Intel Corp (INTC)

[Windsock]

Well now that you mention AMD's failures...

The Performance Rating fraud has stretched to the breaking point just as it did for Cyrix.

http://www.extremetech.com/print_article/0,3998,a=36734,00.asp

Analysis and Conclusion

The PC Mark memory test results on the previous page provide a pretty strong indication that there's a delicate balance between clock rate and cache size.

When Intel introduced the Northwood CPU, with its larger cache size, the slowest Northwood was 2GHz -- the same as the maximum clock rate of the older Willamette core. There was no question as to which Intel CPU would be faster.

Here, AMD is introducing a product that clocks slower than its previous top of the line, but with more cache. The net result is that applications that are sensitive to cache size may run better on Barton, but apps which like higher clock rates (and have tight loops running in L1) will run faster on the 2800+.

This is all somewhat moot, as AMD will not be shipping a 2800+ Thoroughbred at all, opting for the slower-clocked, Barton variant. However, it is interesting to see the tradeoffs here.

Of course, you might look at the Barton 2500+ and note that it clocks more slowly than the Thoroughbred 2400+ -- but that's not apples-to-apples either, as the 2400+ only supports a 266MHz frontside bus.

The real contradiction, though, is that the Athlon XP 2700+ clocks at 2.16GHz, while the 2800+ will clock at 2.083GHz. It's likely the 2700+ will perform better in some CPU-intensive apps than the 2800+. This leaves us scratching our head over the entire product naming scheme.

All of this indicates that AMD is playing a tricky and delicate balancing act with its product numbering scheme. As we've seen, the 3000+ isn't superior to the 2800+ in every case. When compared to the P4 at 3.06GHz, the results are also somewhat mixed -- AMD is faster in some cases, Intel in others.