Dual core AMDs (was Re: [Gllug] Recording Ogg streams?)
Christian Smith
csmith at micromuse.com
Wed Jul 13 15:33:04 UTC 2005
On Tue, 12 Jul 2005, John Hearns wrote:
>On Tue, 2005-07-12 at 16:31 +0100, Christian Smith wrote:
>> On Tue, 12 Jul 2005, John Hearns wrote:
>>
>> IA64 was a move to be "beyond RISC", to make the hardware even simpler and
>> leave parallelism to the compiler, but Itanic has thus far failed to
>> deliver volume penetration due to cost and compiler complexity.
>
>Actually...
>
>we sell a lot of Itanium systems. Really, really.
>They are very good at what they do.
>Mostly our customers use them for pre-processing of fluid dynamics,
>generating the meshes for post processing on Xeon based clusters.
>
>There were great hopes for Itanium - technically they are the best,
>but again it is the volume versus specialism market, and they are not
>going to win. Also hardware wise they are 'server grade' platforms,
>so any ones I've seen are the Intel refernce design with hot swap
>everything.
I find it funny that with all the problems SUN has implementing SPARC
register windows, Intel go ahead and also implement register windows!
Maybe that's why Itanium clock speed hasn't scaled in line with the rest
of the industry.
Technically, they may provide the highest performance per CPU, but what
about performance per watt? Performance per $? I'd say an Opteron probably
beats it on both counts.
And even the instruction set, given a clean slate, is horrible with
arbitrary restrictions on what can be put where in the bundle. Sounds like
a compiler writers nightmare, and a reason why the performance can only be
achieved with limited optimised code:
http://www.extremetech.com/article2/0,1558,1155598,00.asp
>
>Compare with the Opteron - just install your existing 32 bit OS and run
>existing codes, then migrate when it suits.
>And we did just that with one of the first Opteron clusters in
>Manchester, in 2003.
>
Amen, though I'd take on of the new SGI Prism workstations if offered:)
Christian
--
/"\
\ / ASCII RIBBON CAMPAIGN - AGAINST HTML MAIL
X - AGAINST MS ATTACHMENTS
/ \
--
Gllug mailing list - Gllug at gllug.org.uk
http://lists.gllug.org.uk/mailman/listinfo/gllug
More information about the GLLUG
mailing list