6 Months on and... ?

[MrNasty]

Well it's been a while. We've all waited for stuff to come out. And waited. And waited.

So lets review whats happened in Hardware Physics:

1. Ageia brought out the PPU about 6 months late

2. Ageia Proposes to bring out a PCIe version of PPU "Within months" (Eng. samples had dual interface)

3. People discovered mostly there's nothing that uses it completely or in some cases at all

4. Havok announces Hardware Physics on graphics cards (HFX)

5. Ageia proposes PCIe PPU's "By SDK 2.5 - along with x64 support"

6. nVid shows some funky videos and live demos

7. ATi says it's working on its own API

8. nVid and ATi beat each other for 2 straight months about whose physics solution is better
8.a nVid says one card will be able to do physics and graphics
8.b ATi says an x1300 will outperform Ageia's PPU
8.c ATi says an x1600 will outperform the 7 series
8.d Both then say "why not have 3 graphics cards?"

9. ATi announces it will be using HFX After All.

10. And it will do waaaay better than a G70

11. Ageia releases around 27 patches - none of which actually do anything except enable the card for future releases based on each version of the SDK

12. nVid finally drops the idea of doing physics on a 7 series GPU, only discussing physics in regards to the G80 currently

13. Ageia makes PhysX free - fastest project released: Linux port. No games, then.

14. Ageia's "to be released" page looks exactly the same - PCIe slots everywhere are still languishing

15. Havok's "to be released" page looks exactly the same

16. Still a major cause of forum dissention everywhere


That's an impressive amount of trouser waving and one-upmanship with no actual progress in 6 months!

edit: lol Dan, had forgotten about that!

edit 2: lol Quadnad - you reminded me of another, not in a bad way though. no worries

 

[Dan_D]

Don't forget about PCIe x1 PhysX cards that have yet to be released.

 

[kirbyrj]

*yawn*...Physics cards? I thought they went the way of Vesa Local Bus by now .

 

[quadnad]

I suppose both companies are just waiting until the first few rounds of games arrive that support their respective technologies.

I almost feel bad for those who pissed away the money for the PPU when they launched. almost.

 

[SuperGee]

Well game take a long time to make.
Lucky some where already for a long time in produktion like UT2007 a 4 years project or even more.
While CF:R is a short one wich have a lot influence on my expectation. It whill not compete with UT2007 with a large margin.
But it will be the first full scale gameplay physics game out.
And with that short time it will not have to much Q&A out of it so I expect it needs some Patches to.
It will be a game wich depend on it deliverd gameplay.

Well begin next year it will be out I hope.

 

[HOCP4ME]

Lol, that's completely correct. I wish something would actually happen in the physics market.

It will eventually, of course, unless CPU manufacturers get to 16-core processor first. And they may very well do that if AGEIA, NVIDIA, ATI, and game developers don't get off their butts and start giving us what was promised!

Vesa local bus, I have a card that fits that. If anyone complains about video card sizes increasing, I'll just show them that.

 

[DragonNOA1]

I remember hearing about new PPU's that could do more things than the current ones. Of course this info came out before they were even released. They were scheduled to be released when the original product actually shipped but I haven't seen any updated products or suppurt for more features. Did I lose ya? But in the end who wants more features for a product whose original features aren't even being used?

 

[SuperGee]

Well there are a lot of games in the making. So if the SDK supports more features.
The dev's have more cool options to implement in there game. It up to them to make something out of there game to stand out in physics.
Or do existing features more efficient and better.
A good thing for the second wave of games.

 

[Animal]

Well there are a lot of games in the making. So if the SDK supports more features.
The dev's have more cool options to implement in there game. It up to them to make something out of there game to stand out in physics.
Or do existing features more efficient and better.
A good thing for the second wave of games.

So when are we gonna see the first wave?

 

[SuperGee]

So when are we gonna see the first wave?

It started with Graw.

Depends what your definition of first wave is.
Take your pick
1) All PhysX games wich come out before Havok FX games come out.
2) All the games already in produktion wenn PPU came out.

