I'm interested in knowing more about the hardware architecture behind these cards. For example, while I can find tons of information on the exact hardware layout and technical specifications of the R300, I can't seem to find anything similar on the GeForce FX (nv30?).
The information that I need is, given a certain program, say 100 opcodes long (either vertex or pixel shader), after rampup how many vertex/pixels can I expect to complete per clock? Do certain instructions require multiple cycles (probably)?
What I am trying to do is compare current graphics hardware's ability to do fast "math" with the capability of more directed programmable units. The information I'm seeing out there just doesn't cut it. I haven't tried talking to people at ATI or nVIDIA yet, but my guess is that people here know more about it than the people I'd easily get into contact with at either of those companies.
Thanks.
The information that I need is, given a certain program, say 100 opcodes long (either vertex or pixel shader), after rampup how many vertex/pixels can I expect to complete per clock? Do certain instructions require multiple cycles (probably)?
What I am trying to do is compare current graphics hardware's ability to do fast "math" with the capability of more directed programmable units. The information I'm seeing out there just doesn't cut it. I haven't tried talking to people at ATI or nVIDIA yet, but my guess is that people here know more about it than the people I'd easily get into contact with at either of those companies.
Thanks.
