Non-Apple Ethernet cards (Re: Apple hardware)

[William Roberts;]

In <1991Jun14.052133.13353 at panix.uucp> alexis at panix.uucp (Alexis Rosen) writes:

>A few questions about the Tri-Data (I've never seen them):
>Does it have a 64KB buffer? As of A/UX 2.0.0, 16KB is "not supported" and
>could conceivably cause crashes (I think William R. wrote about this).

The 16K vs 64K issue is related to Apple's own cards (the old style ones). I 
can't recall what I said at the time, except that the 64K card made a 
significant difference to the NFS benchmarks I ran, but the eventual facts 
that emerged were:

1) Apple stopped shipping their card with 16K RAM and put 64K RAM on instead 
(this started at Rev L, if I remember correctly).
2) Apple didn't stop supporting the 16K cards - the same driver works for both 
cards since it probes the card at A/UX boot time to find out how much memory 
is available

The new style cards could well be different: they certainly have a processor 
on the card for A/ROSE purposes, but I don't know if they are otherwise 
register compatible with the old ones. I don't think that the A/UX 2.0.1 
drivers make use of the processor.

>Is it register-compatible with the Apple board? That's one reason I've
>stuck with the Asante boards- they can run with Apple's drivers, so I don't
>have to worry about the kind of thing that has happened to owners of the
>older Kinetics/Excellan/Novell/Dayna EtherPort cards (i.e., no support = 
>no drivers = can't use it under A/UX 2.0)

Once again - all drivers which worked under 1.1 still work under 2.0 and 2.0.1.
The difference between a "new for 2.0" driver and an older 1.1 driver is that 
the old drivers don't support EtherTalk Phase 2: they still work fine for NFS, 
IP and so on. My site still has about 50 old-style EtherPort II cards and is 
happily running A/UX 2.0 on them using the 1.1 drivers: we don't much care 
about EtherTalk in our environment so this doesn't cause us problems.

>If the Tri-Data board compares well with the Asante on these three points
>then it's a good card. I won't swear to it but I think I found that FTP
>between a Mac IIfx and a Sun IPC ran at about 110KB/sec with the Asante.
>Assuming that the IPC is as fast as the Sun 3, the Tri-Data may have an
>edge over the Asante. (Then again, that may be a very poor assumption.)

This is a very dubious comparison: a SPARCstation IPC is supposed to be at 
least 3 times as fast as a Sun 3, something like 12-16 MIPS. My NFS 
benchmarking work included a test of pure CPU+Ethernet performance without any 
disk involvement: this showed no improvement from SPARCstation 1 to 
SPARCstation 1+ (despite increasing the rating from 12 MIPS to 16 MIPS), and 
the fastest machine of all (296 units vs 273) was a Mac IIfx with the 64K RAM 
Apple card.


While we're on this subject - an old Kinetics EtherPort II card which has been 
at our dealers under repair for over a year has now been returned.... as an 
new Shiva EtherPort II card (not the same thing at all!). Does anyone know 
where I can FTP an A/UX driver for this card, or do I have to try explaining 
things to my dealer?
--

% William Roberts                 Internet:  liam at dcs.qmw.ac.uk
% Queen Mary & Westfield College  UUCP:      liam at qmw-dcs.UUCP
% Mile End Road                   Telephone: +44 71 975 5234
% LONDON, E1 4NS, UK              Fax:       +44 81-980 6533