Two Different nVidia Cards, One as Graphics, One as Physics - Possible?

Soon I will have a brand new Geforce GTX 285 to replace my current 8800GT.

Will it be possible to move my 8800GT to another PCIX slot, put the GTX 285 into the primary slot and use the nVidia software to make the 8800GT do physics?

I have enough PCIX slots on my motherboard (4). The GTX 285 will take up 2, so the 8800GT should fit in the 3rd.

If anyone knows if this is possible, I would be grateful to be educated!

Solution 1:

How does PhysX work with SLI and multi-GPU configurations?

+ = Supported!

When two, three, or four matched GPUs are working in SLI, PhysX runs on one GPU, while graphics rendering runs on all GPUs. The NVIDIA drivers optimize the available resources across all GPUs to balance PhysX computation and graphics rendering. Therefore users can expect much higher frame rates and a better overall experience with SLI.

A new configuration that’s now possible with PhysX is 2 non-matched (heterogeneous) GPUs. In this configuration, one GPU renders graphics (typically the more powerful GPU) while the second GPU is completely dedicated to PhysX. By offloading PhysX to a dedicated GPU, users will experience smoother gaming.

Finally we can put the above two configurations all into 1 PC! This would be SLI plus a dedicated PhysX GPU. Similarly to the 2 heterogeneous GPU case, graphics rendering takes place in the GPUs now connected in SLI while the non-matched GPU is dedicated to PhysX computation.

PhysX performance review:

In multi-GPU mode, one GPU handles all graphics duties, while the second card is solely responsible for tackling all physics commands. Under this mode end users can mix and match different cards. For instance, someone who purchased a GeForce 8800 GT last year can run this card for physics, and then pick up a brand new GeForce GTX 280 to run as the primary graphics card for gaming. An SLI motherboard isn’t required for this mode either, both SLI and non-SLI motherboards are supported.

Regarding drivers: Which driver should I use?

NVIDIA® ForceWare® Unified Driver Architecture (UDA) supports SLI technology. Simply download the latest GeForce or Quadro graphics driver from the driver download page located here.

NVIDIA continually adds new SLI applications profiles to upcoming drivers. For the most up-to-date SLI application support, always use the latest drivers.

Can I mix and match graphics cards from different manufacturers?

Using 180 or later graphics drivers, NVIDIA graphics cards from different manufacturers can be used together in an SLI configuration. For example, a GeForce XXXGT from manufacturer ABC can be matched with a GeForce XXXGT from manufacturer XYZ.

Can I mix and match graphics cards with different sizes of memory?

When purchasing a second graphics card, you should try to match the memory size so that you are ensured full value and performance from your purchase. However, while it is not recommended, NVIDIA does offer the flexibility to run graphics cards with different sized memory by using CoolBits. Using CoolBits (value set to 18), you can force both of the cards to use the lower of the two memory sizes and operate them together in SLI mode. When dissimilar memory sizes are enabled to work together using CoolBits, the effective memory size for each card becomes the smaller of the memory sizes. Instructions to enable this feature can be found here.

From Wikipedia on SLI (not sure if this is correct...)

In an SLI configuration, cards can be of mixed manufacturers, card model names, BIOS revisions or clock speeds. However, they must be of the same GPU series (e.g. 8600, 8800) and GPU model name (e.g. GT, GTS, GTX).[15] There are rare exceptions for "mixed SLI" configurations on some cards that only have a matching core codename (e.g. G70, G73, G80, etc), but this is otherwise not possible, and only happens when two matched cards differ only very slightly, an example being a differing amount of video memory, stream processors, or clockspeed. In this case, the slower/lesser card becomes dominant, and the other card matches.

In cases where two cards are not identical, the fastest card – or the card with more memory - will run at the speed of the slower card or disable its additional memory. (Note that while the FAQ still claims different memory size support, the support has been removed since revision 100.xx of Nvidia's Forceware driver suite.[16])

Solution 2:

This is certainly possible!

Multiple video cards simply require different drivers to be loaded for each card. Since they are not the exact same model, it will be a simple matter to install both of them together. Additional configuration is sometimes required to get multiple identical cards to all work together.

Solution 3:

Bounty message:

Would this work with 30XX as main GPU and 10XX or 20XX as physics card? What is the state of it in 2021?

Wikipedia says about PhysX:

Initially, video games supporting PhysX were meant to be accelerated by PhysX PPU (expansion cards designed by Ageia). However, after Ageia's acquisition by Nvidia, dedicated PhysX cards have been discontinued in favor of the API being run on CUDA-enabled GeForce GPUs. In both cases, hardware acceleration allowed for the offloading of physics calculations from the CPU, allowing it to perform other tasks instead.

As PhysX has been discontinued, nowadays the question of whether the second GPU can be used for physics calculations all depends on the program that drives these calculations, on whether it supports using CUDA API.

If your program cannot use CUDA API instead of PhysX, then you are blocked.