Quote Originally Posted by dave_graham View Post
i'm sorry, but AMD has stated (at least in SEVERAL emails) that BA steppings are production. whether that's a kludged B1 is beyond me, i'm simply reporting what i've been told.

if that makes me "full of crap," Periander, then so be it. Blindly trusting engineers that live off this stuff day to day is not something that i do, but I do trust those who use this stuff day to day.

I didn't pull the BA stepping out of my ass. It was specifically stated that:
a.) the B1 stepping was not production (could be my week, sure, but...i digress)
b.) that BA/B2 steppings solved a significant errata present in the B0/B1 steppings (aka DVT spins) that increased the nominal performance of the Barcelona in STREAM processes

so, we'll see...won't we? as it stands, AMD is at least 3-4 weeks off from providing the channel with parts, so....since all of August isn't accounted for yet, we still might see some changes..

anyone want to gank the microcode from my BIOS's? that might answer your questions since the northbridges (or MCs) are identified by stepping (at least on the S3992)

cheers,

dave
Dave, it wasn't my intention to insinuate you were lying. I just think you or maybe your FAE may be mis-informed. I feel I have to clarify a couple of things though:

1. There isn't such a thing as a "BA" stepping. Every stepping I've ever seen (along with everybody I've spoken to about this issue) is a letter followed by a number. Maybe it's not the stepping your speaking of...instead, could it be the revision #?

2. Steppings aren't identified by production date. As a matter of fact, the production date doesn't give too much info about the cpu itself. Rather, it's the first five letters/numbers found on the second line of AMD cpu's which give much more pertinent info than the date of manufacturing.

3. Almost ALL final/production level hardware - especially cpu's - contain erratas. These erratas are deemed "acceptable" by the manufacturers as they only affect VERY SPECIFIC situations. If there's a huge problem in the chips, they simply won't be released -> as with the case of the B0 Barcelonas. Successive revisions of the cpu's are designed to fix certain erratas, but clock-to-clock performance shouldn't be affected.

On another note, I wasn't able to pick up the s3992 as I had originally planned..... .....