This is just a little note. Ever since the new GUI was implemented, no matter what game version, once CPU usage started getting maxed out, for example, when running two or more cars per core on a 4690k, the GUI was unusable. It would lag so much that it would be next to impossible to add more cars, while the physics were running just fine. Testing BeamNG in a QEMU/KVM on Linux with VFIO patched kernel for GPU passthrough and IOMMU fixes, this was not the case. I was able to use 11 vehicles with the UI lagging only slightly. BeamNG warned me about the game not functioning well on a single physical core, though in fact all four physical cores are being forwarded natively with intel's virtualization features. I don't know enough to tell what kind of thread scheduling differences might be happening due to this, but I think it's interesting. Below is just a generic vehicle benchmark that spawned my interest and subsequent frustration in using mass numbers of vehicles in the latest update back when I ran Windows native. This benchmark is under Linux, which doesn't detect that the CPU is running at 4.6GHz, which it was at the time I was testing this natively on Windows. http://i.imgur.com/D6ENa3J.png
Run the GUI thread at high priority and put the BeamNG thread to above normal, then bobs ur Uncle and it's faster, but you might loose game performance
I'd imagine, yaknow, such a thing should be done by default, cause you wouldn't be expecting 100% performance for most people once you get to ten or so cars.