A clever chap has managed to enable the PhysX layer on HD 3800 series ATI Radeon cards, and he wants to add more to the list.
Details in the news article
Details in the news article
PhysX on NVIDIA only? - Not anymore!
- Thread starter PV5150
- Start date
Rastalovich
New member
I`m real confuzed. How does his score improve with hardware physx possibilies ?
And.. I do wish they`d sort out the physx drivers to hijack all those games that use the software-physx.
And.. I do wish they`d sort out the physx drivers to hijack all those games that use the software-physx.
I know its a bit off topic..but its cuda related...I just noticed that elcomsoft password hacking CUDA software is listed within the CUDA developments page...clicked on more info....went to elcomsofts page...thought lets have a laugh and see how much it costs...then nearly SH*T myself at cost...
check it out
http://elcomsoft.com/eprb.html
still no sign of the physx cuda beta though ... still says "later this year"
check it out
http://elcomsoft.com/eprb.html
still no sign of the physx cuda beta though ... still says "later this year"
there using the havok engine instead of ageia physx...from what a freind of mine at infrogrammes has said the SDK's they have are pretty similar...he said its no surprise really someone has the physx running on ati...at the end of the day all gpu's are bound by dx and opengl (although dx is the most prominent) so the way they work has got to be pretty similar as they are bound by that standard...I would of thought though that Nvidia would have put some kind of restriction within the hardware....even if it was just a vendor ID tag to stop this happening...even if they have I think its evident whatever hardware protection they have is easily bypassed...
Rastalovich
New member
name='PP Mguire' said:
I know there`s been Ati talk in the past regarding physics stuff, and alt to physx.
But inherently, I can see and understan that all a gfx card manuf needs to do it hijack the physx drivers - legitimately they`d need permission, I`m thinking.
But still, they *could* just attach code to an existing physx driver install and branch off it.
Can`t see it tho in this day`n`age. Every1 wants and needs to be 2000% legit or face the lawyers.
name='Rastalovich' said:
yep...to right...all ati will do is reverse engineer it and rebrand it ... saying that ati and intel have teamed up to face the big green machine
http://www.theinquirer.net/gb/inquirer/news/2008/06/12/amd-teams-havok-physics
there has been other news about it but i cant frankly be bothered digging them all out...its interesting as larabee is beng co produced at the same time...weird intel and amd teaming up tho
Maybe im not getting my point clearly across. Wouldnt you need to have the same type of tech inside your ATI GPU as Nvidia to run Nvidias PhsyX program? Im talking hardware here not software. Its simple to reverse engineer software but hardware? Yeaa im thinking ATI might be slipping a fast one somewhere.
To try and make it more simplified wouldnt it be like trying to run 3dfx Glide on modern day GPUs?
To try and make it more simplified wouldnt it be like trying to run 3dfx Glide on modern day GPUs?
i get what you mean....
lets put it this way...every gpu regardless of manufacturer is bound by the aspects of direct x...the abstract hardware layer has to work with a predefined instruction set...that predefined set is a standard.....cuda is an open sourse software development kit that allows programmers to unlock the potential of the parallel processing power of gpu...these gpu's are designed around the standard (direct x)...so inherently regardless of manufacturer there all designed around the same standard and CUDA exploits that..hence its no surprise ati cards can run any CUDA developed software.....does that make sense....ive had a few beers.....its the whole reason why gpu's have left cpu's for dead in development terms...really really sorry if ive not explained it very well...if ive not i'll edit this post when ive sobered up..or alternatively log on to www.stickam.com and join oc3d gamers password is ineedsleep but im getting more drunk as time goes on
lets put it this way...every gpu regardless of manufacturer is bound by the aspects of direct x...the abstract hardware layer has to work with a predefined instruction set...that predefined set is a standard.....cuda is an open sourse software development kit that allows programmers to unlock the potential of the parallel processing power of gpu...these gpu's are designed around the standard (direct x)...so inherently regardless of manufacturer there all designed around the same standard and CUDA exploits that..hence its no surprise ati cards can run any CUDA developed software.....does that make sense....ive had a few beers.....its the whole reason why gpu's have left cpu's for dead in development terms...really really sorry if ive not explained it very well...if ive not i'll edit this post when ive sobered up..or alternatively log on to www.stickam.com and join oc3d gamers password is ineedsleep but im getting more drunk as time goes on
its really interesting stuff....its quite amazing how many pieces of software have already been developed....most of them are molecular physics stuff...and even though I can grasp some stuff easily just the titles of some of these things baffle the crap out of me....take this fetching title for example "Interactive Visualization of Volumetric White Matter Connectivity"... and thats the shortened version ....
