Yea. I've seen it. But that will change. He'll streamline the code, lower the CPU usage and other stuff. It'll come together right . .

Quote from thisnameistaken :

Do you think some bundled generic engine designed primarily with FPS games in mind will be better?

If you could only use it for collision etc physics while keeping the driving physics intact, or offloading the current physics processing to it, yes.

But still I think all this discussion is pointless, I don't think any of this is plug'n'play sort of thing, it could take years to implement.

Quote from Funnybear :

Yea. I've seen it. But that will change. He'll streamline the code, lower the CPU usage and other stuff. It'll come together right . .

I streamlined my business processes while honing my participatory style and my proactive attitude, all while valuing diversity!

Besides by the time S3 is out we will all have multiprocessor systems and there is no reason at all why LFS at that time couldn't take advanatge of that

i.e.
processor 1 - main game engine
processor 2 - collision detection
processor 3 - main physics engine for your car
processor 4 - physics engine for oponents cars
...
processor 255 - ...

Quote from Krane :

Well yes, but it has flaws or rather limits it can handle.. I guess you haven't seen the "too many physics objects" error message?

Huh?!

I think the PPU might have a limited life, given the other problems people have mentioned (needing it to get the full potential of a game). However read something yesterday that sounded much more realistic - move the Physics processing to the GPU(s).

http://www.neowin.net/index.php?act=view&id=32535

Quote from ButterTyres :

I think the PPU might have a limited life, given the other problems people have mentioned (needing it to get the full potential of a game). However read something yesterday that sounded much more realistic - move the Physics processing to the GPU(s).

http://www.neowin.net/index.php?act=view&id=32535

You need SM3 so the GPU can alter the geometry via pixel shaders.

Though the power of the GPUs still has to increase significantly in order to produce realistic calculations.
The consumer has to pay the price, be it for a PPU or an uber GPU.
Effects physics are easier to do but won't affect gamplay since they're "only" there for calculating flames, smoke, water ect. .

Quote from SladiVadi :

Effects physics are easier to do but won't affect gamplay since they're "only" there for calculating flames, smoke, water ect. .

yes and no ... a lot of descent3 players did and do set the flame effects settings to its loweste value cause it will show you the flaming balls that actually cause damage instead of a wall of fire ... so effects settings can affect gameplay

Quote from SladiVadi :

You need SM3 so the GPU can alter the geometry via pixel shaders.

You must be talkin about Vertex shaders...

Maybe he's talking about parallax mapping?

Quote from askoff :

You must be talkin about Vertex shaders...

It seems that the pixel shaders are more powerful. Those shaders are going to be unified anyway though.

Vertex shaders cannot be used to physics calculations - or they cannot be used effectively. You must use the pixel shader(PS) units, that can execute almost any code, if it is written correctly for PS units. I saw test where the old GF4 Ti4200 was 24 times faster than p4 1.6GHz - but there is a BIG problem. PS effects can be used effectively only when the physics elements can be "converted" to pixels - good example is model of house heating, pixels can represent bricks, mass of air on so on. Im not sure that collision effects or tyre physics can be transtaled on pixel processing.
BTW, PS 3.0 arent needed for the physics modelling, PS 3.0 are just good for it because there is no code lenght limit.

(it is a long time I have read about this, so I might be quite unexact - correct me someone if you will find any mistakes plz)

It seems to me that if physics processing has any chance of becoming mainstream then it has to get incorporated into gaming graphics cards, and a standard API for it incorporated into DirectX 11 or whatever.
Nobody in there right mind is going to buy another addon card for a miniscule number of games from developers crazy enough to support a miniscule number of users.
If its a gaming related function, which it is, then it should be incorporating into our gaming addon cards, and then a generation or two down the line, developers can assume it will be present in the same way they can just assume that pixel and vertex shaders are present now.

Physics addon cars will only become common when physics are more advanced than modern CPUs can bear. That was the same with GFX cards.

Vain

With the increased use of ragdolls and environment physics to increase immersion in Half Life 2 and other FPSes, I'd say we're not too far off from loading the CPU to the point where a dedicated physics accellerator justifies its cost. It won't do to offload physics on the GPU or an extra core on the CPU since they're not optimized for the specialized repetitive physics computations that would be required.