AGEIA's PhysX Processor

[Zarkon]

If you watch the front page, you've noticed me posting an ass-load of PRs from AGEIA about their PhysX physics processor.  Why?

Here's your reason:

http://physx.ageia.com/footage.html

Just check out the difference in GRAW with and without it.

[Daehawk]

I knew physics cards were coming for a few years now. Saw a big info video on them a few years back. I was kinda hoping it would'nt happen. Its just more money i dont have to spend. But they do look to do a kickass job.

[Eduardo X]

So, while that is a cool demonstration, I certainly hope a grenade wouldn't do that much damage in GRAW. It just isn't that powerful.

[Jumangi]

Unless they release diffrent versions that are affordable its going to be a niche product regardless of how cool the stuff looks. PC Gamers are already geting tired of dumping money for faster CPU's and video cards.

[biggercup]

I am tired of spending $$$ buying a new videocard just to have 1% faster frame rate.
I am sick of it and I will not buy the physics cards no matter how cool it might look. :evil:

[Dafones]

I look forward to PC motherboard setups that have dedicated graphics, physics, AI, sound, etc. processors *cough console cough cough* that you buy as one unit, instead of having to buy the cards separately. You update the chips as needed, but you're rocking out one solid, unified, processing matrix. Booya.

[kathode]

Video seems a little misleading to me.  Sure there's a lot of debris blown around in the second one, but the main visual impact seems to come from the much larger smoke cloud and fiery shrapnel.  There's no reason why either of those effects couldn't have been put into the first video.  Even if they just canned the trajectories of the shrapnel in an animation the effect would be very similar.  No question that the second one is better, but there's some PR magic being worked there.

[Atragon]

If I'm not mistaken, in the first one, the truck remains intact, in the second one it doesn't. Specifically, the roof of the truck.

[greeneggsnham]

Video seems a little misleading to me.  Sure there's a lot of debris blown around in the second one, but the main visual impact seems to come from the much larger smoke cloud and fiery shrapnel.  There's no reason why either of those effects couldn't have been put into the first video.  Even if they just canned the trajectories of the shrapnel in an animation the effect would be very similar.  No question that the second one is better, but there's some PR magic being worked there.

Exactly what I thought when I watched it. I'd like to see what this thing can do, but that example just seems to be nothing but PR Spin.

[Zarkon]

The truck all but disintegrates with the PhysX processor added.  Without it, it blows up, but remains pretty much intact.  With it?  You can pretty much see a TIRE fly off, hit the wall, and rebound past you.

[Toe]

The truck all but disintegrates with the PhysX processor added.  Without it, it blows up, but remains pretty much intact.  With it?  You can pretty much see a TIRE fly off, hit the wall, and rebound past you.

I think the point is that you can have the truck blow up and throw debrie around without a physics engine if the developers wanted it to. Yeah, it might be a "canned" explosion that is the same no matter what truck you happen to blow up, but still, a non-physic-engine explosion can (and do in many games) look 10 times better than the the example shown.

[Crawley]

That short demo video had me interested to see other examples but no where near sold on this. Have they spoken to a price point on it? For something that just does physicis it better be in the $50-$60 range at the most.

[Jumangi]

There also going to be having competition from ATI and Nividia who are showing off physics acceleration with their cards. Wasn't sure if this was some sort of addon to the cards or using the current hardware to also help with the calculations but if they can pull off something descent that gamers can just get with their video card purchase I'd think this would hurt and widespread adoption of Ageis's card.

[Zarkon]

Note:  Nvidia's solution requires either the 7600 or 7900 series card, preferably the 7900.  So $500 compared to $250 for the PhysX one.

[Graham]

It's going to be integrated into future video cards, so unless you are DYING to have the physics accelleration, I wouldn't get an add-in board, not for $250.

[shaggydoug]

Note:  Nvidia's solution requires either the 7600 or 7900 series card, preferably the 7900.  So $500 compared to $250 for the PhysX one.

Please explain your requirements or why they are different from this PR:

http://www.nvidia.com/object/IO_30478.html

This says 6 series cards as well.   Are you talking about something better than what they have described here?

- shaggy

[Zarkon]

http://enthusiast.hardocp.com/article.html?art=MTAwNSwyLCxoZW50aHVzaWFzdA==

This quotes 7 series, with recommendation of 7900.  *shrug*

[Atragon]

And note how they say that with 1 card there may be some slowdown, they recommend SLI where the second card does only physics... oh great,  single card performance for dual card prices. Give my the PhysX so I can still have SLI if I want it.

[Hetz]

Oblivion has amazing physics and it runs well enough.  I don't see why it can't be done on todays cards, unless they passed the money hat around to UbiSoft...

[unbreakable]

For my money, I would probably be happier with an upgrade to the in-game physics than the graphical prettiness.

I hope they roll this into graphics cards, and you can turn down the gfx to improve the physics.

[Turtle]

Oblivion has amazing physics and it runs well enough.  I don't see why it can't be done on todays cards, unless they passed the money hat around to UbiSoft...