For CF: revolution
In cas off def 1 ) it could be firstwave, if there isn't annything of havokFX out.
It could be Second wave if there is havok Fx game out.
On def. 2 ) it would be a second wave game with a very short produktion time allocated.
Started after PPU release

UT2007 is a first wave because it was already far in produktion. 1)
If def 2 used it could be second wave. They take the time to make a good game.
If released somewhere mid 2007 it possible there are HavokFX games out.

Knowing the long supportive list of ageia PPU supporting Developers. there are a lot Games in produktion before PPU came out. So in def. 2) That are a lot of games. But most games take or need years to be able to make something good. So the first wave is still on the go in its early fase. It could take 1 or 2 years for the whole wave to come out.
A game could be more or less then a year in produktion and be in produktion for yet another year or more.

This means a 2 year project already for a year in produktion will come out in may /june 2007. while UT2007 is guestimate 3years or more in poduktion and going to be a small year or a half from now on. Thats still a year or more after PPU release.

For havokFX it wouldn't be so much different. Some gamesdevs will adopt HavokFX mid Produktion to. Or come with Patches. These game wouldn't probaly stand out in HavokFX use. But will be the first HavokFX games but using effects physics only.

The better games would be games that incorporate HavokFX or PhysX features in gamedesign stage where this special features are well balanced and merge in the complete design to get a well ballance an intergrated game and be able to take the best out this enhanced physics. This take also time to be able to do so thus 2 to even better 3 years.

 

[griff30]

By the time the second wave games come out the soundcard will do physics way faster than the first PPU. LOL
Glad I didn't pay for research and development of the first PPU buy buying the card when it first hit the shelves.

 

[clayton006]

...Vesa Local Bus...

I think we are on to something here! Let's bring those old mobos off and use the VESA bus for our physics cards.

I just hope for the industry's sake that we get some solution to have faster physics, and offload that CPU.... maybe multiple cores may help this solution ultimately?

 

[Dan_D]

I think we are on to something here! Let's bring those old mobos off and use the VESA bus for our physics cards.

I just hope for the industry's sake that we get some solution to have faster physics, and offload that CPU.... maybe multiple cores may help this solution ultimately?

That's what they are doing with Alan Wake, but internet speculation is that the CPU cores aren't as fast as a PhysX processor. There is still a debate as to whether or not the GPU PhysX processing will be superior to Ageia's PhysX processor cards or not.

 

[edgeprod]

Although I don't regret buying the hardware, I wish it would be supported more broadly for sure. I'd even settle for a timeline I could trust.

 

[clayton006]

That's what they are doing with Alan Wake, but internet speculation is that the CPU cores aren't as fast as a PhysX processor. There is still a debate as to whether or not the GPU PhysX processing will be superior to Ageia's PhysX processor cards or not.

Do you think this will ultimately come down to industry support (the chosen standard)? Or I guess I should ask the question, what developers will find easier to implement?

 

[Manny Calavera]

Or I guess I should ask the question, what developers will find easier to implement?

Whatever is easiest of course,as most progammers are lazy...CPU/CUDA/DirectX 10 will be the firstmost supported in that order.With CUDA potentially taking the forefront,if Nvidia backs it properly.

 

[mucker]

Forget Physics, what about X-RAM

 

[quadnad]

Forget Physics, what about X-RAM

lol, at least the people who paid for the high end x-fi cards with xram can still use it for something.

 

[HOCP4ME]

XRAM is better than physics at this point because it provides a tiny improvement in most games. Physics hardware only provides a tiny improvement in ~six games.

Anyone notice a trend in the last five years? Let's see, we've had 64-bit processing, PCI-E x1, XRAM...

Is physics going to be added to that list? I certainly hope not. Physics hardware has the potential to revolutionize gaming, and I want to see that happen! If only the developers were willing to take a risk and release a fully physics-enabled game.

I know there are several of those in development. Hopefully they will be as good as promised. But we need to start seeing some Havok FX games as well.

 

[mucker]

Full utilization of a multi-core cpu for gaming would be nice too. I know it's getting better but the pace is slooooooooooooooooow

 

[Dan_D]

Do you think this will ultimately come down to industry support (the chosen standard)? Or I guess I should ask the question, what developers will find easier to implement?

