yeah... pwnt... 8800 series have no problem doing your physx card's jobNeXuS4909 wrote:
So my physics card means nothing?
I would not be so sure of that. They got some good stuff in the works.elbekko wrote:
Sweet.
Byebye ATI now
Um....so what u sayin'??!?1 you saying they gonna come out with geForce cards for teh calculat0rz?rdx-fx wrote:
Just a formality, so there weren't any troublesome patent issues for Nvidia, in my opinion.
Graphics cards excel as specialized mathematical processors (matrix mathematics and matrix manipulation being foremost on the list).
Graphics card = specialized math coprocessor.
Nvidia adding a few more transistors to do a few physics equations isn't any sort of a stretch for them.
Put another way:
My TI-89 calculator, using the same processor as an ancient Apple ][ computer (motorola 68hc11) + a small math coprocessor can do differential calculus equations. An Nvidia GPU has a few orders of magnitude more transistors. [TI-89 CPU = 70,000 transistors. Nvidia 8800 GPU = 681 million transistors]. Adding some transistors to do a little physics math? Not even a stretch for Nvidia.
Yet a 3rd perspective:
You can already do SETI@Home number crunching on your Graphics card.
If you can do that math already, basic Newtonian collision physics isn't a stretch.
Another way to put it is that NVIDIA would rather comply with existing programs that already make use of PhysX than call developers and tell them to rewrite code to comply with some new type of hardware.rdx-fx wrote:
No, saying that as humble 70,000 transistor CPU can handle differential calculus - a 681 million transistor monster like the Nvidia 8800 (or, it's descendants) should be able to incorporate some basic physics algebra. Nvidia buying out Ageia likely had more to do with making sure they were covered regarding patents, rather than Ageia actually having any technology that Nvidia couldn't figure out on a lunch break.trippy982 wrote:
Um....so what u sayin'??!?1 you saying they gonna come out with geForce cards for teh calculat0rz?rdx-fx wrote:
Just a formality, so there weren't any troublesome patent issues for Nvidia, in my opinion.
Graphics cards excel as specialized mathematical processors (matrix mathematics and matrix manipulation being foremost on the list).
Graphics card = specialized math coprocessor.
Nvidia adding a few more transistors to do a few physics equations isn't any sort of a stretch for them.
Put another way:
My TI-89 calculator, using the same processor as an ancient Apple ][ computer (motorola 68hc11) + a small math coprocessor can do differential calculus equations. An Nvidia GPU has a few orders of magnitude more transistors. [TI-89 CPU = 70,000 transistors. Nvidia 8800 GPU = 681 million transistors]. Adding some transistors to do a little physics math? Not even a stretch for Nvidia.
Yet a 3rd perspective:
You can already do SETI@Home number crunching on your Graphics card.
If you can do that math already, basic Newtonian collision physics isn't a stretch.
that will never happen tho. Why add a physics chip in gfx card when the GPU can already do its job..... but they might optimize the GPU more for physics calculation...Poseidon wrote:
An nvidia card + Physics card all in one = hax.
Last edited by GC_PaNzerFIN (2008-02-05 01:20:22)
lmfaotrippy982 wrote:
Um....so what u sayin'??!?1 you saying they gonna come out with geForce cards for teh calculat0rz?
I agree. nVIDIA should really do this now, put both chips on one card! That would be worth buying (me thinks).Poseidon wrote:
An nvidia card + Physics card all in one = hax.
Why? That could not be a serious question nor could it be a serious rhetorical comment.GC_PaNzerFIN wrote:
that will never happen tho. Why add a physics chip in gfx card when the GPU can already do its job..... but they might optimize the GPU more for physics calculation...Poseidon wrote:
An nvidia card + Physics card all in one = hax.
Last edited by topal63 (2008-02-05 09:58:27)
that's exactly why nvidia bought AGEIA. They want physx support for their gfx cards. Physics calculation with GPU is at it's early stages tho. The true next generation gfx cards with multiple cores can do it much better. There even might be a seperate core in the GPU that does the physics calculation. There aren't many games supporting the AGEIA or gfx card physics yet. But it is more than likely that companies like nvidia will eventually get more and more support for that. Atm it is pretty tough to add a seperate physics chip on gfx and that would cost quite a lot too. It is more likely that they start selling geforce cards that are optimized for physics only.topal63 wrote:
Why? That could not be a serious question nor could it be a serious rhetorical comment.GC_PaNzerFIN wrote:
that will never happen tho. Why add a physics chip in gfx card when the GPU can already do its job..... but they might optimize the GPU more for physics calculation...Poseidon wrote:
An nvidia card + Physics card all in one = hax.
