If a graphics card has 8 pixel pipes and 1 TMU per pipe, then is there a penalty for accessing more than 8 textures per pass? For example, GFFX has 8 pipes and 1 TMU per pipe. But the total number of textures available per pass is 16. How do these numbers add up?
Thanks!
The hardware does what is known as a loop-back, and can access a single TMU twice.
Using fragment programs, it can do a lot more than 16 dependant texture reads as well.
Originally posted by VirtualVoid:
If a graphics card has 8 pixel pipes and 1 TMU per pipe, then is there a penalty for accessing more than 8 textures per pass?
The number of pipelines is irrelevant in this context. There is an additional penalty for every added texture. That is, accessing two texture takes twice the time of accessing one, ignoring shader length and bandwidth etc.