We are running a number of large runs to optimize share of preference using exhaustive search.
The runs were taking extremely long (3+ days) with our computer here, so we decided to configure an instance of a really good machine on Amazon and run it there.
We set up a machine with 36 vCPUs (threads) and 18 cores @ 3.00 GHz, and 72 GB of RAM. The machine has an SSD so any read/write features would be fast.
However, even with this powerful machine, we haven't seen any improvement in our performance. (Interestingly for the first couple hours, all CPUs were running at 100%, but now it seems to be at 33% overall. Memory isn't an issue. But it isn't faster.)
We can easily set up another Amazon instance if we knew what the secret would be to get better performance. I have found some other comments on here about optimal system configuration, but they are old (and don't seem to match with the failure to improve that we have experienced). Does anyone have any suggestions?