Is the NV20/NV2A frame buffer compression really 4:1?
XBox doesn't have framebuffer compression. It does have Z-buffer compression, which is 4:1.
A few things you might want to know:
4:1 is only a maximum number, sometimes the Z compression won't work at all, sometimes it'll work half way and sometimes it may just compress at full 4:1, and many levels in between. So on average its more like 2:1 compression.
Also the compression is for bandwidth only, so the compression of the z-buffer does not mean that the Z-buffer takes up less ram space, it just uses less bandwidth for the Z-buffer.
What sets apart the flipper from the Geforce 256 in terms of power and performance? Is it really any more powerful than the Geforce 256 or Geforce 2?
There are allot of things that set Flipper appart from those chips. Here's a few things Flipper has that sets it appart from Geforce 1/2 and even 3/4 in some cases:
- On-chip 2mb Z and framebuffer with 7.5gb dedicated bandwidth:
Geforce 1 and 2 have nothing like this. Geforce 3 and 4 do have Z- compression, which is similar in some ways in that it saves some Z-buffer bandwidth, but have nothing like the on-chip framebuffer. The advantages of the on-chip cache is that the main ram only ever gets the final frame and Z buffers. So overdraw does not effect main memory bandwidtt at all, neither does FSAA. Also the on-chip cache could be used for other bandwidth intensive things like stencil buffering ect since it has allot of bandwidth dedicated to it.
- 1mb texture cache with 10.5gb dedicated bandwidth which can hold compressed textures:
Of course all chips have a texture cache, and size isn't always everything
(1mb is massive compared to most other chips). But another thing that sets this appart from Geforce line of chips is the support of texture compression. It can hold a compressed texture in the cache abd only decompress the parts that are being worked on at that time. So its bandwidth is effectively 63gb and its effective size is 6mb!
Geforce 1/2/3/4, texture cache's don't support texture compression AFAIK (NV2A's doesn't so I would think its very similar with Geforce 4 and below).
- Early Z check HSR:
Geforce 1 and 2 don't support this. Of course Geforce 3 and 4 do.
- TEV, basically a pixel shader with extremely flexible texture reads (more so then NV2A's) but slightly less flexible combines then NV2A:
Again Geforce 1 and 2 don't support this and Geforce 3 and 4 has something similar, better in some ways and worse in others.
- Multi-Sample AA:
Basically fillrate free AA. Geforce 1 and 2 don't support this and Geforce 3 and 4 do.
- 8 layer multi-texturing:
Flipper can output 8 texture layers in a single pass. Geforce 1 can only do 1 per pass AFAIR and Geforce 2 can only do 2 per pass. Geforce 3 and 4 can do 4 per pass.
BTW, don't think I'm claiming Flipper is the most advanced chip out there, because its not. After all I'm only mentioning some of its strenghs here. Of course it does have some weaknesses, especially compared to Geforce 4 (not so much compared to Geforce 2). For instance, as you know, one of its weaknesses is that its use of on-chip ram for frame and Z buffers limits the size of those buffers. Which limits its resolution/colour depth.
Basically Flipper has allot of bandwidth saving tech that Geforce 1 and 2 didn't have and also its texturing is allot more flexible then Geforce 1/2. As I said its allot closer to Geforce 3 and 4.
