Question/wondering about Xenos Unified shaders..

[EndR]

So I was wondering about the Unified shaders of Xenos, compared to a separated VS/PS pipelines. In a separated pipeline, there is a chance that some VS/PS units are idle, depending on how PS/VS-intensive the game is etc. (if a game can manage with 6 VS units and the GPU has 8, then 2 are idle)

How does it work on Xenos? Because of the 48 ALUs and the games we have seen so far. Are some of the ALUs being "idle"? Or can they help out with other things as well?

If im fuzzy here, tell me so I can try to explain more but Im curious to know how it works.
:smile:

 

[Bobbler]

So I was wondering about the Unified shaders of Xenos, compared to a separated VS/PS pipelines. In a separated pipeline, there is a chance that some VS/PS units are idle, depending on how PS/VS-intensive the game is etc. (if a game can manage with 6 VS units and the GPU has 8, then 2 are idle)

How does it work on Xenos? Because of the 48 ALUs and the games we have seen so far. Are some of the ALUs being "idle"? Or can they help out with other things as well?

If im fuzzy here, tell me so I can try to explain more but Im curious to know how it works.
:smile:

I think you might be a bit confused as to how they work. You don't use a specific amount of them (like your example of a game managing with 6 VS units and having 2 idle)... you have a set amount of shader progs you need to run, and the more shader units you have, the faster the jobs get done -- the idle time comes with you have a unmatched ratio of vs/ps progs and one section gets done faster than the other (so there is waiting). Unified shaders help to reduce that so the job batches end up done with a much closer proximity to each other on average.

Thare's the broad detail-less answer to your question.

 

[EndR]

I think you might be a bit confused as to how they work. You don't use a specific amount of them (like your example of a game managing with 6 VS units and having 2 idle)... you have a set amount of shader progs you need to run, and the more shader units you have, the faster the jobs get done -- the idle time comes with you have a unmatched ratio of vs/ps progs and one section gets done faster than the other (so there is waiting). Unified shaders help to reduce that so the job batches end up done with a much closer proximity to each other on average.

Thare's the broad detail-less answer to your question.

Ahh, I see!
Apparenlty, you learn something new every day.
Thanks!