3 - NVIDIA's GeForce PhysX implementation
NVIDIA's GeForce PhysX implementation
I have to be honest here, all respect to the PhysX team for making this happen. NVIDIA has created three sets of circumstances on how you can choose to use your PhysX setup from within the PhysX driver, let's have a look:
Standard - one GPU renders both Graphics + PhysX (not ideal as you'll need a lot of GPU horsepower).
SLI mode - have two GPUs render both Graphics + PhysX.
Multi-GPU mode - GPU1 renders Graphics and GPU2 renders PhysX.
For me personally the last option is by far ideal as this is a situation where with any mainboard with two x8 or x16 PCIe slots you can use your old adapter as PhysX unit.
See, you do not need a power-house of a graphics card to deal with the PhysX calculations. Say you have an older GeForce 9600 GT or 8600 GT lying around and you upgraded towards a GeForce GTX 260 (or whatever). Then can you use your 9600 GT (or even 8600 GT) as a Physx unit and the GTX 260 for graphics.
Why do I like that so much? Because you can do something new with an old outdated graphics card and ... and you are not bound to an nForce platform since you are not running SLI mode.
What's there to dislike?
Well, if you choose the SLI or multi-GPU option, the obvious results of running a fairly high-end system with adding another GPU is obviously overall power consumption. With a second adapter even in IDLE your overall wattage will pretty easily rise to over 200 Watts. And when you have two GPUs at full load hard at work, it inevitably will increase the power consumption.
In our test case today we used an nForce 680i mainboard with a 3.0 GHz Core 2 Duo processor, a GeForce GTX 280 and a GeForce 9600 GT. Our maximum wattage peak (system wide) measured from the wall socket outlet was a hefty 432 Watt, so that's really a lot. With a higher end graphics card you'd be able to use PhysX and rendering on the same GPU. So that would be the more power efficient way to go.
For today's multi-GPU test I decided to use a passively cooled Sparkle GeForce 9600 GT for PhysX, it was still in the test rig due to this review. And since I did feel we should not at all be GPU limited to poop out some decent scores, I inserted a GeForce GTX 280, courtesy of NVIDIA, review here.
Now there's a thing you will need to be aware of in the Multi GPU mode, it's actually a Vista limitation but a second monitor must be attached to enable PhysX running on the second GeForce GPU. You must extend your Windows Vista desktop onto that monitor.
To bypass that issue, most monitors have a standard VGA and a DVI connector, right? Just use both. This limitation is related to the Windows Vista display driver model (WDDM). This limitation does not exist in Windows XP. In NVIDIA's upcoming drivers, they will be offering a workaround to improve the experience for Windows Vista users.
With a single card or two cards in SLI mode you will not have this problem.
Anyway, we have inserted the two cards, then we booted up windows, installed the new GeForce Forceware 177.79 drivers, then installed the PhysX drivers 8.07.18 and rebooted.
BTW small hint to NVIDIA - I feel the PhysX options should be moved and located into the NVIDIA control panel, everything under the big umbrella is the best way to go.
In Vista just select your monitor properties / display settings then extend the monitor to the second display (make sure the faster adapter is your primary graphics card with your monitor connected to it). Now under GeForce PhysX properties we have new options available. Let's have a look at them.
Select the GPU you want to have physics processed on, and we are now good to go my man. Check out some of the built in demos to get a better understanding of how and what Physics is all about.
We'll now swiftly walk through the PhysX Software pack that NVIDIA will release next week.