*BSD News Article 20840


Return to BSD News archive

Newsgroups: comp.os.386bsd.questions
Path: sserve!newshost.anu.edu.au!munnari.oz.au!network.ucsd.edu!swrinde!elroy.jpl.nasa.gov!ames!cnn.nas.nasa.gov!wilbur.nas.nasa.gov!tweten
From: tweten@wilbur.nas.nasa.gov (Dave Tweten)
Subject: Re: local bus video?
Message-ID: <CDAxzw.LJ9@nas.nasa.gov>
Sender: news@nas.nasa.gov (News Administrator)
Nntp-Posting-Host: wilbur.nas.nasa.gov
Organization: NAS Systems Division, NASA Ames
References: <jstCDApw0.4E8@netcom.com>
Date: Mon, 13 Sep 1993 17:02:19 GMT
Lines: 37

In article <jstCDApw0.4E8@netcom.com> jst@netcom.com (Jason Thomasser) writes:
>Does NetBSD/FreeBSD support local bus video?

Local bus is not a software issue.  It's a hardware issue.
Net/Free/386 BSD can't tell if your graphics card is local bus or not.
In general, if your card's register semantics and mapped memory areas
are supported by your software, it works, regardless of whether the
card is ISA, EISA, VLB or maybe even MCA.

>If so will the ATI local bus cards work and are they advisable?

The VLB ATI Graphics Ultra Pro works well.  I use it.

>I guess I may as well ask, why would local bus be an advantage?

You want it for graphics update speed, particularly for "frame buffer"
style cards like those based upon the ET-4000 chip.  Processors do lots
of "peeks and pokes" on these cards.  Both EISA and ISA are 8 MHz
busses (EISA is just twice as wide and uses half the bus cycles to
block-transfer data).  If you're trying to write a byte into a frame
buffer (repeatedly!), they're both 8 megabyte per second busses,
divided by the number of ticks to deliver a byte.  VLB for a 33 MHz
486DX motherboard is a 33 megabyte per second bus, again, divided by
the number of ticks to deliver a byte.  With no wait states on the VESA
local bus, you can get about a factor of 10 improvement in "peek and
poke" performance over either EISA or ISA.

The advantage is less pronounced for graphics coprocessor cards (like
the ATI GUP and the S3-based cards) because the required traffic
between your CPU and the card is reduced.  Performance becomes much
more a function of the graphics coprocessor's speed and the video
memory's speed.
-- 
---------------------------------------------------------------------------
Dave Tweten						tweten@nas.nasa.gov
NASA Ames Research Center, M/S 258-5			     (415) 604-4416
Moffett Field, CA  94035-1000				FAX: (415) 604-4377