Back in 2007 PhysX Developers section of Ageia website has some disputable structure (for example, Physx SDK features description was missing, but there was large flash banner with several CG-sequences, not related to PhysX at all) – thus, I’ve decided to recreate this section from my “point of view” and send it to Ageia. And actually, some of my ideas (like feature list) were used later on.

So, just for amusement factor I’ve uploaded whole thing to PhysXInfo – maybe it gonna be interesting for someone, who is still curious on Ageia part of PhysX history.

Proceed to alternative Ageia PhysX Developer website

KitGuru website has brought us news that Manju Hedge, former CUDA and PhysX Solutions Vice President (previously – CEO and co-founder of Ageia) has left NVIDIA to join AMD.

Our own sources at NVIDIA are indicating  – this information is truthful.

However, according to our data, Manju departure won’t affect PhysX (he hasn’t been working on PhysX for over a year) or CUDA development process in NVIDIA, and his new roll in AMD won’t be connected to game physics related projects (instead, Manju is going to be involved in ISV recruitment).

Update: Pursuant to latest press-release, Manju Hedge will lead AMD Fusion Experience Program.

Update #2: from X-bit Labs article

In particular, [Manju Hedge] will manage the developer relations teams that help independent software developers (ISVs) to implement program code optimized for heterogeneous multi-core microprocessors.

We want to thanks Manju Hedge for awesome work on PhysX front and wish him best of luck with this new assignment !

Why ? Because, actually, hardware accelerated PhysX was criticized by AMD/ATI throughout it’s lifespan (more or less). Let’s take a retrospective look at escalation of the conflict:

In June 2006, just after Ageia PPU launch, ATI unveils their GPU physics processing conception called “Boundless Gaming” (also known as Asymmetric Physics Processing) based on Havok FX engine (technology was supported by Nvidia as well).

X1900 XT GPU was promised to deliver over 9x performance of a PhysX PPU card and games augmented with Havok FX physics were supposed to come out in 2007.

Result ? Havok FX died with Intel aquisition of Havok company, no games were released, 16 titles with Ageia PPU support from hardware PhysX side.

To be honest, AMD was threating PhysX enough politely those times, even considering to buy whole Ageia company:

Richard Huddy saying, “we’ve had that discussion, yes. It’s a discussion that goes round every three months – someone turns to me and says “why don’t we buy Ageia?”

Things started to get worse when Nvidia bought Ageia with all patents, personell, software and hardware developments. GPU PhysX was buried alive by AMD beforehand

“There is no plan for closed and proprietary standards like PhysX,” said [Godfrey] Cheng “As we have emphasized with our support for OpenCL and DX11, closed and proprietary standards will die.”

To support its position, AMD comes to GDC 2009 with GPU accelerated Havok demonstrations, this time with promises to port it to OpenCL platform.

Result ? 7 complitely new titles with GPU PhysX support in 2009, no games with OpenCL Havok, but “PhysX will become irrelevant” claims again and hollow “Accelerated physics processing” label to HD4xxx GPU series feature set.

Meanwhile, whole year passed since than, and now new Open Physics strategy is on the GDC 2010 scene, accompanied by more intensive GPU PhysX abuse

And where is GPU Havok now ?! Following Havok FX in it’s way to oblivion, I guess.

So, why do you think, readers – will this hardware accelerated Bullet & DMM duet took up this time, or it’ll end like other AMD’s attempts to settle on GPU physics field for past 4 years ? Are they really planning to bring peace and unity to developers absolutely disinterestedly, while Nvidia is using GPU PhysX to make money ?

While PPU is still good at Unreal Tournament 3 and Mirror’s Edge, it has insufficient power to run latest titles like Batman Arkham Asylum with smooth framerate.

You can read rest of the article here

Following Geeks3D report on that recent driver update disables PhysX hardware acceleration not only for Nvidia GPUs, but for Ageia PPUs as well (if non-Nvidia graphics card is present in the system) we’ve desided to perform small investigation.

Test methodology is simple enough: we’ll use certain sample from PhysX SDK 2.8.1 Build 13 to identify if Ageia PPU is working and initializing properly with different PhysX drivers on several systems with Nvidia and ATI GPUs.

Used PhysX System Software – 8.09.04 (last one with direct PPU support) and newest 9.09.0814 installed above 8.09.04 (PPU won’t be recognised on pure 9.09.0814 drivers)

GeForce System #1 – GTX275 + Ageia PPU

PhysX driver 8.09.04                                                         PhysX driver 9.09.0814

GeForce System #2 – GTX295 + Ageia PPU (thanks to JeGX from Geeks3d)

PhysX driver 9.09.0814

As you can see, in both cases Ageia PPU was recognised properly by the system, and hardware acceleration is working too (test applications indicates that both scene and cloth simulations are running in hardware)

What will happen if we’ll try same scenario, but now with ATI GPU installed in the system ?

Radeon System #1 – HD4850 + Ageia PPU (thanks again to JeGX from Geeks3d)

PhysX driver 8.09.04                                                         PhysX driver 9.09.0814

Radeon System #2 – HD4870 + Ageia PPU (thanks to heuM from Overclockers.ru)

PhysX driver 9.09.0814

While working fine with ATI card on old 8.09.04 drivers, test scene is running in software on CPU, when 9.09.0814 drivers are installed. But Ageia PPU is identified by OS, and hardware acceleration was working with same driver set and Nvidia GPU in the system.

Therefore, we can definitely say, that PhysX computing capabilities of Ageia PPU cards are disabled with newest PhysX drivers when non-Nvidia GPU is present in PC configuration. (at least with driver set we’ve used)

Certainly, PPU installement base is very low and it can’t handle well recent PhysX games, but such Nvidia’s decision looks strange and short-sighted even for us.