Mr. Pedantic
Diamond Member
Of Anandtech or someone else doing a David Kanter-style re-analysis of PhysX?
You guys might remember this article here, written by David Kanter basically talking about how the CPU PhysX is somewhat deliberately written in x87 instead of SSE to gimp performance.
Nvidia said that with PhysX 3.x they'd made much of the changes required for a multithreaded SSE implementation, and I was wondering whether these changes could be seen in both the instruction composition and performance of PhysX running on a modern high-end CPU, and whether someone would be willing to perform the required tests.
You guys might remember this article here, written by David Kanter basically talking about how the CPU PhysX is somewhat deliberately written in x87 instead of SSE to gimp performance.
Nvidia said that with PhysX 3.x they'd made much of the changes required for a multithreaded SSE implementation, and I was wondering whether these changes could be seen in both the instruction composition and performance of PhysX running on a modern high-end CPU, and whether someone would be willing to perform the required tests.
