[peterhess]

Hello,
I am not sure if this question is already asked here.
Anyway, I did not found it in my quick search.

Problem:
Slower parallel computing on multiple workstations than one alone.

Hardware:
- 3 identical Workstations
- On every one: 2 x X5680 (2 x 6 cores)
- On every one: 24 GB RAM
- Windows XP 64bit

Description:
When I solve a problem on one WS, then I am faster than on two or three.

Example:
2e7 elements
One WS for 1000 Iteration (1 x 12 cores) about 4 hours
Two WS for the same (2 x 12 cores) about 20 hours
Three (11 + 11 + 10 = 32 cores) about 15 hours

HP-MPI

All workstations are connected to the same router and sitting in the same room.

Network card 1 GBit/sec on all

Any suggestion??

Thanks a lot

Peter

[litterfree]

Originally Posted by peterhess

Hello,
I am not sure if this question is already asked here.
Anyway, I did not found it in my quick search.

Problem:
Slower parallel computing on multiple workstations than one alone.

Hardware:
- 3 identical Workstations
- On every one: 2 x X5680 (2 x 6 cores)
- On every one: 24 GB RAM
- Windows XP 64bit

Description:
When I solve a problem on one WS, then I am faster than on two or three.

Example:
2e7 elements
One WS for 1000 Iteration (1 x 12 cores) about 4 hours
Two WS for the same (2 x 12 cores) about 20 hours
Three (11 + 11 + 10 = 32 cores) about 15 hours

HP-MPI

All workstations are connected to the same router and sitting in the same room.

Network card 1 GBit/sec on all

Any suggestion??

Thanks a lot

Peter

DID you know MPI? the Message Passing Interface. that is, in parallel computation every core has message passing which costs computation time . the more cores the more computation time costs. so For a specific problem the cores are Not the more the better. just like the grid adaption issue. especially for a simple problem, a single core is just ok.you can imagine this simple problem: 1+1+1+1=? use a single core to compute it , or use four cores to compute it . which fast? absolutely the single core!

[sail]

can you please specify the details of the simulation?
the time to perform one iteration in the various configuration?
have you checked that the domain-splitting gives approximately an even number of cells on all cores?

i take as granted that no work was done on the various ws while the simulation was running and they have their own independent network. (ie the guy next to you was not listening to an internet radio or something