Originally Posted by XFreeRollerX
Oh I know
I'm just annoyed that AMD is SO far behind these days... Experts claim AMD won't make a decent enthusiast level comeback until mid-2009... Thats pretty far away!
It's really all about how AMD had to prioritize decisions on how to compete with the Core marchitecture. I think a lot of people on this forum would agree with the statement "Core2 slaughters any K8." Now people tend to think that to get "back in the game", AMD would have needed to have an answer to Core2. However this isn't necessarily true. Even when AMD CPUs were performance kings, they still lacked the production capacity to really penetrate the desktop market. The most sensible thing they could do is ensure that they keep as much of their server market share as possible. This was the purpose of Barcelona. AMD knew that as long as Intel relied upon a FSB that their CPU scaling would not be good. They had the advantage with the IMC and they had to hold on to it. While Barcelona does offer a singlethreaded performance-per-clock advantage over K8, the real beneifts are multithreaded performance. This doesn't really help with desktop performance at the moment, but for servers it means everything.
What if instead of defending their server market share they would have decided to go for desktop performance? It probably wouldn't have mattered because they still lack the manufacturing capacity to sell enough, and Intel could easily undercut Core2 prices. Having just purchased ATI, AMD also needed the higher profit margins from selling scalable server processors, so going after desktop performance right away would have been a terrible move. I realize that the decision to improve scaling before performance isn't very popular with most of the computer enthusiasts, but really for AMD it was their only sane option. Before going after desktop performance/market share, they had to ensure their server market share was safe, thus Barcelona. It is commonly said that AMD has no answer to Nehalem, but in reality Nehalem is Intel's answer to Barcelona. Both chips have ironically similar features, and like Barcelona, Nehalem's only real performance improvement is going to be for multithreaded applications.
Really AMD's transition from K8 to K10.5 is remarkably similar to Intel's transition from Netburst to Nehalem. Both transitions would bring improved single-threaded performance and improved scaling. AMD chose to go after scaling first, thus K10 and Intel went with single-threaded, thus Conroe. Now, it's reversed and Intel is improving scaling with Nehalem and AMD aims to improve their single-threaded performance with Shanghai/K10.5.