As E3 shows us the fruits of game developers’ latest labors, Nvidia has served up a little something especially for those developers: a new version of its PhysX physics middleware SDK. The updated development toolkit, in Nvidia’s words, "has been over three years in the making and features a new modular architecture and a completely rewritten PhysX engine." It’s available now from this page.
Nvidia goes into more detail about PhysX 3.0 on its official blog. Additions in PhysX 3.0 include the ability to merge multiple actors into a single "aggregate . . . managed as a single bounding-box entity" to simplify collision prediction, more efficient streaming of actors into scenes, and improved artist tools.
Arguably more noteworthy is a new Task Manager and managed thread pool, which "allows games to take advantage of multi-core processors on all platforms." You might recall that, last year, we discovered that certain games completely fail to implement PhysX in a way that takes advantage of multiple CPU cores—or even modern instruction sets like SSE. PhysX 3.0, it seems, is tackling that issue.
Perhaps the new multithreading goodness has to do with Nvidia’s greater ambitions for its physics middleware. The company says PhysX 3.0 targets a "broad spectrum of multi-core gaming devices – such as PCs, notebooks and gaming consoles, as well as emerging gaming platforms like handheld game devices, tablets, and smartphones." Failing to take full advantage of the CPU doesn’t make much sense when you’re dealing with handhelds with very limited computing resources.