[gmx-users] efficiency in HPC

[Albert]

On 08/29/2013 10:31 PM, Prentice Bisbal wrote:
> Albert,
>
> It looks like you are trying to compile for an IBM Power 7 system. 
> What is the 'other machine' you are comparing it to, and did you use 
> the same number of cores in each case?
>
> Prentice

yes it is. here is the configurations:

  * 4 IBM POWER7 processors; detailed configuration:
      o architecture ppc64 (specification Power ISA v.2.06)
      o 64 bit
      o 8 cores
      o core clock rate 3.83 Ghz
      o up to 4 SMT threads per core
      o 64 kB (32 kB instructions + 32 kB data) cache L1 per core
      o 256 kB cache L2 per core
      o 4 MB cache L3 per core
  * SMP
  * 32 cores
  * 128 GB RAM
  * theor. peak perf. 980 GFlops/s
  * dedicated operating system AIX 7.1


I compared with the following one, obviously the following one hardware 
is poor than above one but efficiency is much better. There must be 
something wrong with the compiling....

     AMD Quad-Core Opteron 835X nodes
x86_64 architecure 432 nodes
16 cores each
16/32 GB per node for memobry

THX

Albert