But clearly the transition from simple multicore to many-core processing also represents a fundamental shift in battleground for a company like AMD. Software becomes much more important and to a degree it resets the competition. Intel has as little idea as anyone else, possibly less idea, about how to make efficient use of many-core processor chips. This would therefore be the right time for AMD to jump ship.
While AMD is focused solely on the x86 architecture its primary requirement is to make ICs perform like Intel's. And with multicore developments Intel is becoming harder to track it is quite probably too expensive for a fabless company that is only in the PC market, such as AMD, to do the work continually playing catch-up with Intel.
The only prize that AMD gets for that strategy is lot of expense and always being second to Intel, which may not itself be making all the right moves.
But if AMD joins the extended ARM ecosystem, in return for a few million dollars and a few percent royalty per chip it gets ready-made architectures that it can bolt together leaving it time to focus on performance at the system-level. And with full versions of Windows becoming available on ARM processors, this gives AMD the possibility to address not only the PC market but the broader consumer platform-style requirements of many OEMs.
This can get AMD out from under the thin-margin PC business into a broader landscape that roams from mobile phones to tablets to PCs. It does introduce AMD to competition from the likes of Samsung and others, but AMD has to be willing to back itself as creating value at doing something.
Now the choice between Windows running on x86 and Windows running on ARM becomes one of which is more power-efficient and which is a customers' preference.
Can Radeon GPUs be made compatible with ARM? AMD sold ATI's Imageon GPU technology that was made to measure for ARM, and Qualcomm are now exploiting it as Adreno very aggressively and effectively. So if AMD take ARM licenses, what do they do for GPUs.
License back from Qualcomm? Yuk!
Start from scratch? Yuk!
License Mali? Yuk?
License ARM architecture including Mali? Hmm.
That would allow them to leverage their existing skills to differentiate their products.
Perhaps they could work together with ARM on future GPUs as nVidia say they are doing with CPUs.Just a thought.
Now I see why AMD's CEO. Dirk Meyer leave ... doing business in corporate enterprise market (Opteron: server, HPC) is diff from end-user consumer market (ARM: elec gadget). It'd like asking a traditional dance teacher to dance in the nightclub. But the decision moving to ARM sound logical since Microsoft is also joining the party.
P/S: With exception of new Brazo platform, I think main revenue from AMD's CPU processor did come from server rather then desktop/notebook.
Let me get this right --- the suggestion is that ARM abandons a market where they seem to have a 20% share and relatively stable financial situation (when averaged over longer periods) and go to a chip market where nobody but a few handset makers are making a profit and where instead of being a me-too in the a field of two they would be me-too in a field of 10's. Does anyone really think that TI and Samsung and Qualcomm are just going to roll over? Even if AMD instantly gets 20% of that ecosystem they will still make less money than with 20% of the PC share. And let's not exaggerate the importance of the availability of Windows on the ARM architecture. Windows has been available on Itanium since day 1 and I don't think that anyone is crediting it with the "smashing success" of that particular platform.
Actually, this is a new fact in the ARMx Intel battle.
For AMD is also a good strategy, but they don´t to finish working with x86 products. They can have both. So they will enter in the Qualcomm, TI, NVidea, Marvel market.
Of course, with the experience they have from the x86 market, they will be able to offer very interesting products that will face Intel.
AMD could extend its instruction decoder to support ARM instructions in addition to x86, and it may not even cost much die space. while this might not yet save lots of power, it would be a preparation step away from x86. over time more and more ARM-only cores can be added to the CPU. also, if done right, mixed-mode code could be made possible without help from microsoft. AMD should do it simply because intel was stupid enough to sell its ARM architecture license - a serious strategic mistake.
AMD FANS always happy: Intel's CPU share of a new high - people off the hook sooner or later x86 ARM
Intel continues to lead the process - people focus on R & D IBM
Intel must share the crown of semiconductor - SAMSUNG people make better
Intel's net harvest - people apple blue sky overhead rates
The possibility of only one probable outcome is highly likely; AMD will most likely license ARM cpus. 2 factors help the domination of X86 programs- the amount of installed X86 software, and shrinking die size of future Fusion APU's, Intel has gotten a 30% increase of performance and battery life with every die shrink - Llano should gain, as well as the C class APU's - Monte Carlo simulations and simultaneous HD at the same time, on a tablet or netbook?
David Patterson, known for his pioneering research that led to RAID, clusters and more, is part of a team at UC Berkeley that recently made its RISC-V processor architecture an open source hardware offering. We talk with Patterson and one of his colleagues behind the effort about the opportunities they see, what new kinds of designs they hope to enable and what it means for today’s commercial processor giants such as Intel, ARM and Imagination Technologies.