I was trying to optimize this strategy on NT7, with the default optimizers (brute force and genetic), and also with Piersh GO, and found that only less than 2 cores were used by the job.
Does anyone know something about this ?
It seems here that one core is used for the backtest itself, and the other for keep the charts updated.
With some other platforms, if you have a tons of cores, the optimizer use all of them during this process...

It will depend on the strategy itself based on my experience.

I've got another strategy for instance that doesn't call any external indicators, and it uses about 90% of my eight cores whereas the one you mentioned only uses about 40% of eight cores.

So I think it's to do with the multi-thread syncing, the more external calls to indicators, the more it has to be sync'd.

Also, the first pass of each dataseries takes longer as the bar cache is generated. Subsequent passes are far faster. For instance, if you are optimizing on multiple dataseries (chart input source), it can really slow things down. If you are only parameter optimizing, then after the first run and the bar cache is built, it generally picks up some speed.

Mike