support PhysicsX and/or Havok?

[Feuerfalke]

Both cards are at the moment basically used for eyecandy only. I think it would be pretty hard to make it worth much more. Consider that not everybody will buy this hardware and if you dedicate parts of the physics to these systems, you have to make physics the same for single- and multicore-PCs and both in combination with a physics coprocessor.

As tempting as it may sound, I doubt it will be more than providing additional eyecandy.

[Flyby]

I thought Nvidia was implementing PhysX through it's drivers somehow. Maybe I misunderstood what I read (somewhere)

[Feuerfalke]

I doubt that - at least not as the PhysX-Chip. The statements were merely plans how it could be implemented, but on the technical side, the interface between the card and the mainboard still is a bottleneck, especially at higher resolutions with 16xAF & 16XFSAA.

Considering the bandwidth and functionality of this chip, I doubt a shared interface will be the solution.

On the other hand, there may be parts of the former PhysX-Chip being implemented on the GFX-card to support graphics further.

Basically the same thing with the step from pure 2D-Cards that were aided with 3D-Acceleration by the 3DFX-Voodoo-Chipset. All cried out how completely useless it was, a short while later everybody implemented the functionality on their own chip in one way or the other.

[mondo]

Originally Posted by 99th_Flyby

I thought Nvidia was implementing PhysX through it's drivers somehow. Maybe I misunderstood what I read (somewhere)

All Nvidia 8XXX cards have the ability to run the PhysX software. The new ATI boards about to come out also will be able to run PhysX too because they went into partnership with Nvidia as long as they used the CUDA programming model.

[Zoom2136]

Originally Posted by mondo

All Nvidia 8XXX cards have the ability to run the PhysX software. The new ATI boards about to come out also will be able to run PhysX too because they went into partnership with Nvidia as long as they used the CUDA programming model.

I though ATI went with Intel and HAVOQ.... No?

[Feuerfalke]

Originally Posted by Zoom2136

I though ATI went with Intel and HAVOQ.... No?

That's what I heared, yes. At least the new 4xxx-cards are being released with full HAVOK-Support.

@ urufu's post:

Be advised, that this driver-modification does NOT work with G80 GPUs. Not all 8800-series cards do have the newer G92-chipset, but you will need that in order to have the PhysX-support. Without that, you run in danger of destroying your current drivers and settings!

[Solrac]

Got all now!

Thank you Urufu_Shinjiro! Thx all!

Regards, Solrac.

[Urufu_Shinjiro]

Originally Posted by Feuerfalke

That's what I heared, yes. At least the new 4xxx-cards are being released with full HAVOK-Support.

@ urufu's post:

Be advised, that this driver-modification does NOT work with G80 GPUs. Not all 8800-series cards do have the newer G92-chipset, but you will need that in order to have the PhysX-support. Without that, you run in danger of destroying your current drivers and settings!

Right, thats why I said G92 and up.

[Zoom2136]

Originally Posted by Feuerfalke

That's what I heared, yes. At least the new 4xxx-cards are being released with full HAVOK-Support.

@ urufu's post:

Be advised, that this driver-modification does NOT work with G80 GPUs. Not all 8800-series cards do have the newer G92-chipset, but you will need that in order to have the PhysX-support. Without that, you run in danger of destroying your current drivers and settings!

OK just read something on Tom's hardware... stating that a guy is working on a way to run CUDA on ATI cards... It says that he actually is getting NVIDIA support to do so...

Here is the link:

http://www.tomshardware.com/news/nvi...hysx,5841.html

[Urufu_Shinjiro]

Yeah, nvidia is not being stingy with CUDA or Physx and have stated that ATI is fully welcome to use it (CUDA carries a license fee though). ATI just chose to go another direction.