AMD has a strategy for desktopchips, laptop apu’s and gpu’s, published. The focus on apus and less on the pace of Intel in transitioning to smaller productieproced lwa’s. AMD does not exclude the use of ARM architecture is not.
For the laptopmarkt should 2012 be the year of the Trinity apus. That should be the top and mainstreamsegment operate and will be improved Piledriver cores. The Trinity apus are 32nm produced, but will in 2013 be succeeded by 22nm apu’s with Steamroller cores. The budget will be covered with the new Brazos 2.0-apu with Bobcat cores, which still (40 nm) are produced.
AMD wants to be with a frugal apu with a tdp of 4.5 W, the tablet market to conquer. This Hondo-apu is also on (40 nm) fried. Brazos 2.0 and Hondo save the 32nm-step and in 2013 followed up by, respectively, the Kabini and Temash-apus on 28nm to be produced. Like all apus have these 28nm products a gpu on board.
To the apus, and soon, especially the tablet-apus as Hondo and Temash, quickly on the market and in products get, review AMD are manufacturing strategy. That must be the so-called time-to-market shortening and AMD faster indenting on the needs of the consumer. It is deliberately chosen not to fast on a smaller process, such as Intel’s transition to 22nm and upcoming 14nm processors. Main focus for AMD is, for now, the consumer segment, where the soc will have an important role to play. This would AMD not negative towards the ARM architecture, in addition to the x86 architecture.
In addition to products for the mobile segment, will also continue to cpus for the desktop and servers as part of AMD’s product range, but, partly because of the conscious choice to not on the basis of process technology with Intel to compete, AMD the confrontation with Intel in this area to avoid. Piledriver and later Steamroller should have the performance steadily improve, while the socket will not change. Would the gpu cores thanks to a new technology, Heterogenous Systems Architecture, to better use. HSA distributes a workload across cpu and gpu cores depending on which cores are best suited for the task. Not only x86 and gpu parallel cores with HSA can be controlled, also ARM-cores would can be addressed.
