Will be interesting how hsa works out, but amd's faults have always been a good architecture vs the slow adoption of software that makes use of it.
Things didn't really go downhill for AMD, things went uphill for Intel. Ath 64 was decent, but at the same time AMD didn't have the cash to expand, and had they expanded production two things would have made it not work, Intel was paying out basically billions in pay offs to pc builders to not buy AMD chips and had they expanded when ath 64 launched, you're talking finding somewhere, building and ramping up a fab, basically 4-5 years minimum and several billion.
After ath 64, process technology was rising in cost dramatically, from a billion or 2 for outfitting a fab and research towards 4-5 billion, towards 10billion in the future.
If keller, or anyone else who left had stayed, Intel would still be paying people off, AMD sales still wouldn't have happened, process costs would still have spiralled out of control of AMD's budget(hence the spin off which saved AMD, zero question about that), and Intel would still have gone from half arse-ing it to actually bothering and their pockets and resources.
That phenom or bulldozer are able to compete so closely with a company spending MASSIVELY more with a freaking huge process advantage actually paints in incredibly good picture for the quality of AMD's chips. Intel's process advantage is monumental for them and something AMD simply couldn't afford. If Intel wasn't paying off people to not go for AMD cheaper alternatives for like, a decade, AMD would have made significantly more and likely expanded their production before costs spiralled out of control, but that would also in the future have spread the cost of process R&D across 2 big fabs and twice as many chips produced, which would have let AMD be profitable rather than make a loss. Intel paying off people to not go AMD cost AMD any chance to be truly competitive with Intel long term.
I have no figures for R&D but lets assume Intel have spent 10-20bil in the last 5 years and AMD have spent 4-5billion, and they are making chips a full generation of process's behind(dropping power by 30-40% and die size by a similar amount) we should absolutely expect AMD to be making chips that use 30-40% more power, are 30-40% bigger and significantly slower. Bulldozer CAN beat a Ivy in a bunch of situation and lose in others, the fact that it can get at all close is absolutely fantastic.
In terms of HSA, one thing going for it is a lot of major players have gotten into the alliance for it, so its not AMD pushing software, its like 20 companies doing it. Having all the consoles and importantly their cpu's in them means a lot if not all games will be better optimised for AMD hardware, cpu and gpu, which will end up looking good in reviews for sure.
ULtimately what AMD needs is Linux, game dev's to move to open gl or some new API, not DX, not tied in to Windows where most/all software is produced on Intel compilers.
Have a look around for AMD on linux performance, there are LOADS of software packages that on a properly compiled linux optimised for AMD gains 20+ % in performance.
Cheaper linux pc's, with all games working, with AMD not being sabotaged by software would be amazing for AMD. But MS has done incredibly well to lock everyone in to DX, its probably a large reason they got into consoles, as it keeps all games DX, and prevents a move away from Windows for the entire market
?
