Cheap, good tulips...
Bob Drzyzgula
bob@drzyzgula.org
Thu Jan 7 22:43:20 1999
On Thu, Jan 07, 1999 at 05:13:50PM -0600, alex wrote:
> I also wrote to Wisecom, and I got a response. Unfortunately, I think
> they answered me because they thought I was building a cluster.
Has anyone contacted Intel and/or Dig^M^M^MCompaq?
I seem to recall those companies having at least a passing
interest in Linux. It seems to me that we would have a
few facts to present to them:
* The 21140A is one of the best, if not the best 10/100
chip for Linux. The reasons for this include the
excellent design and the availability of good
programming information.
* The 21140A is commonly used, when available, in the
construction of Beowulf clusters; thus customers who
buy them often purchase dozens at a time.
* While the i82558 is a strong alternative to the 21140A
for Linux applications, a variety of factors
(lacking Don's perspective, I hesitate to enumerate...)
have resulted in the 21140A remaining the superior
100Mbps Ethernet solution for Linux clusters.
* In the months following Intel's acquisition of Digital's
Semiconductor operations, a number of popular 21140A-based
cards have been redesigned with alternative chips, often
clones of or workalikes to the 21140A. The results in
Linux have been mixed, but generally most designers of
Linux systems and clusters would much prefer to have
a genuine 21140A in their systems. Nevertheless, good,
inexpensive, generic 21140A-based cards are becoming
increasingly difficult to obtain.
It seems to me that it might possibly be worth the effort
to contact Intel and Compaq directly with this background,
and to respectfully suggest that we could use a little help
here. We are, after all, just trying to purchase their product,
and are having a difficult time doing so. Things that they
could do to help might include:
* Release of a Compaq- or Intel-branded 21140A-based
PCI 10/100 Ethernet card, hopefully at a low to
moderate price point ($25-40). Possibly doing so
for sale by brand houses (like Hi-Val, for example...
companies that don't make anything but packaging)
would be an OK way to do this.
* Inclusion of the 21140A chip on select Intel and
Alpha motherboards. (Alphas may already have this,
I wouldn't know).
* Giving someone like Jaton (who never saw a video
chipset they didn't like and who can probably
design a PCI card in their sleep) a reason to build a
straightforward 21140A-based card.
I wouldn't think that any of these actions would cost
those two companies all that much, and it could result
in quite a popular product.
We could perhaps all do this individually, or alternatively
we could send a jointly-signed letter making such a request.
It would be great if we could collect data from everyone
on the list regarding the volume of such product they
would be likely to purchase on an annual basis, along with
statements regarding sucess or lack of sucess with Tulip
clone chips as well as the 82558/82557.
I would be happy to draft such a letter, especially if
interested parties were to give me input on what they
would like it to say.
Opinions/Thoughts/Comments/Fla^H^H^H^H?
--Bob
--
============================================================
Bob Drzyzgula It's not a problem
bob@drzyzgula.org until something bad happens
============================================================