Wisecom

Douglas Eadline deadline@plogic.com
Tue Jan 5 10:22:41 1999


On Sun, 3 Jan 1999, Steven N. Hirsch wrote:

> All,
> 
> Ok, after Mark Martin was kind enough to post Wisecom's web site, I tried
> to grab a data sheet from it.  No joy, as the link is broken.  So, I next
> tried the two retailers listed.  Again, no luck.  Neither one will admit
> to carrying their network cards.
> 
> So, back to square one.  Anyone know where I can mailorder a PCI 10/100
> ethernet card with a real Tulip chipset?


We will be posting an announcement about this in a few days.

First a little background, Paralogic builds "turn-key" Linux
clusters (beowulfs).  Obviously we are very concerned about 
tulip availability. We have found both the Intel and Tulip
NICs to be very good for Beowulf use.  However, we have made the
following observation.  The Intel NICs have a single "blood-line"
and as such the driver does not have to track constant 
releases of the new hardware (read as single sourced).
The tulip on the other hand, while 2-3 times less in cost,
seems to have multiple "blood-lines" and as such, Don is constantly
"#ifdef"ing the driver. Indeed, the industry switch to the
Lite-On has cause considerable havoc for people like us that
buy NICs by the case.  (Particularly when the product number
stays the same! - which leads me to the conclusion that commodity
hardware is more defined by "a windows driver exists" than the 
actual hardware in the card)

In any case, we have found a supplier (KTI networks) of DEC 21143 NICs
that (so far) seem to work with driver .90f.  (Don worked on
the driver at SC 98 in our booth). Later this week we
plan to announce the availability of these NICs for
about $39 each.  KTI makes other NICs so just ordering a KTI
100Tx NIC is no guarantee that you will get a tulip.  

So as a service to the community (we just want to cover our administrative
overhead), we will offer a tested genuine DEC tulip NIC (21143)
for those in need of the "real thing". 

BTW, we have been running these NICs using a variety of tests
(netperf, netpipe, NASA parallel) and they seem to work just
like the "tulips of old".  

If you have further questions, please contact me directly

deadline@plogic.com

no need to clog the list with this thread.


Doug

-------------------------------------------------------------------
Paralogic, Inc.           |     PEAK     |      Voice:+610.861.6960
115 Research Drive        |   PARALLEL   |        Fax:+610.861.8247
Bethlehem, PA 18017 USA   |  PERFORMANCE |    http://www.plogic.com
-------------------------------------------------------------------