On the first part: Without question. Ultimately, whatever software developers and Microsoft decide to support is what will win the physics processing war.

On the second part: I have no idea which is easier to code for. I suspect multicore CPU and probably GPU. According to the Havok FX page on the Havok website, this is in fact the case.

I also suspect that multi-core usage for physics calculations will probably be the slowest method in terms of overal actual performance.

 

[clayton006]

On the first part: Without question. Ultimately, whatever software developers and Microsoft decide to support is what will win the physics processing war.

On the second part: I have no idea which is easier to code for. I suspect multicore CPU and probably GPU. According to the Havok FX page on the Havok website, this is in fact the case.

I also suspect that multi-core usage for physics calculations will probably be the slowest method in terms of overal actual performance.

Thanks for clarifying Dan. I wanted to make sure I was thinking along the right lines here.

 

[SuperGee]

On the first part: Without question. Ultimately, whatever software developers and Microsoft decide to support is what will win the physics processing war.

On the second part: I have no idea which is easier to code for. I suspect multicore CPU and probably GPU. According to the Havok FX page on the Havok website, this is in fact the case.

I also suspect that multi-core usage for physics calculations will probably be the slowest method in terms of overal actual performance.

I think PhysX and Havok+FX are API's.
Both supporting CPU's Single core, dual , quad. Also all nextgen consoles.
They are Physics API middleware. They offer highlevel Physics features ready to use or adopt to a Game projects on a higher level. They take the need for reinventing the physics wheels over and over.

The differens is that next to the common Platforms (CPU/Consoles) PhysX supports next to the others PPU and not GPU.
While Havok FX supports not PPU but GPU instead.
Both are CPU API to.

For the Game developer they are all the same the difference is what gamer need to play the game in optimal form.

A PhysX game the player could push the Physics setting high wenn PPU is detected, if not it must set on medium or lower.
While Havok For a PPU it does not support so have nor a setting for it, it depens on what for GPU resourses you have and want to spend to physics to keep enough for rendering.
So if you got much GPU power you could put it on high. Or else medium. Because there is a wide variation is glider option would be the best way to set physics in a game.

Cuda is low level Shader and Physics coding. First uses as a C shader language.
C is not so easy as C++ but beats assembler in ease.
Plus with Cuda don't have the extended feature libary as full features Physics middleware packages have like Havok FX, newton, physX, tommahawk.
It's not middleware but more Hardware related and specific for nV shaders only.
While DX10 have HSML for it. But not aimed at Physics.

Well easiest to code for I think HavokFX and PhysX are more the same. PPU and +FX is more a intergral part. What's more is a matter off initialisation of the Physics API.

While PPU suport for PhysX is Free.
For Havok FX extension there must be paid

For consoles Both API arn't Free. And PhysX is also use in that market.
RedSteel for Wii is a PhysX game.

Microsoft is the firm off standards supporting al Hardware.
Ageia must deliver a PPU Directx physX driver.
While Havok+FX is completly bypassed as it's core busness is Physics midleware or API

While Ageia get a hardtime in the near future with competition from GPU's and HavokFX combo. HavokFX gets a hard time on the long run wenn DirectX coming to play in the Physics market.

While Ageia is also doing good in the console market,havokFx probaly to.
Maybe ageia and Havok are saved by the larger console market in the long run.

 

[HOCP4ME]

Do you think this will ultimately come down to industry support (the chosen standard)?

Of course it will. A product can't survive if no one will use it. Of course, right now, none of the physics solutions are being used. AGEIA has a small lead, but I don't think that wil matter much when things really get going.

I just hope Microsoft (or some other company - it doesn't really matter) will hurry up and make a standard physics API. Because if they don't, we're going to need a PhysX, 8800GTX, and a quad-core CPU just to be able to play all physics-enabled games.

As for multicore usage in games, I think raytracing is better than physics for that purpose. It's already been shown that, for multicore physics to be worth it, you need quite a few cores.