[Beowulf] Re: dual core (latency)

Vincent Diepeveen diep at xs4all.nl
Tue Jul 19 05:27:48 PDT 2005

At 10:31 PM 7/18/2005 -0700, Greg Lindahl wrote:
>On Tue, Jul 19, 2005 at 06:42:02AM +0200, Vincent Diepeveen wrote:
>> You keep mixing latency and bandwidth. That's ugly, to say polite.
>If you're through berating Stuart, can you be bothered to explain why
>you think TLB misses should be so cheap? Many architectures have much
>worse TLB miss performance than the Opteron. Go bother the Power5

Point is that latency to a single core in a dual core dual system is far
slower than a single core in a dual system or a single core in a quad system.

Idemdito a single core in a quad dual core system is also dead slow to access.

In short a TLB miss at a dual core has a larger latency than a TLB miss at
a single core, also when the other cores aren't busy.

This for a read, which seemingly cannot be done in parallel by dual cores.

>-- g
>Beowulf mailing list, Beowulf at beowulf.org
>To change your subscription (digest mode or unsubscribe) visit

More information about the Beowulf mailing list