Quoting Tracy R Reed <[EMAIL PROTECTED]>:
That is
completely different and does not suggest that breaking backwards
compatibility every now and then is a bad thing.
Technically this is true... from a business/market standpoint, go ask
Intel how they feel about that now. :P
Itanium has been out for years now and I have still never actually laid
eyes on one or seen an ad for one. I wouldn't know where to buy one if
I wanted it. No wonder it's dead.
Cost: Astronomically high for fast processors.
We have/had 13 total itaniums. They were niche machines doing the
last stage, high ram, heavy FP math stuff for chip designs. X86-64
beats them in every way now (except maybe FFTs still edges out on
ia64). But we needed the fastest cpus for our engineers (time is
everything) and they ran upwards of $11k per cpu! Well, that was HP's
cost but there weren't any competing smp high-ram setups. Of course
the RAM cost craploads too.
The 32bit support, sucked majorly. Even Intel would tell you to avoid
it as much as possible. Users would sometimes submit a job and their
wrapper would default to the 32bit binary without the right flag, and
they would freak when the job that should've taken 4 hours was still
running 24+ hours later, and nowhere near done. Of course, most jobs
on these hosts were multi-day level jobs which made such mis-steps
even worse.
As I understand it, the low-end, power saving IA64 processors aren't
all that bad cost wise, and could actually do pretty good at being a
web farm server or some such. But the x86 family overtook it so fast
anyway (which was the real failing of ia64 if you ask me.. it never
lived up to the speeds they were claiming) that there really just
isn't a driver to move to it.
--
Mike Marion-Unix/Linux Admin-http://www.miguelito.org
Principal Skinner: "Pull Willie, Pull!"
Groundskeeper Willie: "I'm doin' all the pullin' ya blouse wearin' poodle
walker." ==> Simpsons
--
[email protected]
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list
