Hardware Physics Engine!

confoundicator

Distinguished
Feb 15, 2002
813
0
18,980
Here is a good example, I think, of a real benifit of paralellism. Offload the processor intensive parts (just like graphics) to a dedicated piece of hardware, in this case a PCI (maybe only PCI-e) card that handles the physics engine in games. According to the INQ we can expect to see this hardware and games that support it in a year or so.

<A HREF="http://www.theinquirer.net/?article=21648" target="_new">Clickie!</A>



<font color=blue>Good judgement comes from experience.
Experience comes from bad judgement.</font color=blue>
 
Yeah, I can't wait to see what these babies will be capable of :smile:

<font color=green> Woohoo!! I am officially an <b> Enthusiast </b>!! </font color=green>
<i> <font color=red> One new Firefox fan </font color=red> </i>
 
One of the cool things I saw in the Advanced Visual Effects with Direct3D Tutorial was a cloth and particle system simulation on Geforce6 hardware. It was basically using the vertex shader to calculate a tension matrix. The cloth was deformable and very realistic and all virtually all processing was done on the GPU. Apparently this a shader v3 only operation.

Dichromatic for your viewing plesure...
 
Its a nice idea, but I doubt its a good long term solution. It reminds me of the FPU coprocessor, which at one point was also a standalone chip before it got integrated into the CPU.

I can only wonder how much data communication is required between the CPU and this physics engine, but I would guess its pretty huge, meaning, its not all that usefull if you can't bring it closer to the CPU or eventually integrate it. It will be interesting to see how it performs though, maybe I'm competely off here, and "physics engines" will be the next GPU's, but my guess its going to be the next FPU. In other words, it will be assimilated by the CPU.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
 
I doubt it would get integrated into the CPU, as not every user will need it. If such a CPU did exist it would have being designed as a gaming chip, and since the high end gaming market isn't <i>that</i> big I can't seeing it selling, at least not for a profit + R&D.

Just my opinion mind, just like you said yourself, I could be way off.

______________
Who's the man with the master plan?
 
I can only wonder how much data communication is required between the CPU and this physics engine, but I would guess its pretty huge, meaning, its not all that usefull if you can't bring it closer to the CPU or eventually integrate it.
Well, with a standard PCI Express or PCI interface it's going to have to be a pretty low bandwidth system. You're right though, integrating it is the better idea. You're wrong about integrating it with the CPU though. The much more effective point of integration in this case is with the GPU. (So for that matter would be the integration of the sound card into the GPU for handling 3D sound effects.)

I've been dreaming of the day when some bright spark finally integrates the 3D graphics, sound, and physics engine into one core API for years. I mean imagine how much data is needed to produce the graphics, sound, and physics seperately that could be saved by integrating these into one device.

Well ... perhaps two devices, with one interface. I could easily see an SLI-type setup with a graphics card and a physically seperate physics/sound card.

IMHO anything short of that level of integration is wasteful. Not that this goofy PhysX chip wouldn't be a step forward, but really I'd rather wait two years for a fully integrated Game PU (name chosen for its acronym equivalent to Graphics PU) than one year for a completely seperated Physics PU.

Of course the graphics and coding would likely be a lot different under the fully integrated Game PU than it is now, because you'd be more concerned with higher-order object associations than with direct verticies and normals being drawn. But updating the GPU programming world to the 21st Century is another topic all together. **LOL**

<pre>Antec Sonata 2x120mm
P4C 2.6
Asus P4P800Dlx
2x512MB CorsairXMS3200C2
Leadtek A6600GT TDH
RAID1 2xHitachi 60GB
BENQ 16X DVD+/-RW
Altec Lansing 251
NEC FE990 19"CRT</pre><p>
 
Not every user needed an FPU in the old days, actually, only CAD used it. Not every user needed a GPU (Voodoo), only some games needed it.

But I see an "integrated physics unit" more general purpose, sort of like SSE on steroids.

Oh, and of course gaming is a big enough market, ATI and nVidia make a living on it :) What is high end today, is mainstream tomorrow, don't forget that.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
 
You're wrong about integrating it with the CPU though. The much more effective point of integration in this case is with the GPU.


yea good idea but how many freaking transistors can they fit on a single die? the damn things will be 10cmx10cm square if they add anymore

-------
Work sucks.
 
hey, we're pretty close in the folding race, don't you think? Been fighting for place between us a bit now!
Yeah I did notice. You're less than 100 points ahead of me. Must... add... more... 'puters... 😎

<font color=blue>Good judgement comes from experience.
Experience comes from bad judgement.</font color=blue>
 
Anyone have any idea when this PPU is going to be availiable on the market. I read the artical on the inquirer and the thing sounds damn cool, apparently it can perform like 300 times more physics calculations than a normal CPU.
 
You can expect to see such cards in shops by the end of the year, and the company will release the things when enough developers finish their titles. µ



<pre><font color=red>°¤o,¸¸¸,o¤°`°¤o \\// o¤°`°¤o,¸¸¸,o¤°
And the sign says "You got to have a membership card to get inside" Huh
So I got me a pen and paper And I made up my own little sign</pre><p></font color=red>
 
So says the inquirer...

<pre><font color=red>°¤o,¸¸¸,o¤°`°¤o \\// o¤°`°¤o,¸¸¸,o¤°
And the sign says "You got to have a membership card to get inside" Huh
So I got me a pen and paper And I made up my own little sign</pre><p></font color=red>
 
Here is a interview with the company's founders on Gamers-Depot, as well as some pics of a reference board.
<A HREF="http://www.gamers-depot.com/interviews/agiea/001.htm" target="_new">http://www.gamers-depot.com/interviews/agiea/001.htm</A>
 
>You're wrong about integrating it with the CPU though. The
>much more effective point of integration in this case is with
> the GPU.

I don't see why. A physics engine could be usefull in the GPU if all that you want to do with it, is things like animated water, smoke, clothing, that sort of purely visual stuff. But shaders should be able to do that anyhow, so I don't see the necessity of adding a PPU there.

Now if you want to be able to interact with these physical properties, as clearly is the idea now, things like ragdoll effects, you'll need to transfer the information back to the CPU. The GPU would be a poor place then, a PPU coprocessor linked through hypertransport or something inside the CPU would make a lot more sense.

>I've been dreaming of the day when some bright spark finally
> integrates the 3D graphics, sound, and physics engine into
>one core API for years.

You mean like DirectX ?

>I mean imagine how much data is needed to produce the
>graphics, sound, and physics seperately that could be saved
>by integrating these into one device

Whats the point of integrating sound onto the GPU ?? Why not an etherner adaptor or RAID controller while you're at it ?

= The views stated herein are my personal views, and not necessarily the views of my wife. =