[Top] [All Lists]

Re: ARC 100 Prices

To: andy@waldorf-gmbh.de
Subject: Re: ARC 100 Prices
From: Drew Eckhardt <drew@romeo.cs.Colorado.EDU>
Date: Thu, 27 Jan 1994 21:13:21 -0700
Cc: riscy@sunsite.unc.edu
In-reply-to: Your message of "Thu, 27 Jan 1994 20:59:11 EST." <9401260815.AA12641@resi.waldorf-gmbh.de>

    > I believe Drew offered to handle the linux scsi driver for the NCR53cf94 
    > (very nice chip from what I hear) chip if we decided to use it.  I'm not 
    > sure this offer still stands.
    Someone wrote that the Emulex FAS216 is compatible with the NCR 53cf94.

Could be.  Emulex makes clones of a few of the low-midrange NCR chips,
as well as their own high-end SCSI chips.

Unfortunately, when I asked for people to commit firmly to using one of the 
chips in the NCR53c90 family, I saw only one response and nothing out of 

With no consensus, I wasn't about to waste my time on a driver that 
wouldn't be used.

I made the offer when I still had plenty of free time - enough time 
to wire wrap a prototype board, write the driver, and debug it,
and wasn't working on any other interesting projects.

Now, I'm working full time as a programmer, if I wasn't I'd still have 
enough consulting to keep me busy, I'm in the middle of snowboarding
season (12" of fresh powder on Loveland Pass!), and am working on the 
NCR53c800/700 series driver under Linux.

The marketoids bill the NCR53c810 as having an onboard "2 MIPS processor",
it's busmastering, runs code out of main memory, takes about 600ns to start
a SCSI command, has no limit on the number of discontiguous scatter/gather
buffers, and can do just about everything without host processor intervention.

And writing a SCSI driver that runs mostly on the SCSI chip, with a 
scheduler to handle multiple threads of execution (one per command),
is much more fun than writing a SCSI driver for a simple chip that
has nothing more than a simple onboard sequencer :-)

Not to mention the fact that hardware was provided for the NCR53c810
driver, meaning a hot 486 PCI board (Real nice.  0ws cache, interleaved
main memory so writes alternate between 0 and 1 wait states, allegedly 
transfers from main memory to PCI video boards at > 30M/sec) :-)
<Prev in Thread] Current Thread [Next in Thread>