[Beowulf] Barcelona numbers

Mark Hahn hahn at mcmaster.ca
Mon Sep 10 15:46:31 PDT 2007


> For many people here the random read latency to BIG ram sizes is quite 
> important.

are you suggesting TLB effects?  for 8 threads, the test is effectively
using 256MB, which seems plenty large to dilute the effects of 2x 2M 
caches and 8x .5M caches...



More information about the Beowulf mailing list