Good morning,

I need to increase the size of the model I simulate: I used to work with a 3.5millions cells model and a simulation used to take 12 hours with a Xeon quad core 1.86Mhz and 8Go of memory (7.2 used during the simulation).

I do not want my simulation to take more than 17 hours for a model of 10 millions cells. I wanted to upgrade my work station to a Xeon X5365 3Ghz and add 8Go of memory more.

Could anyone tell me if this configuration is okay?

Thank you very much.

How could anyone possibly answer this question? Where is the bottleneck in your current model? Is it memory bandwidth (very common)? Then the speedup will be proportional to FSB speedup which is probably much less than the CPU speedup. How is the algorithmic scaling for your solver? Is it really mesh-independent? Is any part of your simulation explicit? If so, increasing the resolution will increase the local CFL number so you may need to do many more time-steps. If your larger mesh is adaptively refined, this will be worse than uniform refinement.

In general for meshes of that size consider using a cluster might make sense.

Try to get more cache memory. 12MB would be great

This is highly dependent on the workload. Bigger cache is not always better.