Originally posted by: instantcoffee
In my previous post, I am linking to AMD's strategy about implementing dedicated hardware support for Havok API in their CPU's.
And how does this negatively impact Nvidia's push for GPU-accelerated PhysX? Nvidia's GPUs will run on any platform that supports CPU physics, both Intel and AMD, and benefit from any such optimizations. The end result is no advantage gained and no penalty from CPU optimized physics.
I didn't know that Cheng asked me a question I was to answer.
No he didn't, he gave the traditional non-sequitur marketing babble that when translated means "we don't have jack." Of course his comment begs the question, "when does it make sense", which is the question I'm asking
you. When do you think it makes sense for Intel to license a GPU accelerated version of their propriety physics engine, given they do not have a discrete GPU solution capable of accelerating physics beyond that of a CPU. When do you think it makes sense for AMD to back a GPU accelerated physics engine given they have no direct control over any physics API or middleware and are completely reliant on 3rd party interests or open standards that are months away from public consumption?
AMD haven't answered the question as far as I've read (I've paid attention to physics for many years now). I would assume, from what I've read, that they are referring to ingame physics that is crucial to the gameplay itself, not only as decoration.
Of course they haven't answered the question, as they do not have a solid answer. This is what makes this so comical, there's AMD and people like you advocating "nothing" or "anything" over PhysX when PhysX is clearly the better solution right now with at least an equally optimistic outlook.
In-game physics "crucial to gameplay" with Havok are going to run into the same road blocks that we see with current titles that use PhysX. Simply put they won't exist until there's a
de minimis baseline that allows for the same gameplay functionality with all hardware configurations. Until then we'll be left with mostly scalable eye-candy, which bodes well for Nvidia given many titles now are console ports. Having the SDK in the hands of those developers will make scalable PhysX much easier as a result.
My personal interpetation: When they have signed the deal with Havok.
They already signed the deal with Havok. They signed it 3 years ago with Havok FX and they re-signed it 9 months ago as a paper response to Nvidia's GPU accelerated PhysX. And we still have nothing. Oh wait, we have a blog entry by Terry Makedon about a possible tech demo. Great.
Worst case for PhysX?
Currently, Havok is the most widely adopted middleware for physics. Featuring over 200 AAA game titles and then some lower budget ones. Over 100 developers have signed up with Havok. Intel is the largest GPU maker (Nvidia is largest for the discreet GPU market), while ATI is the third largest. Intel is a heavy player and owns Havok. Microsoft, another big player, have a unique deal with Havok:
Let's cut through the marketing BS and look at the numbers that matter. Intel does not have a GPU that can accelerate physics beyond the capabilities of a CPU, nor do they produce any GPUs that are capable of more than basic gaming, so their market share means nothing. Period. There is no hope of this changing until Larrabee at the earliest and even then there's much doubt whether Intel's solution will be competitive.
So that leaves us with the discrete GPU market, which Nvidia has dominated for years with around 60/40 share, a percentage that was as high as 70/30 at the height of G80/G92's dominance. So again, in a worst-case for Nvidia, you're advocating a segregated market and physics solution with Havok that actually results in diminished market share of GPU accelerated parts. And how exactly do you expect the situation or adoption of hardware physics to improve?
Microsoft-produced games will use Havok's physics-based tools and animation systems until the end of time:
http://ve3d.ign.com/articles/n...ng-Deal-With-Microsoft[/quote]
LOL. I guess its probably a good thing for PhysX that Microsoft has closed down nearly all of its gaming studios. They shut down most of their PC studios years ago, the last few shut down earlier this year when Ensemble and ACEs closed their doors.
Havok is running on the non propritary open standard OpenGL, while PhysX on the propritary closed CUDA.
Again, where's this OpenGL Havok client you're referring to? HavokFX? Or did you mean OpenCL, a standard which was only recently finalized but not yet available for public consumption?
PhysX can't run on anything less then 8000 series of GPU, leaving GPU acceleration on consoles out in the cold. Havok FX has been run on older GFX cards even and might be adopted to consoles with GPU acceleration.
GPU accelerated PhysX requires a DX10 capable GPU, as would any GPU accelerated version of Havok that relies on OpenCL or DX11. PhysX runs just fine on consoles and is equally capable of anything Havok is on the same hardware. Your continued references and level of expectations with regard to Havok FX are comical although I'm not sure you realize they undermine all of your arguments with regard to GPU accelerated physics.
Havok might be run on both ATI and Nvidia cards, while PhysX only on Nvidia GPU's. Developers wants to reach the largest audience and PhysX won't do that.
Again, if developers want to reach the largest audience, the clear choice is PhysX as Nvidia dominates the discrete GPU market. All market share indicators and anecdotal evidence relating to DX10 capable parts corroborates this. Also, as its obvious you're bent on misrepresentation and misinformation, you seem to be ignoring the fact ATI parts can run PhysX:
- Does AMD block PhysX on Radeon development?
While Nvidia opened up and provided access to its software libraries, engineers and hardware, Badit noted that AMD was less helpful. It appeared as if AMD was silently blocking the development of PhysX for Radeon.
Of course none of this should really come as a surprise as parts from both vendors need to satisfy the same compute shader requirements for OpenCL and DirectX 11 compliance and compatibility.
Worst case scenario will then be that PhysX dies.
Best case scenario will be that PhysX will be ported to OpenGL for everyone and try to compete with Havok there. Havok is still bigger and wider supported, so it might still die.
You mean ported to OpenCL? Maybe you should've read Ben's post, as has been linked previously, Nvidia has already stated numerous times that PhysX will fully support both DX11 and OpenCL. Again, as mentioned, it probably helps that Nvidia's VP of Embedded Technology sits as chair of the OpenCL group at Khronos. This would be like claiming the President of the US doesn't speak English, lol. DX11 compute shader support is fully backward compatible with DX10 parts.
End result is you have some of the largest publishers on the PC (EA, 2K, THQ) and 2 of the 3 major consoles (Wii and PS3) along with some of the most influential game engines (UE 3.0 and Gamebryo) all signing and announcing major licensing agreements with PhysX in the past 8-9 months since Nvidia announced GPU accelerated PhysX. I'd say support is clearly growing in PhysX's favor, if anything.
Havok was used without special AMD coding which, if I understood it correctly, makes them compatible with Nvidia GPU's and even ran on Nvidia GPU's. This means that Havok covers 100% of the marked. Here's from mhouston
System Architect, AMD:
<text>
http://forum.beyond3d.com/show...p=1280212&postcount=24
Ya sounds like you misunderstood. They're using Havok as middleware on top of OpenCL's standardized API as the HAL which then interfaces the various hardware. Good thing PhysX can and does the same exact thing now with their own API through CUDA and will also be fully compatible with OpenCL and DX11.....
Here you go. Perhaps GPU accelerated Havok physics for everyone. One really have to be in the pocket of Nvidia for not liking that!
PS. I had to cut a bit on your post for not making it too long and repeatative. Please ask if there were questions you wanted answered.
Perhaps we'll see GPU accelerated Havok this year, but not until it "makes sense" for Intel.
In the meantime we can see tangible benefits today with PhysX with the promise of better support in the future.
Facebook
Twitter