AMD FANS HAVE BEEN WAITING for the new Barton core to sweep them off their feet for some time. It hasn't quite managed that. In fact, it seems a bit of a let down at first. The new Barton core with its 512KB of onboard level 2 cache was supposed to be the bee's knees. Instead it has turned out to be questionable as to whether the new top of the range, the 3000+, has earned that performance rating.
The new chip's real speed is 2.17GHz. That's the same speed as an Athlon XP2700+ with a Thoroughbred core. Does the extra 256K of cache really give an 11% increase in performance? A quick look at the various sites which have been benchmarking the Barton shows that the answer is, only sometimes.
Had AMD chosen to release the new processor at the same clock speed as the previous top of the range XP2800+, 2.25GHz, there would almost certainly have been no complaints from anybody. As it is, in some tests, the XP2800+ actually runs faster than the new processor. That's a rather nasty own goal on AMD's part.
Joe D'Elia, Director of Research for Compute Platforms with analysis firm iSuppli, gave his take on the situation. He was convinced that AMD had made a mistake in not releasing the Barton at the same clock speed or more as the older Thoroughbred. That small difference of 0.08GHz could have pushed the new chip to be definitively faster than its Intel competitors. Instead the top of the range Pentium beats the Barton in several tests.
D'Elia also made a valid criticism of the AMD PR rating. While it was once a useful indicator of processor power, that idea seems to have gone by the wayside. When the PR rating was first introduced, it was a comparison of the current processor against the original 1GHz Athlon. AMD seems to have moved the goalposts with every release of a new core. The PR rating has a life of its own that bears no relation to anything other than AMD's marketing decisions.
D'Elia thought AMD were guilty of a PR rating too far and, having read reports on many sites, it seems likely that he is right. Marchitecture has won over common sense. AMD could have released the new processor at 2.25GHz with very few problems. Reviewers have been managing to overclock the chip far beyond that speed with ease. The chosen speed has to have been a political choice rather than a technical one. But why?
After some consideration, several suspects can be removed from the list. It's not a speed problem, 2.25GHz is easy to achieve with the Barton core and the 3000+ was originally roadmapped to run at that speed. It's not a competition problem, AMD would have been happy to be ahead of Intel again. That leaves the problem of time and Microsoft.
Microsoft failing to deliver a 64bit version of Windows for Clawhammer has cost AMD dearly. It means the Athlon XP has to struggle on as the company's top desktop processor for at least six months longer than originally intended. AMD must have a fair idea of how far it can push the Barton and it seems that it's not far enough to last the course.
The answer was obvious. Trim back the clockspeed by at least one notch. The processor won't storm the market as it was meant to but it will last a little bit longer. µ