We were wondering why Hector Ruiz actually pushed AMD to K10, native 65nm quad-core, and after a long investigation we have a piece of the puzzle unriddled.

Since its early days in 2000 and the success of original Opteron and introduction of the first 64-bit chip, Hector realized that there was a lot of money in the server business and that it's very profitable to sell chips to server guys. They learned to respect the power envelopes, that TDP and power saving matter and that you need a lot of performance to be able to sell your chips and compete with Xeon-based products.

They did a good job with K8 chips and after single and eventually dual-core server chips, but the market wanted a quad. Intel delivered a quad, but not a native one. It took two dual-cores, bridged them to a quad-core and offered it to the market. This chip was doing well, but it was still not as fast as the native one would have been.

AMD saw an opportunity and Hector (and probably Dirk) ordered to go for K10. They probably didn't have enough engineers to do both dual and quad K10, as two separate projects which would make more sense, and AMD has decided to go for a quad K10 in the hope it would destroy Intel's bridged two dual-cores in the server business.

Apparently, it was too early for a native quad-core and two dual-cores with a bridge made more sense. Intel did two bridged dual-cores, while AMD went for native K10 and failed to dominate. In late 2008 and approaching 2009 native is the way to go and with 45nm both Intel and AMD are doing the native quad-core, something that works well for servers; and Intel even indicates a native eight-core CPU at some point in 2009.

The bottom line was that AMD failed to conquer the market in the server business and at the same time it lost a lot of ground to Intel in desktop and it lost dramatically in the notebook market. Dirk Meyer's new strategy is to gain ground in all of these markets, server, desktop and notebook, as well as the graphics business, that is doing quite well compared to anything else in the company.

The only problem of this plan is that it takes a lot of time to develop new chips and fill the gaps in the existing roadmap.