Newsgroups: comp.parallel
From: spike@trinity.llnl.gov ( Richard J. Procassini )
Subject: Re: Cray Computer is *NOT* Cray Research
Organization: Lawrence Livermore National Laboratory, NCD
Date: 04 Apr 1995 17:20:54 GMT
Message-ID: <3m95ha$kki@usenet.srv.cis.pitt.edu>

on 2 April 1995, thus spake Christopher Potter:

>It is true that Cray Computer (Cray-3,Cray-4) is bankrupt now, but
>Cray Research (Cray-YMP,Cray-C90,Cray-T90,Cray-T3D,...) is a *totally*
>different company and is not at all in a bad financial situation.

>Best regards,

>Christopher Potter                              
>CRAY T3D support/Parallel Applications Engineer 

	Well it is true that CRI (Cray Research Inc.) is not in BAD
financial shape, but attendees of the latest Cray Users Group (GUG)
meeting in Denver in March were shown a side of CRI that many users of
their products may never have seen before.

	CRI is certainly in a defining period of their corporate
existence.  They are attempting to change from a "we build the world's
fastest supercomputer PERIOD" company which gave us the Cray-1, Xray
XMP, Cray-YMP, Cray C-90 and Cray T-90, to a company that is
attempting to be many things to many customers.  They are attempting
to design, build, market, write software for and maintain four new
types of machines:  the CS6400 file server clusters, the J-90 class
PVPs (parallel vector processors), the T-90 class PVPs and the T3D/T3E
MPP (massively parallel processor) machines.  This increase in product
deiversity is coming at a time when the supercomputer market and the
company are downsizing.

	It will be very interesting to see if CRI can maintain it's
reputation for high quality products and support in this new era of
diversity and downsizing.  Trouble is already beginning to rear its
ugly head in terms of software support for the T3D MPP.

	CRI WAS planning on discontinuing software development for the
T3D in favor of starting development of software for the (relatively
different architecture) T3E MPP.  This statement was made several
times during the CUG meeting.  CRI software developers continually
asked the (somewhat irked) users of the T3D (including yours truly)
whether they would like continued software support on the T3D, or
stable software on the T3E, which is scheduled to ship in early-mid
1996.  Can you guess what the current users of the T3D replied?

	On should keep in mind that CRI sold 31 T3Ds in 1994 for a
total of $130 million: ~ 15% of their total sales.  CRI employees have
also stated that the T3D was really a prototype machine, and was never
planned on being sold to the MPP user community.  This is not a
loss-generating product, especially in light of the fact that CRI is a
relative newcomer to the MPP community.  It seems clear, at least to
me, that if CRI wants to see current users of T3Ds upgrade to T3Es in
the future, they shouldn't "orphan" the T3D before it's out of it's
infancy.

	To CRI's credit, they have APPARENTLY reversed their decision
to no long develop software for the T3D.  From an application
developer's perspective, this is a VERY wise move.

	So to sum up this rather overblown message (I didn't intend
for this to be soooo long at the outset), CRI is not at a point that
bankruptcy is imminent, or even a possibility.  On the other hand,
their current growing/shrinking pains they are experiencing in a time
of shrinking funds to buy supercomputers may be an indication of a
troubled future.

	So what do you think?????

--
				Dr. Richard Procassini
				Methods Development Group
				Mechanical Engineering Department
				Lawrence Livermore National Laboratory

