Question about the simulation frequency/fps

The actual beam calculations are a significant % of the whole CPU workload, but considering the big picture it is a small part of the whole (collision detection for example is ~30% by itself). Even so, the original version of the Rigs of Rods beam core (where we were previously working on) was actually trying to reduce the updates on the beams that were relatively "calm" (until i removed the checks after doing some benchmarks), in a similar manner to what you are suggesting above. This made sense in the older CPU generations where "checking" (conditional "if" in programming) to see if something is "calm", is computationally cheaper than doing the actual physics computation.

In newer CPU generations, the "checks" are quite costly and CPUs have become a lot faster in number crunching. This means that now-days it is cheaper to not do any checks and just blindly do the actual physics computations.

Concerning the optimizations of the physics engine, it is very heavily optimized but it still has room for more optimizations. In more detail, i don't expect the single threaded (simulation of a single vehicle) performance to change by that much. For the many vehicles case, there are still significant optimizations that we are looking into. Also there are some parts of the physics that we are planning to move to the GPU side. The main constraint right now is that the upgrade from DirectX9 to later versions will have to happen first.