Three things I will never buy......unless they are built into a mobo or video card.

Gee lets PC gaming even more complicated by adding three more pieces of hardware to maintain drivers for....and game programmers to program for. As if PC games were not buggy enough!!!

Future problem……”weird flashing and lag during online game playing of BF5” Tech Support: “Well have you checked for….OS updates, mobo BIOS, video card drivers, sound card drivers, physics card drivers, Killer NIC drivers, AI card drivers, game patches….are over clocking any of the above????”

The only thing this will do is accelerate console gaming. Especially if you consider the cost of having it all it. These three new add on cards will cost about $500 or more. Now your gaming rig with Quad SLI and XFI went from being 3 times as much as a Xbox 360 to 4-5times as much.

Oh and what sane person would buy that “Killer NIC”???????? It reminds me of that software you can buy to speed up the internet. I am sure that NIC and its software is more efficient than XP’s generic network stack…..but the speed of light is not going to get faster….nor is the distance to your game server.

My Kids and I play warcraft3 pretty often due to the varied gameplay different maps provide (DOTA, TD, Hero wars, etc) one of the map types we play is no build time no cost maps, so we easily have 400+ units on the map in a matter of minutes.. I think an AI accelerator would be great in this scenario as currently the game crawls to a halt trying to path and move that many units at a time.

I also agree with an ealier poster who talked about letting a server use this for pathing and such for computer AI, although this might create more network data as users would have to upload their commands to let the server path out the units and send that data back...

Although.. i don't actually know of any RTS games that use a dedicated server? Nice idea tho (for FPS yah, bf2 with 32 bots.. rawr).

JUST ANNOUNCED: PLANT GROWTH ACCELERATOR

Now you too can watch your simulated grass grow with stunning life-like realism! Also included is the DFTS, or Daylight Flower Tracking System. This feature run the calculations for all your simulated flowers to track the position of the in-game sun! (Note: Moving positional Sun-tagged light source must be coded into the program for this feature to work).

ACT NOW and be the first to own a PGA!
List price: $699 (ISA slot only, VLB to be announced)

I'd buy an AI accellerator over a fricking network card accellerator.

However an AI accellerator is only as good as the AI it is running. However if it is even 10 times faster, never mind 100x or 200x that means that AI algorithms can get more involved and interesting.

By providing an API to code to, and presumably a software version (multithreaded perhaps) they can get games written to that API. Then the hardware can accellerate it if it is present. Also a lot of the AI algorithms aren't simple things to program, so having them available in a verified third party library could be very tempting, and worth licensing for a game company.

However some things like Line Of Sight seem far more accelleratable on GPUs, basic terrain/distance could be done with a small z-buffer rendering from the AI's point of view, with highlights for special objects. The GPU has all the terrain and object data as well, which would help - it seems odd to send the scene to the GPU (for rendering), to the physics card, to the audio card (for 3D audio), and now to the AI card. It seems such a waste of resources.

if it cost $10 then u got one customer, but one more thing NO PCI .. it should be using PCI-E 1x ... u got the deal...

It would be interesting to see RTS, FPS, and RPG multi-player servers use this. That would pass along the benefits with no extra cost or hassle to the players.

If these accelerators were really going to take off, wouldnt one of the big four just buy them and incorporate it into already existing platforms? i.e. intel and amd have both discussed adding specialized cores to future "many-core" cpus. That would seem the ideal to me, a few large general purpose cores for general software, then if you need say AI acceleration you shut down(C1e or whatever) most of the power hungry general purpose cores and use the much more efficient specialized one(s). I think amd or intel were calling them "array processors" at one point.

Add all the BS phsyics and eye candy you want, it's no substitute for gameplay, and far too many game developers are forgettting that. Make the game addicting, fun, challenging, thought provoking FIRST! - Then you can work on making it look nice.

And as far as "programing for a second core is too hard" goes, why would it be that much more (if any) more difficult than programming for a third party device altogether? I'm ignorant of what is really being programmed here, but what "runs" the ageia cards, its a processor of sorts isnt it? What makes it so polar opposite of a cpu?

Why don't game companies just program the games to be multithreaded? A lot of gamers have the ability now, why program for a 3rd party proprietary units like a physics cards etc, we already have a spare cpu! Let us use it!

There are a finite number of PCI slots on motherboards now-a-days.

Slab this AI accelerator, the physics accelerator and one of those uber-NIC's into one and a same card, call it a game accelerator, and sell it for a humane price. Then I might take these things seriously..

I'm all for the off-loading of things to specialized processors (isn't that what the cell processor does?) but I think where these guys (Ageia, Bigfoot, and now these) are missing is the price point. Personally, I find it hard to justify spending $200 for a piece of hardware that only marginally affects gameplay, if noticeable at all. However, that said, I would pay 50 bux "just to see" or hell, "just to say I have one", but not $200. Especially if I cant be guarenteed that new games will be utilizing the hardware.

I wonder, are AI tasks so much different that phisics to justify extra chip? If I am not mistaken Ageia promises acceleration of path finding as well.

I certainly could see some uses for it but I guess we'll have to wait and see how they develop it.

Skylab and the matrix style AI is always for the better :p