Coming to this thread rather late, here, but a few brief points....
1) People really, seriously need to stop spouting the "15 respins" bullsh*t. In the first place the A15 revision was only ever a rumour, and it's a rumour that has since been firmly squashed. The released version is A13.
More importantly, the two digits in "A13" refer to
two different things. One digit refers to the
silicon revision, the other to the
metal revision. I can't actually remember which is which
but if we assume, for the sake of example, that the first digit is silicon and the second metal, A13 would be the first silicon revision combined with the third metal revision. Whichever way round it is, "A13" is the
third revision of the chip, not the 13th. PLEASE GET IT RIGHT!
2) It is often stated that DX9 vs DX10 game screenshots must be bullsh*t, because there is no visual effect doable in DX10 that you can't do in DX9. This statement is both true and misleading. It is true to say that any visual effect doable in DX10 is also doable DX9, but the level of
performance that you get will
not necessarily be the same. What this means in practice is that there are some effects that can be done in DX10 which cannot
practically be done in DX9, because, if you tried to do them in DX9, the performance level would drop to the point where the game would be unplayable. Thus, for practical purposes there
are effects available in DX10 that are not in DX9. DX10 allows you to do them in a
game rather than only in a still picture.
3) If R600 is really only scoring 200 3DMark points higher than 8800GTX, this is dissapointing.
4) It is not impossible that R600 is 65nm, but I have to say it's unlikely.
If it does turn out to be 65nm then clearly what has happened is an exaggerated version of what happened aropund the release of R520 and R580. R520 and R580 were worked on by two entirely independent teams. R520 suffered delay after delay, but R580 had almost no problems. The result was that R580 was actually ready to launch only about two months after R520. If we imagine that R520 had been delayed by yet another two months, it would have been quite possible for ATI to say "sod it, let's just release R580 and forget about R520 completely".
We
might be seeing the same thing here. Maybe ATI originally intended to do exactly what Nvidia did, which was to launch the high-end R600 part and then follow up at least 6 months later with mid-range parts on a smaller process, and then, once the smaller process was stable, release a die-shrink version of R600. R600 is now at least 6 months late (and coinciding with the release of the mid-range chips). If the shrunk version is being worked on by a different team that has
not experienced delays, it's possible that R600 will be scrapped and they'll go straight to the first revision.
However, I still regard this as
highly unlikely. Almost all of these rumours can be traced back to a single news-story, and it's perfectly possible that the 65nm figure was quoted in an AMD presentation, but applied only to mid-range chips (or only to a laptop-version of R600 due out later) and that one website misunderstood THAT, and everyone else got the idea from them.
My prediction remains 80nm.
Facebook
Twitter