I wonder how this will affect benchmarking. It seems like the drivers would probably cap frame rates and use the freed resources to improve physics detail. So faster cards may have similar frame rates but better physics calculations - so how will that be measured? We could have adaptive image quality now (extra GPU cycles used when available to improve IQ), but since that would be difficult to express and since higher frame-rates sell more-costly graphics cards, it's not provided. If the GPU physics component doesn't produce a bragging number, it will never have serious support.