Digitimes says that Athlon64s will start showing up sometime in August rather than September.
That’s nice, but not terribly significant.
What is a little more significant are the speeds at which the first Athlon64s will be introduced: 3100+ and 3400+. That indicates AMD hasn’t had any big breakthroughs with any major improvements to the chips; 3100+ and 3400+ indicates 1.8GHz and 2.0GHz actual speed. That’s not so nice, but again, nothing major: they aren’t doing worse than what they promised.
The most significant piece of information in the article is the tidbit about the expected shipments of Athlon64s through the end of the year: 200,000-300,000.
To put that number in perspective, for the last four months of the year, AMD would normally make eight-ten million CPUs.
Of course, you wouldn’t expect millions and millions of Athlon64s to be sold immediately after introduction, but that figure is rather low even after taking that into account. A figure of 400,000-1,000,000 would be more like it.
There’s a number of reasons why this could be the case: lack of Windows x86-64 support for some or all of the rest of the year, problems/cost of making a 130nm processor with a 1Mb cache, lack of competitiveness outside of niche markets at 130nm against current processors, both from AMD and Intel, or any combination of the above.
In this case, the why isn’t as important as the what. Given the low numbers, Athlon64s (at least the full 1Mb cache version) are likely to be rare and expensive.
Those with a serious reason to prefer it over others will buy it. The rest of us will wait for 90nm.