Given the recent X-bitlabs review, I wanted to summarize just how far ahead Intel is from AMD, clock for clock, across the various applications in the review. I am comparing 3.0GHz from AMD against 3.0GHz/1333 from Intel. This will most likely represent Intel's lead in dual core through the end of the year, and also illustrate the hill (or mountain) that AMD must climb with Phenom, just to beat the current Conroe design, let alone Penryn (and also assuming that by some miracle, AMD can reach higher clock speeds than they are currently projecting!).
After just browsing through the benchmarks, I was surprised with Intel's healthy lead. I just wanted to quantify it to see if it falls within the expected 15% integer gains and 20+% SSE gains that AMD expects to get from Barcelona micro-architecture.
All results are in percentage points that Intel 3.0GHz Core 2 Duo E6850 is ahead of AMD 3.0GHz Athlon X2 6000+. I'll only go to 2 significant digits. I'll also call out specific benchmarks that are GPU limited, which I think are worthless for comparison. Note that any results within 5% are in the noise and imperceptable to end users.
Synthetic Benchmarks 3DMark06: GPU Limited 3DMark06 CPU: 21% PCMark05: Includes Tests Such as Hard Disk PCMark05 CPU: 25% Fritz 9 Chess: 23%
Office Apps Word Doc Compare: 21% Excel Data Sort: 84% (this most likely exposes an instruction level bottleneck on K8 micro-architecture) 7-Zip Compression: 17%
Image and Audio/Video Editing Photoshop CS3: 54% Sonar 6.2: 47% Premier 2.0: 25%
Rendering 3DS Max 8: 29% Cinebench 9.5: 12%
Scientific ScienceMark 2.0, Primordia: 1.3%
Games Quake IV: 17% Half-Life 2: 30% F.E.A.R.: 3.7% Company of Heroes: 23% Supreme Commander: GPU Limited Valve Source Engine: 33%
Summary
Wow, there's a lot of data here, and many of these seem far more optimistic than I could have hoped. AMD stands to close the gap, assuming we take the optimistic K10 scenario defined above (+15% boost in integer and >20% boost in SSE). This is reasonable to assume given the micro-architecture changes, but no one says that AMD necessarily hits these targets.
On the other hand, it may not even matter. On many benchmarks, Intel is >15% ahead of Athlon X2 in integer tests, and often >20% ahead on SSE based tests. Can Phenom close the gap? Can Phenom even ramp in clock speed close to 3.0GHz...? It will be interesting indeed to see if it can. And by some miracle, if AMD can close the gap fully with Intel, what benefits will Penryn bring to the table?
Based on my analysis, I think it's reasonable to assume that AMD will have a hard enough time just catching up with Conroe on all apps. On some apps, such as Cinebench, ScienceMark, and maybe some games such as F.E.A.R., the gap seems easier to overcome. On many others, such as many of the audio/video tests, or 3DS Max, it seems inconceivable that AMD can pull out ahead.
Average speed increase of 2% on 1333MHz bus speed-E6850
This indicates that a single socket dual core Conroe is not significantly constrained by the FSB bandwidth. Howover the situation is likely to change for a quad core and other multi socketed systems. These systems suffer greater degradation by the constraint of a common FSB, and the performance scaling falls off with increasing number of cores.
So I expect a much greter improvement due to the faster FSB for quad core and multi-socketed systems. The impact on the V8 configuration should be a lot better than this measely 2% improvement.
Market Data 
Markets 
AI
