Personally I see two companies that are so much centered into getting their opponent down that they forget to do some actual improvements that their should-be-buyers could actually profit from.
Nvidia has done virtually nothing since they bought up Ageia - and Ageia themselves hasn't done much more then developing the PPU and fitting API, since they got the Novodex-engine. So half a decade later we still stand with a engine, that can do hardly anything more, then go'ole Havoc, except calculating excessively high numbers of uselessly small increments, IF provided with excessively powerful, expensive and specialised hardware. So asside from some game developers enjoing Nvidias physics-engine-for-free offer, the whole project is nothing more then a nice topic for magazines to add in a couple of pages with those other benchmarks. To become a real revolution, one would require something that makes a real difference in gameplay, but creates only a 10-20% more load on existing hardware. This would allow game designers to integrate it mandatory, because it will run somehow, but at the same time it would force people to upgrade for best performance. Obviously, this would also require, that one can upgrade. Instead of replace. So no we-won't-allow-you-anything-as-long-as-you-have-bought-anything-from-any-of-our-competitors as Nvidia has been done for years now with virtually any development, that went beyond "display a picture".
But as long, as we only get 321235252 particles were once 1233 were sufficient and even this only on a few selected products and as long as we don't want certain other products...
Thats for Nvidia, who do a lot in marketing but nothing sensible in my opinion. (one might pretty well call their please-ad-laughing-real-to-emphasis-so-called-funny-part "cartoons" their most typical product)

As for AMD: All they do is bicker about Nvidia and release a bit of hot air, when nvidias marketing gets to much attention (read: every couple of days). The last time, I've seen something new coming purely from the direction of ATI, was back when they tried (and failed) to introduce their curved-surface implementation. The rest of the time they just made hardware, that fits a new DirectX level (okay: They also participate in the development of these, but there are several more players their und we would be only half as dependend on them, if ATI would put any efforts in OpenGL for their desktopproducts). Now this would be completely fine - we need companies that simply just offer a affordable, well desigend product capable of the basic functions. But AMD can't hold just one week from bragging, how they are the most advanced and revolutionary company to hit the globe since wheel.inc, sometimes resorting to pure insulting, if nothing else comes to minds (remember the real&faked quadcore campaign? Though the AMD vs. Intel CPU-bickering would be complete seperate, although quite similar story).


(please note: this personal opinion was asked for and is not based on representative tests, but on a couple of personal experience and mostly on the frustating pasttime "beeing mod in a forum full of nvidiafanboys and fanATIcs" that makes one feel every single press release like a slap in the face)

A much more interesting read: AMD and PhysX: history of the problem
And some figures: physxinfo.com/articles/?p...

Advanced physics in games are a growing market. AMD/Ati failed to jump on the boat and now they want "open" physics. Meanwhile the good games that use hardware PhysX are still few and far between. Will something new eventually emerge from all the corporate crap ?

I'm used to the term "propietary" applied only to interfaces that are limited to one manufacturer. So GLide was propietary-api to 3dfx because they didn't allow any other hardware to use it. PhysX on the other hand runs on PPU, Cuda, x86 and Power architectures and if I understood nvidia correctly, it could rund on even a lot more, if the manufacturers of these devices developed a physX implementation. (or, in the case of ATI, even if they "just" opened all the information necessary to implement Cuda on Radeon)

"open source" is familar to me for software, whose code is freely avaluable for anyone to use as they please.
So first its a lot opener: You may not just implement it in your own project, you may even enhance it, take only parts and generally profit in all thinkable ways of the knowledge and work of someone else, without giving anything back, not even your help to establish a widely comaptible basis. (taking the physX example: You would be allowed to use the full GPU-PhysX-kernel, rename all the commands so that it gets incomaptible and market it under your name -lets say: DirectPhysics- while at the same time forbid Nvidia to use your new command names, so they might at least have the marketing benefit of compatible hardware)
Secondly it applies to Software, not APIs. Softwarecode actually represents a lot of intellectual work. An API like PhysX on the other hand is just a defined set of commands. Using an existing set will not give a competitor any advantages or savings aside from beeing compatible with existing solutions (an advantage, that works both ways), but he still has to create his own code that will either give the defined commands or be able to do the requiered calculations. (In case of PhysX, the former one -the PhysX-physicsengine- can actually be licenced for 0$, but it is not openly avaluable. Though no matter how you stretch the term, you could not call it propietory, as it is issued to wide range of developers. On instance of the later one, the PhysX-hardware implementation in Nvidias drivers and GPUs on the other hand is highly closed. No one has access, no one can manipulate or adopt it for his own hardware. So you might call this "propietory" if you have to. But you would not address the PhysX-API in this case)



Actually most PhysX games to neither feature demanding fluid, demanding cloth or exceptionally demanding particle effects and behave exactly the same, wether you have a geforce, radeon or gma. (well: okay, with gma that usually behave exceptionally crappy graphic wise, but the physics stay the same )