There is more and more excitement building up on an upcoming physics hardware acceleration engine. If this thing succeeds, it's going to be big. Anyone remembers 3dfx that revolutionized the game industry with their cards?

But, anyway, I just read three consecutive articles in ExtremeTech about Ageia, and I think it's interesting to see how the overall tone is changing over time. The first article is generally positive, but the next article ends up with a sarcastic remark:

Ageia claims that when add-in boards go on sale this fall, there will be 10 to 12 titles optimized for it. They may have great technology, and the PhysX chip might be the next "gotta have it" PC component, but boy, does the company not know how to make a launch in the game industry. To make a splash with gamers, you've got to show real-time demos that make people say "Wow!" You have to give them lists of AAA titles that will use your technology, quotes from rock-star developers about how it's going to revolutionize gaming. You've got to tell gamers how much it'll cost them. There should be before-and-after AVIs of a hot upcoming game all over the web, showing just how incredible it looks and reacts with the PhysX chip.

Ageia has done none of this, and a gaming public already staring down expensive video-card upgrades and next-generation console systems is understandably jaded.

The third article was written after E3, and the overall tone is different now. This time, the article starts with the fresh facts:

So far, ASUS is the first to announce a partnership with AGEIA. They'll have a PPU board with AGEIA's chip on the market in the fourth quarter of this year, with 128MB of GDDR3 memory, for roughly $249 to $299. Initially, it will only come only in a PCI card, with PCIe cards expected further in the future.

And some game vendors that signed up:

AGEIA was able to name some software titles that will enjoy physics acceleration, too. Rise of Nations: Rise of Legends from Microsoft, Ghost Recon 3 from UbiSoft, City of Villians from NC Soft, and Atari's Matrix license Path of Neo are some of the highlights they told us about.

And, finally, the first live demos:

The first demo was graphically simple, but still fairly impressive. A large rocky hillside had about 4,200 boulders dropped at the top, which all bounced, tumbled, and interacted in a realistic (and speedy) fashion. AGEIA claimed that a dual-core CPU can handle maybe 800-1,000 in a demo like this, but was quick to note that 4,200 boulders was nowhere near the capability of their chip. There's a driver issue right now where a lot of the timings need to be worked out between the massively parallel math units in the chip. Within a couple of months, the company will have a new driver which will enable them to raise the boulder count to 32,000. They're confident they can reach that number, but even if they can only get halfway there, 16,000 to 20,000 boulders is a lot better than a CPU can do.

The other demo was somewhat less impressive, showing particle-based fluid dynamics by displaying a shiny car that had fluid streams "sprayed" on it, like a primitive car wash. The particles could be change to plasma, soap bubbles, water, or whatever, but it honestly didn't look that great. Chalk it up to having programmers make demos, and not having real artists involved. The demo involved 6,000 particles, but again, that's the driver timing issue rearing its ugly head. The fixed driver in a couple months should be able to handle 40-50,000 particles.

I am not sure how things will evolve over time, but one thing is clear. Sooner or later, everobody will need physics hardware acceleration in their games. We'll see how this works out...