Re: You're comparing launch frequency of a new design (Core 2) on a very mature Intel 65nm process, with that of a mature design (initial 90nm K8) on a very new (i.e. very new when initial 90nm K8 was launched) AMD 90nm process!?!?
Very well, how about comparing it to AMD's first dual core processor, which launched about 6 months after the first 90nm core, just like Core 2 launched ~7 months after Intel's first 65nm part. It clocked up to 2.4GHz, and by the EOL (upcoming 3.0GHz part) it had 3 additional speed bins and 25% more frequency. Applying this to Core 2, we should at least expect 3.66GHz out of it. Although, I think that would be too good to be true, but I wouldn't be surprised to see at least 3.33GHz later next year, soon after which Intel will begin shipping 45nm parts.
Re: I agree that K8 is nearing EOL. Still rev G will improve faster in speed than 90nm rev F because I think it will see more process improvements. Fafaik the roadmap for CMW doesn't feature major speed jumps either. For now it might not be unreasonable to assume that until AMD comes with either the new mobile core, or rev H, that the current performance gap between Intel's CMW line and AMD's K8 line will remain approximately as big as it is now (comparing DC CMW's with DC K8's), with the gap being larger in desktop and mobile than in the server segment.
I don't disagree with this, but I'll reiterate my previous point, which is that AMD's competitiveness will not improve with the Rev-G cores, since they will underperform the 90nm generation for quite some time to come (at least the next 3 quarters).
Re: 65nm rev G allows AMD to shift to dual core X2's at a fast pace. X2's even at somewhat lower speed perform better and are hence more competitive than single cores. So in the sense that rev G allows a much richer DC/SC ratio rev G is more competitive. Only if you compare an 'average' bin 90nm DC with an average bin 65nm DC you are probably ('probably' because we don't know what 'average' is) right that it is a step backwards.
It doesn't exactly improve AMD's competitive position when they move forward at the same pace as Intel. Right now, Intel said they will be 90% converted to dual core in the performance segment by the end of the year. AMD is trying to ramp dual core production, and 65nm is necessary to even meet Intel's pace, but it falls far short of beating it.
Re: It seems you're saying that rev G will outclock rev F some time (e.g. two to three quarters or so) after launch. If so I agree with that. I think the design goal of rev H was obviously to improve IPC, not clock speed perse. Therefor evolutionary jumps in speed are more likely imo (comparing DC rev H with DC rev G) than any revolutionary jumps.
Actually, what I think is that AMD will not staff the Rev-G core to catch up to Rev-F in frequency, but will rather staff Rev-H to achieve clock rates on par with the 90nm generation. Rev-G is only an interim core designed to ramp 65nm prior to Rev-H. While I agree that rev-H is more of an IPC push instead of frequency, the fact that it will meet Rev-F specs while Rev-G will not actually makes it appear as if it's receiving a frequency boost.
In other words, expect Rev-H dual core parts to clock around 3.0GHz, give or take a speed bin (quad cores will clock lower), while G-step may not clock far above the 2.6GHz launch frequency. So when comparing G->H, it looks like a boost in frequency. When comparing F->H, it will be more clear that AMD prioritized IPC gains over frequency. Later, when the core has been proven, they will continue to ramp frequency well into the 3.xGHz range. Of course, that will be using aggressively targeted 65nm transistor sizes, just as AMD's current 90nm uses aggressively targeted transistor sizes.
News 
Market Data 
Discover 