Most current games have physics, but the key here is that those physics are only working on a few things at the same time, and are limited in how they can react and collide with each other.  

The videos they show have hundreds of not thousands of physics objects that can all be affected by physics at the same time.

You know how in some games with physics objects will get stuck inside each other?  This is because collisions in video games or the digital world in general are very difficult to handle correctly.  It takes quite a bit of power to handle collisions well, a physics processor can help a great deal with that.

[Jumangi]

Another turn off for me I read at Firingsquad. This was from a demo of a FPS prototype game called Cell.

One thing we did learn is that the demo we saw was LAN-only; because of the limitations of Internet bandwidth with all of the physics Cell Factor would not work as an online game.

I pay $300 for the card but won't be able to use it on any online games? Yea guys let me know when you have something worth my time and money.

[Turtle]

Wow, what'd they do, kick your dog?

It's perfectly reasonable that you can't use this kind of physics interaction online simply because no internet connection besides giant corporate T1s could ever handle the amount of data being transferred for these physics interactions.

It's not their fault the internet only has so much bandwidth at a cost.

That said, I'm not sure I'd buy a physics only card for $200-$300 right now, but it'd be a big factor in my next big PC purchase.

[Zarkon]

And Dell -does- have a PC with the card for like $1300.

[Jumangi]

Wow, what'd they do, kick your dog?

No idea were this is coming from, I mean heaven forbid someone should have something besides a glowing reaction with the latest and greatest technology to come out....:roll:

It's not their fault the internet only has so much bandwidth at a cost.

Doesn't matter to me as a consumer. I care about what I would get for the money I payed or might pay for a product. For me this is a turnoff.

[stiffler]

I'd say a lack of internet play is a big turnoff for these cards.

I'd still be interested in seeing it integrated into my video card.  It would be a nice little bonus for when I play the SP campaigns.  The limitations would keep me from buying a standalone card regardless of price.

[Crawley]

I thought one of the first games to support this thing was City of Villains. Is that lack of use on the Internet 100% positive?

[Zarkon]

Somehow, I think the interpretation of what was actually said got missed.

And yes, City of Villians supports it.

As does Vanguard: Saga of Heroes.

Personally?  I don't see why internet play -can't- be involved.  The physics doesn't effect game play substantially (as far as how the game actually plays).  The game is no different if it's on or off.  So on internet play, just turn it on for someone with it, turn it off for those without.  It should be almost like a graphics option, IMO

[Jumangi]

The server would still have to send the data that lets every computer know what is happening to the hundreds or even thousands of objects that are being tracked when the truck blows up. That would quickly overwhelm most current cable/DSL speeds.

[Turtle]

Physics objects on internet game servers works in one of two ways.

The first way is that the physics object is entirely client side.  This is usually the case with ragdoll corpses, such as those in Counter-Strike.  When you see a ragdoll online, it won't be in the same spot as ragdolls on other players' computers.  The same can be said for other non-important objects that are there just for show.  Players can still interact with these objects, but nothing they do to it will show up on other players' screens.  So, if you ever wondered why a CS player will sometimes shoot the ground near a corpse in an odd pattern, he's shooting at the corpse on his screen.

The other method is that the physics object is server tracked entity, just like the players.  For example, in HL2 deathmatch, there are large objects on the map that can be picked up using the gravity gun and launched at other players.  In CS, some maps like Office have such objects, like the filing cabinets that people often use to block entrances.

That gameplay trailer with people blowing up all those boxes and pipes on the Agea site is with people playing on a lan.  On a lan the data transfer rate is at least 10mb/s if not 100mb/s.  So lan games can handle a lot of information being passed between computers, but we couldn't see that kind of stuff for online games unless T1 connections became standard.

[Zarkon]

You know, if AGEIA gets widespread enough, I can see them going and saying, "SEe, this is why we need faster network speed on the net in the US."

[Atragon]

It occurs to me that if all the clients have a PhysX card then you could reasonably expect the clients to be able to track most of the interactions... (yes, that would leave things wide open for cheaters with most server-side cheating detection schemes in effect) So if you tell ALL the clients that grenade exploded at point A, every client should be able to come up with the same result, cutting down on the required bandwidth.

Of course, with all the asshats out there, I don't really think this would be viable really as it leaves many holes for cheating.

[Turtle]

Actually Atragon, you can't.  Game netcode is a mishmash of assumptions and predictions of player movement and what he can see currently.  This is then matched up to the actual reality of what the player has done once the server gets the real info and sends it to clients.

Sure, it's certainly possible to tell a client that box A got bounced in so and so direction from a stand still position.  But what you arne't factoring in is that now that the box has moved, clients much update its position every frame from the server, just like another player.

Not only that, what if another rocket detonates nearby, or a bullet impacts the box in mid air?  What if it lands on a player, or maybe even a dynamic wind pushes it a little?

Well, yes, they are currently doing this in games where you have server side physics objects like HL2 Deathmatch.  However, you'll notice that these objects aren't too common since each of these objects is essentially another player's worth of data to transmit.