Great article and very realistic with constructive criticism, compared to the one on tom's with the usual BS like the future is fusion/vision, etc.
The recent "bold strategic move" of AMD to ARM suggests that something is very wrong. Looks like they want to enter the ARM business but only 2nd to everyone since their license doesn't allow modification of the design like the other half dozen competitors'.
As you say, Graham, it's "own mismanagement, some bad predictions" and I would add that they lack good common sense, since this is the second time AMD shift focus, the last time was 6 years ago announcing fusion/vision after burning $ 5.4 bn for ATI, and in the meantime producing a spectacular failure in their main business like Bulldozer which is more like a modern Netburst. At least they are competitive in PC graphics but they don't excel financially and have superior competitors in HPC with Tesla, and now Xeon Phi.
I didn't know what to think back in 2006 when AMD bought ATI but now I'm sure it was a huge mistake and it is the main reason why the company is in that mess.
1st, the whole idea of the merger was a vision of fusion only on paper as the future for AMD.
Fusion was not even technically feasible until 32nm. Anyway there is at least the TDP as a limiting factor to a SoC so massive and complex which means that performance can never match the high end range nor can be the most efficient.
As for the real application, most software is not ready for such parallelism, in fact only a fraction can run on so many cores like rendering and decoders/encoders.
2nd, it was an expensive acquisition by a company that has barely made a profit and the alternatives were plenty to say the least.
It has puzzled me for a long time why AMD wanted nVidia or ATI so badly. I guess easy-come easy-go: after 2-3 years of fat profit common sense was gone.
The cheap alternatives make more sense when realizing that AMD needed only a platform to compete with Intel. And I don't mean VIA as it is best left in its niche x86 position.
The first candidate would have been SiS as it offered a full chipset with graphics, sound and networking while being quite cheap and stable. Now SiS has vanished from the PC marked.
Other candidates (better than ATI chipsets) were Broadcom, a server chipset and network chip maker, or ALI/ULI, a chipset maker, which was bought by nVidia. As for graphics chips (good enough to be integrated) there were PowerVR, now only in ARM SoCs, or Matrox Graphics. For network and sound chips there were quite a few companies to choose from.
But history has been made and AMD needs to focus on the future. I hope they learn from the mistakes and know what to do with some good common sense.