An SMP / Multithreaded type science computation is in development over at another (starting?) project, and BOINC is only learning about this since recently, so NO, you can't.

Sorry

I think (not sure) what he meant was can you split a single running science app into 2 threads instead of one and run it on different cores at the same time. That way, in theory, you would complete the task in 1/2 the time.

As Sekerob said, that would require symetric multithreading. This is not to be confused with simultaneous multithreading (as found on new Intel i7 chips as well as older architectures). Both have the same acronym so it gets confusing.

Apps have to be specifically coded for the first kind of SMT. This is regularly done with apps running on super computers but is still uncommon for end user apps. The program has to be structured so that different parts of the processing don't have any dependence on the results produced by other parts of the program. So it is fairly straight forward for apps that are easily broken down into independent segments - things like telemetry processing, graphics encoding, etc.

I don't know how linear the science apps here are. In other words, can you simultaneously try different ligands on the same molecule or are the calculations interdependent. IDK.

These checkpoint save every minute too, in fact if you let it, the default "at most" is 60 seconds, but HCMD2 can do them every few seconds depending on the position on hand ;>)

Caveat: The oh so wise of Berkeley have made the checkpoint saving a BOINC wide setting in 6.6. It was per-science. So 60 seconds on an Octo is now 8 minutes, at most... just that you know if running with 6.6 :O|