- = Just the CPU-accelerated portion, though.
That's the fly in the ointment, as it were. Nvidia obviously doesn't want to give away PhysX completely, to allow AMD to highly parallelize their own version of PhysX. But at least this will allow AMD to implement their own version of PhysX, with a little bit of work of their own. So CPU-bound limitations of PhysX should be lessened thanks to Nvidia's move, and AMD should now be able to improve their CPU performance of lower-end systems running PhysX-enabled games.
This move should help reduce power consumption of CPUs, decrease CPU utilization in games (which should improve performance), making AMD GPUs more interesting to a wider audience. It's a great publicity move by Nvidia, after the whole GTX 970 3.5GB fiasco, but it seems to have been driven by the release of source code for Unreal Engine 4 and Unity 5.
I'm hopeful that this will not only improve performance on lower-end CPUs, but when combined with Vulkan (glNext, or future unreleased OpenGL version) and DX12, will decrease unnecessary CPU usage by passing on more work (like Physics and graphics driver work) onto the GPU, instead of the CPU. This offloading of work from the CPU to GPU should make gaming systems more affordable, allowing Nvidia to encourage sales of high-performance gaming systems, since a less powerful CPU will likely cost less. (This also gives head room for game developers to use more of the CPU for AI and other purposes.)
It could also allow more of a gamer's budget to be spent on GPUs, which is likely where Nvidia is hoping this move will go (along with the great publicity). While not as nice as the release of CUDA highly-parallelized PhysX source code, it's a step in the right direction.
Now if only we could do something about the release of their Linux and Windows drivers.