Who runs large capability jobs

[Mark Hahn]

> If you had only said "Note that the SGI won't scale beyond 512", you
> would have provided useful information and no one would have argued

actually, what always pisses me off about Greg's comments
is that they _don't_ provide any useful information,
except about his self-image.  if he said something like 
"oh, their bisection bandwidth is only 3.14 giga-trits/fortnight",
that might be useful.  or "they claim 400ns worst-case memory
latency, which sounds much better than multiple us latency
for a fabric like myrinet, but 400ns only happens if all other
CPU's are idle."  (note that all content except the 400ns
was from /dev/random.)

regards, mark hahn.