riscy
[Top] [All Lists]

Re: 21066 board (was Wild idea)

To: af4@ukc.ac.uk, riscy@sunsite.unc.edu
Subject: Re: 21066 board (was Wild idea)
From: Drew Eckhardt <drew@kinglear.cs.Colorado.EDU>
Date: Mon, 31 Jan 1994 18:52:24 -0700
In-reply-to: Your message of "Mon, 31 Jan 1994 06:55:42 EST." <199401311128.AA27976@SunSITE.Unc.EDU>
--------

    Well video is handled by the CPU so the only things to argue about here
    are the amount of VRAM and choice of RAMDAC. As for SCSI, Drew Eckhart
    is in favour of a rather fast NCR PCI SCSI controller for which he has
    written a driver. 
    
A more accurate assessment may be "my driver is under development" _ 
although the code was essentially done before I got my sample
hardware, now that I have the sample hardware and before the first
public release, I'm doing some radical rearrangements on the guts
so that it's cleaner and avoids some nasty race conditions.

The "rather fast" bit is a bit of an understatement - the time 
needed to start a SCSI command is measured in nanoseconds rather
than milliseconds as with other SCSI bus masters, there's no limit 
on the number of discontiguous scatter/gather buffers, with a 40Mhz
clock you can do 10M/sec sync, 6M/sec async, etc.  The chip's
also very robust, featuring some amount of onchip filtering and 
active negation so you shouldn't have the cable problems that 
plague the newer Adaptec boards.

    I see no reason for not going along with this unless
    the NCR chip is particularly expensive.

It's a nice chip, fast, elegant, requires no glue to plug it 
into the PCI bus.

Cost should be in the $50-$60 range from the rough pricing estimates
I got some time ago, I'm not sure where the price is now.
    
--------

<Prev in Thread] Current Thread [Next in Thread>