[lug] Linux hardware

[Scott Herod]

On Thu, 6 Mar 2003, Daniel Webb wrote:

>   I just put together a 10-machine cluster for research at CU.  We found
> that for memory-intensive search algorithms (particle collision), Intel
> wins by 50% over AMD.  For integration (ODE systems), AMD is twice as fast
> as Intel (integration loops are all in L1, I guess).  We don't even have
> hard drives in the machines (except for the master).  I think we built the
> whole thing for about $5000.  I would suggest either build it yourself
> from newegg.com, or find someone local who is an expert at this sort of
> thing.  Maybe I should go into business.
>

Out of curiosity, Daniel, what compiler are you using for the Intel and
AMD processors?  I recall hearing that Intel's floating point operations
are not nearly as good as AMD's but that gcc doesn't take advantage of
some of the AMD's capabilities.  Does that seem reasonable based on your
experience?

Scott