GeForce 177.39 PhysX drivrutiner släppta

Ursprungligen inskrivet av jOnÄTÄn
Jag kör dom på ett 9600gt.
Nu kan jag köra fysiken på very high och när jag spränger ett hus i luften blir det inte de minsta lagg.
Förr körde jag med det på high och då laggade det till rejält när husen flög i luften, så jag är nöjd.

Ursprungligen inskrivet av jOnÄTÄn
glömde nämna det. Crysis.
Nu ska ja såva.

With the latest Nvidia drive to make people jump onboard the PhysX bandwagon, we did a quick and dirty test with the help of Nvidia’s own particle fluid demo. We actually have a PCI PhysX cards from the days of Ageia and we thought it’d be interesting to put it up against a Geforce card to see how both perform against each other.

There were a few unexpected surprises that we ran into; first of all we noticed that the Ageia solution was unable to render as many fluid particles as the Geforce option, which in this case was a Geforce 8800GT. In fact, it couldn’t even do half the amount of fluid particles that the Geforce 8800 GT card was able to do.

Now this might just have been down to faster hardware, but we noticed that the render was stuttering on the Ageia hardware and we couldn’t quite figure out why, as this wasn’t an issue when using the Geforce PhysX option in the drivers.

Running Fraps revealed that the Ageia card only managed about 15fps while the Geforce 8800GT was doing on average about 24fps. Again, we simply put this down to the Geforce solution being faster. Then we happened to glance over at the CPU usage widget in Windows Vista and noticed that there was a massive difference in CPU usage between the two.

The Ageia card only used between 30 to 40 percent of the CPU, taking into account that at idle at the desktop about 20 percent of the CPU is being used. Switching to the Geforce 8800GT the CPU usage jumped up to between 75 to 80 percent and at times even higher.

Even running the demo in pure software mode doesn’t use as much CPU as the Geforce PhysX option, although the demo is vastly slower using the software mode, which relies purely on the CPU. Oddly enough the software mode would produce the same amount of fluid particles as the Geforce card, no matter if the Geforce or Ageia option was ticked in the drivers.

This makes us wonder if Nvidia is playing fair and square, does the GPU actually do all the PhysiX calculations or are Nvidia offloading a whole bunch of calculations onto the CPU. Is this just a demo issue, or is there something more to this? At this moment we don’t have an explanation as to what is going on, but we’re sure that we’ve just stirred up yet another hornets’ nest.

After our initial PhysX test results, we figured we’d investigate this a little bit further, as something is going on here. A new benchmark that tests PhysX has just been released and it’s called FluidMark and we decided to give it a go.

However, we wanted to test to see if there was any difference in performance depending on CPU usage, so we also did some tests running CPU Burn-in, which loads the CPU, and as we’re using a dual core CPU we ran two instances of CPU Burn-in to load both cores to 100 percent.

The results are intriguing, to say the least. With no extra load on the CPU, the Ageia PPU scores 1,753 points, or an average of 30fps, while the Geforce 8800 GT card manages 4,921 points or an average of 83fps. A clear lead for the Geforce card, but things are set to get interesting.

When we loaded the CPU to 100 percent things changed dramatically for one of the cards, and it’s not the Ageia card. Even though the PPU took a performance hit down to 1,282 points or 22fps, it was nothing compared to the performance hit that the Geforce card took, as it dropped all the way down to 1,344 points or 22fps.

It seems like the Geforce PhysX implementation is highly CPU dependent, and although some of this might have to do with the CUDA implementation, we didn’t expect to see the kind of performance drop we saw. It doesn’t bode well for PhysX in CPU heavy games and as yet we have no idea how well it works in GPU heavy games.

We’d expect some more digging to be done on this subject, but we don’t think Nvidia will be all that happy with our findings. Hopefully they can provide us with a reason for why the performance is so poor during heavy CPU usage.