Software and Hardware vertex processing

Generally, hardware is better. It depends on the software/hardware matchup.

Hardware vertex processing is done by the graphics chip (GPU).

Software vertex processing is done by the host CPU.

Your host CPU is likely to have a faster speed than the GPU, so if the CPU is *ONLY* doing vertex processing, and doesn''t have anything else to do (physics, AI, audio etc), then software VP can be faster.

"Speed" isn''t the point of hardware vertex processing at all though...

Hardware vertex processing is meant to do the work SO THAT THE CPU DOESN''T NEED TO. Hardware VP frees a lot of CPU cycles - you can use those CPU cycles for **other** time consuming things such as physics etc.

The idea is the CPU and GPU are meant to be processing stuff IN PARALLEL. While the CPU is busy with your physics, procedural texture generation etc, the GPU is busy transforming and lighting the render vertices. So which is faster shouldn''t really matter.



Do make sure you''re being fair to HW VP when profiling though - check your vertex data is in the correct type of memory for the vertex processing method in use, the correct flags have been used for buffers, API states are set correctly, not causing any stalls, retail drivers/API runtimes etc.

Thanks for your replies, guys.

Just wondering, when I run the system performance checker on Windows, the CPU load indicator seems the same whether I use software or hardware vertex processing. Why?