[mpich2-dev] processor comparison between Fusion and Magellan

[Rusty Lusk]

Thanks, Susan.   Yes, that would be a great thing to do, since it  
would demonstrate, at significant scale, just exactly what the  
overhead of virtualization is for an MPI job.  (Some previous  
publications say it is terrible, but the experiments were not entirely  
satisfactory.)   We'll let you know when we are ready.

Rusty


On Tuesday,Feb 16, 2010, at 1:02 PM, Susan Coghlan wrote:

> Rusty,
>
> We did get a faster processor than Fusion:
>
> Fusion:  2.53 GHz, 8MB L3, 5.86 GT/s, 1066 MHz DDR3, 80W
> Magellan:  2.66GHz 8MB 6.4GT/s 1333MHz 95W
>
> Perhaps, when you guys are ready, we could some Magellan time and  
> run it in basically regular cluster mode so that you could run  
> comparison performance runs on the same hardware?
>
> -Susan
>