I've heard about nVidia doing something where the GPU also helps out with game physics. . . anyone know more about this?

Specifically, this is what I'd like to know:
1. Does it work in Linux too?
2. Does it require SLI?

it's not nvidia that's actually doing it.. it's ageia.

http://www.ageia.com/

Ageia has created a chip that you can insert on you're mainboard to cooperate with you're gfx card. Just as in the playstation 3 that's currently in development. Read ageia's site. it's very interesting. Unfortunatly I am not able to tell you if it'll work with linux. On the other hand i can tell you that PS3 will run on linux and it works there. But i don't know if it's a sony made linux distro or one bought from a company like RH or mandriva.

No this is something else. It's a program that sends all the physics processing to the GPU. But that's all I know about it.

If you don't know much about it how can you tell puntjuh that he's wrong? Software that could load-share between your CPU and GPU, well I doubt it very much and if it exists, it's bull. GPUs are very different to CPUs, they are manufactured in very different ways and their internal structure is designed to do very different things. It would take something very intelligent (and thus something that used too much CPU time to run off the host CPU) to be able to interpret what could and couldn't be sent to the GPU and then manage the different threads redirecting the end result to where it needed to be.

I've been looking for the software you spoke about but haven't found it. It's doesnt exist is what i can tell you. Here's what i managed to find out. Asus is working together to get physics chips with ageia. Nvidia on the other hand is going to work with Havok, whom also specialises in physics. If you chose for Nvidia + Havok physics then YES you are required to use SLI. You can then chose to either use the default Gfx + Gfx setup, or the GFX + Physic setup. So default is 2 gfx cards doing gfx, 2nd is 1 gfx card doing gfx and 1 gfx card doing physics. Ati on the other hand will (this hasn't been confirmed yet) use ageia's card, so they can use crossfire for dual gfx, and add a ageia chip in probably the pci-e x4.

So, the software thing does NOT exist. My source is VNU net, for the dutch people here is the link:

http://www.vnunet.nl/nieuws.jsp?id=1199410


And again i am not able to say if this will work with linux, the only thing i can say is what i already sayd in the previous post.

cheers!

Wish I could read Dutch, looks swish

haha lol, well i managed to get babelfish doing what it's supposed to do and here is the link where it's translated to english.. hope it works .

http://babelfish.altavista.com/babel...jsp?id=1199410

You sound pretty sure of yourself.

According to http://www.gpgpu.org there's actually quite a few projects that use GPUs for general computations, and using the GPU for this sort of thing is quite a bit faster than just using the CPU.

I don't see why you couldn't use the GPU for physics if you can use it for complex math equations. That's really what physics is - just a bunch of math equations where you use numbers taken from the environment, such as acceleration due to gravity and the mass of an object.

In fact, it wouldn't be much different from the nVidia/Havok solution except that I don't need to buy a new GPU and it's probably a little bit fancier. There'll be a piece of software on the machine that'll give all the physics processing commands to one of the GPUs. That GPU will feed info back into the machine, which will then use that info to make decisions.

If you only had one GPU you could have some of the GPU's vector units working on physics while the rest are still working on graphics. Voila - physics and graphics on the same GPU.

ATi and nVidia are both working on software to put load from the CPU to the GPU. It's meant to be used with sli/cross-fire. Ageia is also releasing their standalone PPU. These are two different technologies.

First hit on google:

http://downloads.guru3d.com/download.php?det=1357

It could be useful, given a cheap vid card could be less than a PPU. And in either case, it's always good to have competing technologies and techniques.