Gopher, Re: "AMD may have good engineers but people would not be playing games on the platform in mid-May of the first silicon was not until end January."
Intel first demoed the McKinley platform during Spring IDF 2001 - *two weeks* after first silicon. They were able to boot three different operating systems at the time, but it's still taken a year since then to debug the systems enough to get a high margin of reliability (especially since McKinley is being aimed at high end systems). AMD's systems managed to boot Windows 64 and show a Pong screen saver. By mid-May, they could have managed a full layer stepping, which would have solved a number of fetal bugs, allowing enough stability for more demonstrations.
From what I've seen, "stability" can be masked quite easily if you know how to avoid a given problem. Most problems occur either during bootup, during heavy traffic, or during "rare" traffic (bus locks, config accesses, etc). If you can avoid these situations, and boot up with a minimal config, you can probably run games to your heart's content, even on very early silicon.
Not that I'm trying to argue against the possibility that AMD taped out earlier in 2001, but chances are they didn't, and the press and enthusiast community have pretty much agreed that first tapeout was at the end of last year, offering first silicon in January. But your interest in finding out a more exact date is to gauge the possibility of an early Hammer release. Although the answer to this is probably moot, it's more likely that AMD will stick to their revised roadmaps, which have suggested a fourth quarter launch of Hammer. If you think this can be pulled in further, I doubt you have much to back that up besides speculation and wishful thinking, but go ahead if it makes you feel better.
wbmw |