....ATi stuff currently uses the Havok system that was bought out.
nVidia stuff use CUDA (renamed PhysX).
ATi 3 series was always meant to have physics help in it and anyone reading the website when first announced would no this.
The good bit is that nVidia has said that once they release the first public CUDA software that they will release a complete open source SDK stuff for all developers to enable their kit (although one step behind nVidia's funkiness)
nVidia stuff use CUDA (renamed PhysX).
ATi 3 series was always meant to have physics help in it and anyone reading the website when first announced would no this.
The good bit is that nVidia has said that once they release the first public CUDA software that they will release a complete open source SDK stuff for all developers to enable their kit (although one step behind nVidia's funkiness)
name='rrjwilson' said:
err i think u need to do a bit more research...CUDA is the open source SDK for nvidia unlocking the power of GPU's in a variety of ways not just physics for games.although the physx development is being done under the cuda sdk..check the nvidia website under technologies....physx and havok are pretty much the same thing from the 2 companies both dealing with parallel processing of the gpu being used to help with ingame physics...CUDA has already 30 pieces of software available through the nvidia site...mostlymolecular physics...a few helpful pieces to do with video encoding and password hacking ...
for more reading check
Cuda Developments
http://www.nvidia.com/object/cuda_home.html#
Physx Development
http://www.nvidia.com/object/nvidia_physx.html
Err no. If you will notice I said a complete open source SDK not just an open source SDK. The SDK currently available is not the whole hog. This is exposed in an interview with the CUDA development leader in a CustomPC article one month ago. He states that once CUDA enabled drivers are up and running for nVidia then a compete SDK with appropriate support will be released to make full use of the CUDA hardware. He also states that the reason for this approach is to make sure the problems are experienced by nVidia first and so they have the better knowledge for application use.
Technically they are different but as the nVidia developer was asked "what have you done with PhysX since buying it?" and he responded by talking about CUDA enabled hardware having PhysX stuff, I believe if he can say it so can we. PhysX is part of all 8 series and above hardware and CUDA is the language to set it free, kind of like Ali Baba saying "opensesame" to the cliff in his story. It isn't truely PhysX it is the next step up for PhysX by unleashing the power of the GPU.
Another thing he says in the interview is that CUDA enabled games will be able to run both graphics and physics processing on one card as well as two, three or four. This slightly worries me as an SLi user. Currently my system pretty much pimp slaps every game out there (including Crysis but i've been told later levels will test it better). Now with the physics side being moved from CPU to GPU the CPU can handle the actual data part more quickly reducing in game lag. But if graphics and physics are on one card even if parallel the performance gain from shifting the physics will be null and void as there isn;t enough hardware to go around (unless the game is lacking on the graphics front). Over an SLi computer the CPU can handle data transfer using all cores, GPU1 can handle the physics behind the data and GPU2 can render it all in glorious high resolution. Surely this would refine the gaming experience whereas a single GPU setup you would have to find the best way to run for that game.
Just seems like more work than is really needed on their part. Hopefully it will be worth the wait though.
Another thing he says in the interview is that CUDA enabled games will be able to run both graphics and physics processing on one card as well as two, three or four. This slightly worries me as an SLi user. Currently my system pretty much pimp slaps every game out there (including Crysis but i've been told later levels will test it better). Now with the physics side being moved from CPU to GPU the CPU can handle the actual data part more quickly reducing in game lag. But if graphics and physics are on one card even if parallel the performance gain from shifting the physics will be null and void as there isn;t enough hardware to go around (unless the game is lacking on the graphics front). Over an SLi computer the CPU can handle data transfer using all cores, GPU1 can handle the physics behind the data and GPU2 can render it all in glorious high resolution. Surely this would refine the gaming experience whereas a single GPU setup you would have to find the best way to run for that game.
Just seems like more work than is really needed on their part. Hopefully it will be worth the wait though.
