blu
Wants the largest console games publisher to avoid Nintendo's platforms.
I think we got sidetracked a bit here. The original post I was responding to was claiming that any performance issues during scenes of heavy translucency usage would be likely due to BW limits. I was disagreeing with that, resolve or not. I was saying that the BW to an embedded fb is more than sufficient to handle something which was originally written for a lower fb BW, and that there are other factors, not related to BW, that can affect the performance of a GPU under such conditions. Getting back to tex resolve, though. A texture resolve:It is a pipeline bubble you can schedule work around and thus manage, but it still remain a pipeline bubble for the GPU (rendering to texture and then using said texture take a much longer amount of time and thus GPU clock cycles, so that is more latency you have to hide with math ops) and every resolve you do to main RAM does reduce the bandwidth available to the CPU.
1. is a fixed cost that normally happens to the same amount every frame, translucencies or not, and
2. can affect CPU BW negatively, but is an overall negligible bubble for a modern desktop-ballpark GPU. A resolve of a render target is normally limited by BW to main RAM. For example, an RGBA8 720p target resolve to RAM @25GB/s would take ~150us (assuming the bus is entirely saturated by the resolve). That's not such a big deal for the CPU to work around (remember, at this time the game logic is still finishing preparing the next frame).
An embedded fb of sufficient size is pretty much the best thing there is, IMO. Of course, 'sufficient' being a key term here.It is still a worthy compromise between the very nice approach the Gs used and the first Xbox 1 with its single memory pool.
It doesn't make sense to me, that's why I'm asking ; )Well, I could say ... because it makes sense , but I will have to leave that to "birds".