opteron VS Itanium 2
rbw at ahpcrc.org
Thu Oct 30 11:07:25 EST 2003
On Thu, 30 Oct 2003 07:57, Andreas Boklund wrote:
>Just a note,
>> For bandwidth/memory intensive codes, I think the Opteron is a clear
>> winner in a dual processor configuration because of its dual channel
>> to memory design. Stream triad bandwidth during SMP operation is
>> ~50% more than a one processor test. Both the dual Pentium 4 and Itanium
>> 2 share their memory bus and split (with some loss) the bandwidth in
>> dual mode.
>This is true as long as you are using an applicaiton where one process has its own
>memory area. If you would have 2 processes and shared memory the Opt, would
>behave like a small NUMA machine and a process will get a penalty for accessing
>another process (processors) memory segment.
>To quote D. Barron, "If it seems to be to good to be true, it probably is!", i have never
>yet seen true linear scalability, and with Ahmdahl out there i doubt that i ever will.
Agreed. Of course, in the case of dual Pentium and Itaniums, even non-
overlapping memory locations buy you nothing bandwidth-wise. Small or large
scale perfect cross-bars to memory are tough and expensive. The Cray X1, with
all its customer design effort and great total bandwidth on the node board,
targeted only 1/4 of peak-data-required iin its design and delivers less under
the full load of its 16-way SMP vector engines.
And it's node board is probably the best bandwidth engine in the world at the
Beowulf mailing list, Beowulf at beowulf.org
To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
More information about the Beowulf