dividebyzero
Posts: 4,840 +1,271
AMD's main problem is that they like to release roadmaps...unfortunately the roads are built at the same pace that the Crazy Horse memorial is being sculpted at.
As long as AMD remain a hardware only company I think that they will find it hard gaining any significant ground on Intel. Magny Cours (Opteron 6000 series) was supposed to be the "Xeon killer" in servers. Reception....luke warm at best. Intel are ahead in both process node (and by extrapolation performance/watt) and the range of applications favouring their achitecture - hardly surpising given that Intel's compilers seem in widespread use both in workstation and HPC enviroments (note: AMD compilers = zero).
APU will probably be a damned-if-you-do-and-damned-if-you-don't scenario. At present AMD can sell a CPU and a discrete graphics card to the same customer. While selling that same customer an APU will probably gain marketshare in CPU sales against Intel, it also amount to one less graphics sale to the same customer. Lest we forget the vast majority of computer users will have no need for an on-die (CPU) graphics + discrete card. They are going to have to get Fusion selling in vast quantities to maintain revenue.
The elephant in the room concerning graphics is that the demands of the majority of future games is unlikely to require any significant upgrading of graphics hardware in the foreseeable future. Cards are already in release that can max out all but the most demanding of games at HD screen resolution, and sales of this generations cards (and those to be released soon) are predicated almost entirely upon the graphics horsepower needed for 5760 x 1080 + resolutions (Eyefinity/nvidia surround) and 3D gaming. I would doubt that the customer base requiring either would grow at the same rate that each successive generation of cards is being released at.
The third part of the equation is linked to the first. AMD has had to spread it's R&D resources fairly thinly due to fighting on both the CPU and graphics fronts. They definitely do not have the same resources to call on that Intel can, and probably never will, and are only making significant headway in graphics due to nvidia's slipups and TSMC's 40 and 32nm woes. I sincerely doubt that AMD would have have received such a shot in the arm that the revenue from the HD5000 series brought had nvidia's original targets been met a year ago (i.e.marketing what would effectively be the GTX 580/570/460/560 against the HD 5970/5870/5850). If history tells us anything, it's that one bad/late misstep in graphics can be catastrophic for a company (see HD 2900XT, GF100) and that the next catastrophe can be your next graphics decision.
AMD can't afford to slacken up on either the CPU/APU or graphics fronts (they also need to adhere to their roadmap timetable...or not release them). Intel I don't think will be that forgiving in the marketplace, and if the scuttlebutt doing the rounds that Intel and nvidia have reached an out-of-court settlement (regarding nvidia's suit concerning being locked out of the chipset market) is true - and that Intel will be handing over ~$US1 billion to nvidia (presumeably including licensing of SLI by Intel and possibly nvidia being able to manufacture boards using Intel CPU's) then nvidia's coffers would be well and truly bulging with potential R&D funds. nvidia have proven to be fairly resourceful even in the face of overwheming odds ( HD 5000 series) to the point of posting a profit of $US85m despite only fielding one good volume graphics card (GTX 460)....and an extra billion dollars would surely translate into more competition for AMD.
I most certainly hope that AMD can carry this off. Put simply, they need to execute rather than talking about executing. Roadmaps and pretty slide decks raise expectation and every missed deadline and every missed design goal erodes confidence in those whose expectations have not been met. From a graphics standpoint they most definitely need the "Gaming Evolved" program to actually produce results- not for the advertising revenue and gaming community kudos (though very welcome they would be) but to further transform PC gaming to the point where a discrete graphics card (or 2) remains a necessity for good gameplay. Relying on third party software developers to keep pushing the graphics boundaries without help is wholly unrealistic.
(Sorry for the editorial! -but the subject seemed to require something a little more involved that "do better")
