AMD says PhysX will die if it remains a closed and proprietary standard

i thought havok fx was gpu accelerated:3eyes:

 

Havok FX got canned when Havok was acquired by Intel. It looks like AMD are pushing for it to be brought back for their graphics cards, but it depends on whether Intel are keeping it for Larrabee, or just limiting Havok to CPU acceleration only.

 

I just hope that we won't be forced to have physx with nvidia and havok with ati...
If so then developers would most likely only support one for physics effect and continue to use cpu for core physics elementss. No developers wants two support two different code paths to be able to use different brands of hardware. The first one to make a physics api and implement it on both all hardware will win.

 

No, the one which is most open will win. It's the same reason why 3dfx Glide died.
It was a revolution when it was released and later just lagged behind. The fact that it was 3dfx's propertiary API was not doing it any favor at all...
If HavokFX could be accelerated by S3, Intel, NVIDIA or anyone else that is building GPU's, it's a clear winner from the start.

 

haha BS, when PhysX was owned by Ageia and was an add-in card nobody said anything, now that nVidia owns it nothing really changes, simply ATi is jealous because there's potential to PhysX.

 

Donno on what planet your living on but on planet reality, we had a massive argument about how retarded the ageia solution was with the only defenders being retards who bought into it.

 

That just kind of lame to call a group of people retards because they belive in what they saw was a device which could change the way we game. Just because you guys or some folks has a diffent view on it then those who liked what it did or owned it.

Because I was for it and I didnt even own the card but wanted one, because I liked what it could do for us PC gamers. It might not be up to par. But I still liked the idea of better phsyic in the game. So Am I a automatic retard for not being neg about it. Even tho I do still fill they have a long way to go still to improved on it.

If it wasnt over price and PCI only at the time, I prolly would have had one my self but I never bought it because of it price. Then I knew some one was gonna buy them sooner or later to make it easier to get.

As for AMD, all they need to do is talk to nvidia about using it on there cards, if it can really be used on non nvidia hardware like the guy with the tool showed.

 

Just like they're buying into NVIDIA's bullcrap Mass physics in games are great, can't deny that. But a single sided solution, well thats where is the problem...
The problem with Ageia was that a) their card was way too expensive for what it could offer to the user and b) it actually caused massive performance penalties just because all the traffic had to be shared through external bus. Especially where physics and graphics always have to be synced to work properly. I've seen that for long time and i knew it can only be properly done if graphic and physics processing units are on the same card or in the same chip (the GPU).

 

Ehm no, this one of the issues which pissed me off greatly, when ageia came out with their cards It was pretty obvious it was going to be an epic fail and that the company would disappear in the belly of another.

The reasons are simple old hardware, old interface, price and a device which had poor support and nearly zero advantages. basically it was an overpriced deadwieght which would sit in your PCI-slot and do squat.

When the company was baught by invidia the above was confirmed, from becoming a deadweight consuming power to just a deadweight probably responsible for some form of global warming.

 

PhysX card for $200....mid-range vid card for less....where did the price tag on the PhysX PPU ever make any sense? Ageia claimed it would improve game play by adding hardware accelerated physics. Ok, nice idea.....will it ever go beyond simply being eye-candy? If not...it's a completely pointless topic. Most users aren't going to enable something that will be detrimental to their FPS unless they have the funds to buy top of the line graphics cards. Ageia was destined to fail for 2 reasons...they marketed a processor that was only capable of physics calculations and priced it like it was a damn upper mid-range graphics card. The bus used really made little difference. Most people weren't willing to pay the price for the limited functionality and massive performance hit.

As for NGO getting their PhysX port working and out in the public....I have no doubt they can get it working...my doubt is in it ending up in the hands of the public. I just don't see it getting mass distribution....nor do I see nVidia allowing mass distribution without some sort of licensing fee.

 

Here's how I see this playing out

-> Developers include software physX options and hardware options in their games just as they did when graphics cards were just becomming standard.

-> Hardware accelerated physics becomes a standard just as video graphics became standard. DirectX/Cuda allows most of PhysX's features to be accelerated via graphics cards.

-> For ATI the point becomes moot as they begin shipping most of their Mobo's with integrated graphics that can be converted to physics processors should the user upgrade to a discrete solution.


I think this will conclude in about 5-8 years with most motherboards having integrated graphics that double as physics/cuda/gpgpu's should the end user opt to add a discrete graphics solution.

/predicting the future since 2008
//where's my damn flying car!
///jetpacks would be cooler

 

Did I disagree with you? My point was that it is certainly not cross platform, as my other post indicated. ATi users are left out.

 

Ati users are only left out because the physics cheque went to havok.

Nothing wrong with Physx or havok and for the most part it won't affect users much.

Theres much more important issues to address like pushing for 2Gb+ ram as the new standard for graphics cards, adoption of a new high-density format for storage and solving through-put and latency issues. Basically, all the things needed to raise overall pc performance across the board and lay the foundation for the next generation of games software.

 

i wouldnt really consider Nvidia to have any huge advantage since its users would be on CPU accelration in any Havok game. those with decent quad cores would be just fine on either account.

yet it never ends, you cross one hardware border and achieve better performance then a software comes out to make it crawl to unplayable frame rates and users complain. its always been like that. those issues will continue to be addressed just as fast as more problems arise from it. over and over again until were all in the Matrix.

 

I wanna learn kung fu!!!

 

then watch some jackie chan movies or something

 

With the upcoming DX11 providing support for physics through properly used GPGPU calls, nVidia's proprietary PhysX support is the equivalent of 3dfx's proprietary support of Glide 2/3 graphics language. It's great while it lasts provided you have the required hardware and the game supports it, but as soon as a cross-platform alternative comes out, even if it may not be so efficient on the hardware which supported the old dedicated solution (like Glide, or PhysX), it will replace the proprietary language.

As soon as DX11 and OpenCL are viable alternatives (rather like the days when Direct3D 5/6 and OpenGL became viable alternatives to incompatible graphics languages from each card manufacturer), support for the likes of hardware PhysX will be doomed. Dedicated hardware will be supported not by a specific driver for a particular manufacturer's product, but instead through the DX11 or OpenCL interface which that hardware manufacturer has to support. It happened about 11-12 years ago with graphics, and will no doubt happen again with physics, and is the only sensible way to go forwards in the long run.

 

DX11 and OpenCL are just platforms that can possibly provide a Physics solution. They won't replace PhysX but something on the platform COULD replace PhysX. More then likely PhysX and Havok will just be ported over to the platforms and used that way. A more accurate thing to say would be DX11/OpenCL will replace CUDA.

 

CUDA will run over the OpenCL layer, therefore so will PhysX. Last week Nvidia announced full support for OpenCL.