Looking at page 6 of http://www.amd.com/us-en/assets/content_type/white_papers_and_tech_docs/25759.pdf, AMD did not find it necessary to package the last two K8-steppings for Opteron-products. I assume they would certainly have done so if there were any substantial 90nm delays or 90nm products would not be expected to become available for high-performance parts.
OTOH, a new stepping just weeks before deploying 90nm parts carries the message that 130nm SOI will still be manufactured for considerable time; "great yields" on 90nm probably just means "good enough for ramping wafers for shipments of server-(and FX) products".
it occurs to me that AMD really has no reason to rush out the FX57(90nm) - it's not like the current FX has any real competition - so what's the hurry?
Just curious but did you accept this exact same rational when Dothan was delayed several months ealy this year?
90nm is nice but not really *that* important at the moment. While 130nm SOI K8 can be sold at an asp for more than it takes to make AMD will be happy. Intel are quite happy with the current pricing structure too as they are getting ~$200 for the 2.8/3.0 which are their most popular P4 models now and also a binsplit Prescott is happy to be produced in quantity in :-). 90nm could be quite comfortably delayed another year from now imo before it started to bite and was actually needed. 130nm SOI K8's high asp is probably the difference between AMD's Cpu division making a profit or not and is a testament to the value of the design.