, then lets see them.  A year or
two ago, I had a hard time getting the rep. to show me a million
polygons a second for the VGX.  In that case, there was no problem in
obtaining the benchmark via the net that achieved the quoted rate.  Now,
in this case there was no specifically quoted rate, but just a vague
description of performance.  The answer from the local rep. would seem
to indicate that such a set of performance numbers is available.  All
I'm asking is that they be published, if they exist.

>The most likely situation is the one I mentioned earlier:  It takes
>time and effort to do a good job on the PLB, and if there's little
>demand for the result, then it makes sense to use the resources on
>other projects.  That's just normal business practice, especially in
>lean economic times.

Granted.  No company has infinite manpower to do all things.

However, the PLB allows reporting of unoptimized and optimized results.
The criterion for what constitutes "unoptimized" and "optimized" is
spelled out in the PLB, I believe.  Given the rep's assertion of
performance on these benchmarks, it seems there should be at least
some unoptimized numbers that can be published.  Given that there are
numbers reported for one benchmark running on a VGX, were the other
benchmarks were run, but not reported?

The thing I'm trying to point out is that you need meaningful comparisons
in order to objectively evaluate the performance of hardware.  The PLB
comes closer to this than other benchmark programs produced.  Most of
the existing benchmarks besides PLB are slanted towards one particular
vendor, usually because the vendor has produced the benchmark to show
off their machine in the best light.  That's only natural for a vendor,
and there is nothing wrong with showing off one's product.  However,
for a customer to compare machines from different vendors, these benchmarks
are almost useless.  I would like to see more apples to apples
comparisons in the industry.  I feel that this is a "win" for the
end-user.

>SPEC is different because it figures prominently in many RFPs.  If
>there was comparable demand for PLB data, then I imagine SGI would
>pursue the PLB project more energetically.

I have seen interest in the PLB's grow significantly over the past
year or so now that more numbers are being published.  Of all the hardware
manufacturers in the membership list, SGI is the most
underrepresented (only one set of numbers).

                            -- Rich
--
    Repeal the personal income tax; vote Libertarian in 1992.
Disclaimer: I speak for myself, except as noted; Copyright 1992 Rich Thomson
UUCP: ...!uunet!dsd.es.com!rthomson            Rich Thomson
Internet: [e-mail]    IRC: _Rich_        PEXt Programmer