May 2nd, 2000 -HotRail has announced it will discontinue development of
a chipset intended to allow Advanced Micro Devices (AMD) CPUs to be used
to build multiprocessor servers. HotRail has said it will refocus its
efforts on the networking sector.
Using HotRail's chipset with its own Athlon processors, AMD was expected
to break into the multiprocessor (up to eight CPUs) server market for
the first time, later this year. AMD is continuing to work toward getting
a chipset that will allow it to ship a dual-processor server later this
division planning manager Byran Longmire said AMD plans to come out with
its own two-way chipset later this year and expects to move beyond that
late next year when it introduces its 64-bit, next-generation processor,
code-named SledgeHammer. "We wish (HotRail) well, but we are disappointed,"
This really messes up AMD's plans for world domination of the PC and server
world. In an earlier piece, we estimated that it would take AMD a year
to be a credible server supplier. Even though AMD plans to release a dual-CPU
chipset, this market segment is a relatively-low-margin segment - not
as low-margin as traditional desktop PCs, but significantly lower than
four-CPU and eight-CPU systems. Unless AMD has already been working a
dual-path scenario with Reliance Computer/ServerWorks, this stretches
out that timeframe for the larger machines for at least another year.
Although AMD has made great strides with its PC microprocessors, it needs
the high-margin server market to enhance its long-term viability. They
have become profitable by focusing on the PC space, and will probably
continue in that vein, but adding servers to the revenue stream means
they can generate more cash to help sustain the low-margin PC space.
understand HotRail's desire to jump on the networking/telecom bandwagon,
but this indicates someone in AMD may have taken his eye off the ball
in the HotRail deal. There are two alternatives regarding AMD's situation:
either they anticipated this drastic change and have already taken steps
to minimize its impact; or they were blindsided, and are now in trouble.
If it is the former, then we expect to see Advanced Micro announce a development
relationship with Reliance Computer (or someone similar) in the near future.
If it's the latter, then they need to find someone ASAP, and commit the
necessary resources (human and $$$) to the project. AMD-philes can only
hope this wake up call will help AMD refocus its efforts.
Users who want an alternative to the Intel server hegemony will have to
wait awhile longer. Although AMD could (technically) field a single-processor
server now, that particular market segment (except in special circumstances,
such as server appliances) is dead or dying. The talked-about dual-processor
server should be tested extensively (and exhaustively) by potential users
before committing to a large order.
For now, Intel-architecture customers will need to stick with Intel. This
is not a big deal for most, if not all, of the current customer set. Intel
generally delivers quality products, even though there have been some
missteps with things like the Profusion chipset. (Ref. TEC News Analysis
Dell Announce Eight-Way Intel Servers, published in September 1999).
The only immediate benefit we expect AMD's servers would
provide to a customer is the expected price reductions that would result
from competition. Given the generally lower pricing of Intel systems as
compared to Sun/Solaris or Unix on IBM or HP, pricing may not be as important
as it is in the desktop PC market.
We are resisting the temptation to talk about longer-term
benefits, because it is still unclear whether AMD will actually ship something
in this greater-than-two-CPU space, ever. Because we like marketplace
competition, we hope AMD succeeds, but they need to "get serious" about