nVidia to acquire AGEIA

[Strider]

http://www.nvidia.com/object/io_1202161567170.html

SANTA CLARA, CA — FEBRUARY 4, 2008—NVIDIA (Nasdaq: NVDA), the world leader in visual computing technologies and the inventor of the GPU, today announced that it has signed a definitive agreement to acquire AGEIA Technologies, Inc., the industry leader in gaming physics technology. AGEIA's PhysX software is widely adopted with more than 140 PhysX-based games shipping or in development on Sony Playstation3, Microsoft XBOX 360, Nintendo Wii and Gaming PCs. AGEIA physics software is pervasive with over 10,000 registered and active users of the PhysX SDK.

"The AGEIA team is world class, and is passionate about the same thing we are—creating the most amazing and captivating game experiences," stated Jen-Hsun Huang, president and CEO of NVIDIA. "By combining the teams that created the world's most pervasive GPU and physics engine brands, we can now bring GeForce®-accelerated PhysX to hundreds of millions of gamers around the world."

"NVIDIA is the perfect fit for us. They have the world's best parallel computing technology and are the thought leaders in GPUs and gaming. We are united by a common culture based on a passion for innovating and driving the consumer experience," said Manju Hegde, co-founder and CEO of AGEIA.

Like graphics, physics processing is made up of millions of parallel computations. The NVIDIA® GeForce® 8800GT GPU, with its 112 processors, can process parallel applications up to two orders of magnitude faster than a dual or quad-core CPU.

"The computer industry is moving towards a heterogeneous computing model, combining a flexible CPU and a massively parallel processor like the GPU to perform computationally intensive applications like real-time computer graphics," continued Mr. Huang. "NVIDIA's CUDA™ technology, which is rapidly becoming the most pervasive parallel programming environment in history, broadens the parallel processing world to hundreds of applications desperate for a giant step in computational performance. Applications such as physics, computer vision, and video/image processing are enabled through CUDA and heterogeneous computing."

AGEIA was founded in 2002 and has offices in Santa Clara, CA; St. Louis, MO; Zurich, Switzerland; and Beijing, China.

The acquisition remains subject to customary closing conditions.

More details about the acquisition will be provided during NVIDIA's quarterly conference call, to be held on Wednesday, February 13, 2008 at 2:00 PM, Pacific Time. The Company's prepared remarks will be followed by a question and answer period, which will be limited to questions from financial analysts and institutional investors. To listen to the conference call, please dial 212-231-2901; no password is required. The conference call will also be webcast live (listen-only mode) at the following Web sites: www.nvidia.com and www.streetevents.com.

Replay of the conference call will be available via telephone by calling 800-633-8284 (or 402-977-9140), passcode 21354792, until February 20, 2008. The webcast will be recorded and available for replay until the company's conference call to discuss its financial results for its first quarter, fiscal 2009.

Very interesting. Maybe we can expect Graphic/Physics combo processors in the future?

 

[GenMarshall]

What applications would a Physics processor have use for, other than Gaming?

 

[Strider]

What applications would a Physics processor have use for, other than Gaming?

For the average users possibly desktop effects, but the major effect will be for people who create animated videos and the like on their computers.

 

[GVBN]

Can't you just use CUDA to calculate physics? Or am I missing something?

 

[Speedo]

Can't you just use CUDA to calculate physics? Or am I missing something?

Problem is that only the highest end GPUs have any horsepower to spare for physics work when they're also rendering a game. And even that isn't always true... example being Crysis.

 

[warpus]

Can't wait for the chemistry cards to start coming out.

 

[mdwh]

Problem is that only the highest end GPUs have any horsepower to spare for physics work when they're also rendering a game. And even that isn't always true... example being Crysis.

Yes but it may still be more efficient to do it on the same chip, just by having more cores, than having to have two separate cards or processors.

If you were willing to buy two cards, it'd surely be better to have two graphics cards via SLI that also did the physics on the GPU, than two have one graphics card and another card that could only do physics.

And since GPUs are already general purpose processors, it's not clear how a "physics/graphics combo card/processor" would differ from simply a new GPU that had more cores. I guess it might be that you no longer have to fudge the physics calculations through a graphics API that deals with things like vertex and pixel shaders. But I don't know if that's much difference on a hardware level (especially as we now have unified shaders rather than separate shaders for vertex and pixel programs).

Essentially, a GPU is a processor that is very good at parallel processing (it just so happens they are mainly used for graphics, but they can be used for physics), whilst a CPU is very good at running single threads as fast as possible. It's not clear to me what defines a "physics processor", and how it differs from these two extremes?

 

[Speedo]

Essentially, a GPU is a processor that is very good at parallel processing (it just so happens they are mainly used for graphics, but they can be used for physics), whilst a CPU is very good at running single threads as fast as possible. It's not clear to me what defines a "physics processor", and how it differs from these two extremes?

It's hard for me to believe that a GPU will be good at all the physics details you'll be looking at in the future - rigid & soft bodies, collision detection, hair/cloth sim, finite element analysis, etc - as a dedicated PPU tweaked for the purpose. I'm sure what you'll see within the next few years is graphics cards that included a dedicated PPU. It's pretty much the only way that PPUs are going to make it onto the market in any real numbers.