Xref: utzoo comp.sys.dec:2266 comp.unix.ultrix:2286
Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!mailrus!purdue!decwrl!crltrx!max.crl.dec.com!jg
From: jg@max.crl.dec.com (Jim Gettys)
Newsgroups: comp.sys.dec,comp.unix.ultrix
Subject: Re: 3100 mix and match
Message-ID: <1392@crltrx.crl.dec.com>
Date: 6 Dec 89 14:34:33 GMT
References: <4305@helios.ee.lbl.gov> <1989Nov28.191803.18970@sun.soe.clarkson.edu> <8732@cbmvax.UUCP>
Sender: news@crltrx.crl.dec.com
Reply-To: jg@max.crl.dec.com (Jim Gettys)
Organization: DEC Cambridge Research Lab
Lines: 20

The GPX on the PMAX would have been a "graphics decellerator".

For example, at best, the GPX can do around 20k characters/second.
PMAX does around 50-60K characters/second (in color, monochrome is
yet faster; R4 numbers are around 100k for color).

Similar numbers apply for almost all graphics operations, with
the exception of aligned raster-op, where memory bandwidth on the
PMAX is less than that available on the GPX.  The unaligned case,
which one might have expected the hardware to help on, turns
out to be slower in hardware than the PMAX does in software.
Everything else is much faster.

On the whole, for most applications, the PMAX outperforms a GPX system
by a large margin.  For complex stuff, a large multiple indeed.

So go look at real numbers before coming to any conclusions on graphics
performance.  The x11perf program will give you an interesting set
of performance numbers to work from.
				- Jim Gettys