Linus Torvalds (torvalds@linux-foundation.org) on 2/7/09 wrote:
---------------------------

>But look at it this way: if you're trying to convince
>somebody to stray from the POWER fold, what the hell has
>IPF to offer? Absolutely nothing. This whole argument that
>IPF is competing against POWER seems broken to me. If it
>is, it is losing.
>
>POWER has better performance, a better infrastructure, and
>a vendor who doesn't screw its customers over like clock-
>work every decade.

Ahahahaha, that's really funny. Talk to some AS/400 users
about how non-screwing IBM is. Or mainframe customers
who had hoped for restoration of multivendor competition
with the likes of PSI.

In the early days Power had many advantages over IPF but
with each new generation of IPF more of these advantages
disappear. Montecito added multi-threading and dual core.
Tukwila adds huge amounts of bandwidth from integrated
memory controllers and system routing fabric. Poulson
will catapult IPF from usually trailing Power in process
technology to leading it. Where can IBM take Power to stay
ahead of IPF? To 300W+ chips packaged five at a time in
1600W MCMs run inside refridgeration loops with wind
tunnel backup cooling like z10?

Looking ahead it appears that IBM is strongly redirecting
Power towards the HPC market and preparing mainframes
for center stage role of running mission critical commercial
workloads. If that is true then IMO IBM is acknowledging
the war between IPF and RISC for general purpose high
end servers is almost over and IPF will win. IBM just can't
compete with Intel processes, fabs, or merchant processor
business model *head on* which is why it has spent the
last decade transforming itself more and more into a
software and services giant that dabbles in chip making
and low volume, high margin proprietary ISA servers on
the side. (IBM's x86 server business is in such disarray
one must wonder if its days are numbered).

Big system users don't switch architectures on whim, they
really do "marry" them for 10 or 20 years or more. Someone
choosing a platform today will ask themselves "will Intel
still be in the chip making business 10 and 20 years from
now? What about IBM?"

IMO buyers see IPF and z mainframes as pretty safe long
term bets (mainframes processors are distinguished by
their ISA and software base, not computational muscle;
being fabbed by IBM Micro vs TSMC doesn't change the
value proposition of the platform much), Power is less
certain, and SPARC has red flags all over the place.

And x86? Well x86 servers will continue to leverage the
processor cores, process techology, and OSes bought
and paid for by PC MPU sales and own the vast majority
of the entry level server market. The same Intel process
technology leveraged by IPF thank you very much.