[gmx-users] Re: latest benchmarks (Itanium/Xeon/Opteron)

[Erik Lindahl]

Hi,

>
> OK, Thanks Erik for the prompt reply. Can I add the factor 2 to your 
> quite
> by David that "According to Erik Lindahl a 900 MHz Itanium is slightly 
> faster
> than a 3 GHz Xeon for gromacs (next version)", so that a 1.4GHz 
> Itanium2 would
> about three times the speed of a 3GHz Xeon?
>

No, IIRC that was with the assembly loops, but for a water system - for 
a non-water system the performance on Itanium is relatively better 
compared to Xeon.


> Then remains the relative performance of Xeon versus Opteron. Any 
> thoughts
> on that anyone?

Until the AMD issues with SSE are sorted out I won't recommend Opteron 
to anybody. It's not horribly bad, and you'll always be able to use 
3DNow, it's just that the Xeons work 100%, and they are not that 
overpriced anymore.

>
> We're reviewing several hardware offers now, production simulations 
> won't
> start in the next few months, so I could assume the remaining hitches 
> in the
> assembly code and hardware specs are sorted out before then... ;-)
>

Apparently IBM have started shipping 90nm versions of the PPC970, so 
there are rumors Apple will release faster G5 machines very soon. A 
dual 2.6GHz G5 would be a pretty nice machine - 64-bit, huge memory 
bandwidth, and about as fast as Itanium2 (faster than Xeon or Opteron).

Cheers,

Erik