Endgame
by: ptnewell 10/5/00 3:00 pm
Msg: 168512 of 168534
The Rambus investment thesis depends upon gaining royalty rights
to a significant share of the DRAM market, either through the success of RDRAM
or through the enforcement of broad patent rights essentially covering
all current and proposed forms of high speed memory. Lately discussion
has centered on the latter approach, stressing vigorous enforcement of IP.
However the spectacular collapse of the recent DDR "releases" raises the
question whether victory through market share will moot the legal battles.
Is DDR dead? Certainly not. The correct term is "stillborn". DDR, after
all has never drawn the breath of life outside graphics cards.
Many DDR skeptics doubted whether 266 MHz DDR was possible, and whether
even at the 200 MHz level a DDR motherboard could support more than two
DIMMs. Of course if the address lines are registered, a larger number of
DIMMs could be supported, but this would add considerably to the cost
(and incidentally introduce additional latency).
But reviews by Anand and SharkyExtreme of chipsets by Acer and Via
show that even the bearish view of DDR capabilities was too optimistic.
Even running at 200 MHz, these boards were so unstable that only a pitiful
few benchmarks could be completed. Sharky actually described going through
"hell" just to complete this token task. The crowing blow, of course, is
that the effort required was wasted: the benchmark scores are shockingly
low. Meanwhile the number of aborted DDR launches continue to rise,
as Apple missed a supposed launch of a 266 MHz DDR chipset in August,
Transmeta is proceeding with its launch using ordinary SDRAM, and the
first version of Intel's Foster is now RDRAM only. HotRail abandoned
its efforts to develop a DDR server for AMD, whether for a "better business
opportunity" as stated or otherwise.
The DDR proponents have always had a "you first" mentality. Each thought
DDR was a great idea, if it was going to be cheap and plentifully available,
with lots of other chipsets out by the time they launched their own efforts.
But now AMD is up against a hard place. If DDR is to succeed, it has become
obvious that they will have to go first. Yet this involves high costs,
especially at the outset. And of course the problems encountered by Via
and Acer, as well as the non-standardization of DDR faces them.
A speculation: suppose AMD is in talks with Rambus on the production of
a RDRAM chipset? We know that AMD initially hoped that Via and Acer would
produce Atholon DDR chipsets for them (originally by the time the Thunderbird
arrived in June). AMD hoped to get out of the chipset business, now they
know better. We know AMD's hopes for HotRail failed. We know that in the
spring AMD began running a series of advertisements for engineers with
RDRAM experience.
If I were at Rambus and saw such advertisements (and they must have),
I would have called AMD and volunteered RMBS staff effort. AMD might have
been reluctant to accept given the ties between RMBS and INTC. Yet we
have recently learned that AMD and Rambus are in talks. CMP media
speculation is that these talks refer to licensing SDRAM. But AMD will
only say that they are an RDRAM licensee, and therefore have cause to
talk to Rambus.
The CMP interpretation is certainly a quite likely one. But it is not
the only possibility. If AMD has decided to go with RDRAM, they might
well eventually accept RMBS assistance (as Intel did). If AMD converts,
the game is over. The lesser players will abandon DDR in droves.
Rambashing may end in the ice of lawsuits or in the wildfire of exploding
market share. If Rambashing must die twice, ice is nice. But if Rambus
can only prevail once, let it me through the more satisfying fire. |
