I should mention that Rambus screwed
up the engineering in two obvious places.
The first was that they wrote the spec in
such a way that RDRAM chips had to be 25%
larger than regular SDRAM memory chips.
There is some talk about redesigning RDRAM
to reduce that area penalty, but it is a
little late. Rambus has been working on
this stuff for 10 years, couldn't they get
it right already? You would think that die
area costs would have been something they
would have thought through. The only explanation
for them having missed this is that they really
don't have any competence in memory chip design.

It's as if total novices tried to tell the
industry what to do, and then got Intel to
back it with near monopoly muscle. This was
a stunning screw up, truly an amazing error.

But that wasn't the crowning screw-up. The big
one was leaving so little room for margin
(voltage and timing) in their interface that
the companies that had to build products to
that interface standard were unable to do so.
It was this error that cost Intel big time. I
doubt that Intel execs are returning much in
the way of telephone calls from Rambus right
now...

The world tends to blame the workers when a
company produces a product that is defective.
And the world has largely left the reputation
of Rambus intact, and has, instead, blamed the
chip makers, memory makers, and board makers
for the problems with systems using Rambus
memory. But those of us who work as design
engineers for a living know that the real
cause of most manufacturing screw ups is lousy
engineering. This is the cause of the destruction
of the reputation of RMBS among design engineers.
Bad design.

Rambus should take as a marketing slogan
something like: "Hire Rambus to Design Your
Memory Interfaces!", with a subtitle something
like "even losers need to have a chance." At
least the company that hires them won't be able to
blame their own engineers.

Rambus is quite dead. It may have a few more
twitches as it subsides into coma and decay, but
its future is quite obvious. The shares have been
largely shoved into the hands of mom and pop, and
mom and pop may run the short interest, but in the
end, the company is quite dead. The chance that the
memory community (or any other engineering community)
will ever again trust Rambus with the design of an
interface is zero. The royalty and fee income will
slowly decline, and Rambus will close its doors.

As far as the above prediction goes, we should note
that Intel sure seems to be producing a lot of those
little chips that make DIMMs look like RIMMs...