AMD demonstrating Havok on Radeon GPUs at GDC

Sorry for resuscitating an old thread.

Couldn't the difference between the way the cloth moves in the two demo's simply be due to the fact that they are simulating two different types of material? All cloth is not the same, some is naturally stiffer and coarser, like denim, whilst others flow and move more, like silk?

So to be able to really compare the two wouldn't you have to know what type of cloth they were trying to simulate the of movement of? Just a thought.

 

Indeed.
We don't know anything about this cloth simulation.
They could be simulating different materials, or with different detail levels (a lower detail mesh will always have a stiffer/coarser look-and-feel than a more high detail mesh, using the exact same physics simulation), different types of forces applied, etc.
The simulation algorithms are generally approximation schemes of an iterative nature. By using less iterations, you can get higher performance, at the cost of stability/realism (we've all seen objects that just jitter endlessly because the physics routine never reaches a stable state). This makes it very hard to measure performance.

It would be interesting if someone made some kind of PhysicsMark, where the exact same meshes were used for all the major libraries, and the test conditions would be as close as possible, as far as the API differences allow.

 

Hm, the drivers are out, the SDK is out. Isn't it time for a big OpenCL demo?