Here is why - pathway (parallel processing). Multi-threading in applications, as a norm, is really not that far off. A few years from now it will be necessary so that game creators can keep up with the competition in terms of performance.
Having a separate CISC* pathway means the GPU, which is a tessellation engine/rendering engine/pixel shading engine (FPU**), wont have to do it. The physics-simulation will have a separate pathway and run concurrently. I have a physics card and can tell you that it doesn't matter if the GPU has the potential to do physics calculations/simulations. It works better (there is better performance) when a separate pathway is doing the physics calcs. The problem with physics simulation on a card - is the lack of games supporting it and not the implementation/design of the physics-card.
If games could make better use of threaded code (take for example: AI) on a Skulltrail platform you would get incredible environment-behavior combined with incredible performance. If/when they actually use (finally apply as a norm) multi-core processing (threading, multi-pathway, parallelism, etc).
*CISC: Complex Instruction Set Computer-chip.
**FPU: Floating Point computing Unit .
Last edited by topal63 (2008-02-05 12:33:03)
so you are saying they don't make a PPU of their own which was what I ment?topal63 wrote:
I doubt it... you're confused.
First-off Physx is primarily a software physics SDK, like Havok is a software physics SDK. Games already are developed without the need for a PPU (Physics Processing Unit); there is already a bunch of Physx & Havok titles that use one or the other SDK. That is the point and reason for the SDK accelerator-card; to off-load the calcs for either the CPU and/or the GPU. The Ageia Physx (formerly Novadex?), SDK, as a card is an accelerator that off-loads physics-calcs that the CPU is doing (in the SDK). Having the GPU do that will LOWER the performance of the GPU-card and overall system performance; when compared to a system with 3 seperate dedicated pathways: CPU, GPU, PPU.
Nvidia bought a software SDK and a PPU-card implementation. If you think they want to kill parallel processing, by incorporating physics calcs all-in-one on a GPU, than I think you are mistaken. That is the point of the PPU. The business IFs are: how to make a PPU a, more or less, design standard, can you incorporate it onto a main-board (thus not waste a PCI-slot), incorporate the chip and a pathway on a GPU-card (and not interfere withe render-pipeline), things like that.
If a Geforce-card is optimized for physics than it will be for the Physx SDK (the one Nvidia bought); but that already exists it's called an Ageia Physx PPU-card.
Last edited by GC_PaNzerFIN (2008-02-05 10:38:54)
So? I know AGEIA's PPU is hardware support for the SDK....topal63 wrote:
Physx is an SDK. Ageia Physx as a card is hardware support for an SDK. What aren't you understanding?
____Tech ARP: Wow.. That was some purchase!
NVIDIA: Yes, but it was in the pipeline for some time now. We only finalized the deal with AGEIA moments ago.
Tech ARP: Really? Why? Don't you already have SLI Physics?
NVIDIA: We decided that running physics on the GPUs just wasn't going to cut it, not even with the amount of power our GPUs have.
NVIDIA: We will continue to support SLI Physics and Havok FX, but eventually we want the PhysX silicon to handle physics acceleration, leaving the GPU to do what it does best - graphics.
Last edited by topal63 (2008-02-05 12:25:44)
why are you thinking that they would use the AGEIA PPU? there is nothing superior with their tech compared to what nvidia can do.topal63 wrote:
That's exactly the point. A multi-core GPU/PPU card would be EXACTLY like having an Ageia Physx card built into a Geforce-card (if they make one); considering Nvidia now owns the Physx SDK.
And, if Nvidia releases their own separate PPU-card it will support the Ageia Physx SDK, maybe that will be called this instead the "Ageia Physx Card powered by Nvidia?" What's in a name? It will still serve the near exact same function as the currently existing Ageia Physx card.
now THAT is what I have been trying to say. Since the PPU doesn't have to be the original AGEIA one.PS: What would be a good thing is this! A PPU that is both hardware support for the Physx and Havok SDK all in one.
Last edited by GC_PaNzerFIN (2008-02-05 11:15:19)
Last edited by topal63 (2008-02-05 11:32:24)
Nvidia bought 3Dfx.Gooners wrote:
NVidia Made the worlds first GPU?
