At AMD there is a major shake up of how their business is run under the watchful eye of CEO Rory Reed. The roadmap sees the mainstream APU take centre stage with Trinity leading the way for 2012 followed by the 2nd Gen FX CPU 'Vishera'. Trinity will be on current 32nm fabrication and will feature 2-4 'Piledriver' cores that will deliver 25% more performance over Llano's 'Stars' cores and can operate at just 17w TDP for portable products.
As we can see, 'Vishera' will be holding the top spot throughout 2012 and 2013, which echoes how AMD stated, at least for the time being, will not be pushing out for the enthusiasts but instead is focusing on HSA (Heterogeneous System Architecture) and their APUs. A long list of 28nm APUs will be seen in 2013, Trinity's successor codenamed 'Kaveri' will be based on 'Steamroller' cores (only 2-4 still?), sport a GCN (Graphics Core Next) GPU and feature HSA application support. The 'stop gap' Brazos 2.0 will be replaced by the 'Kabini' APU in 2013 which will feature a GCN GPU and 2-4 2nd Gen 'Bobcat' cores which are codenamed 'Jaguar'. Finally down at the ULP level, (Ultra Low Power) for tablets and the like, we can expect to see the 'Hondo' APU this year ousted by 'Tamesh' which is AMDs first true SoC as they integrate the FCH.
The client graphics roadmap outlines how integration of HSA features takes precedent and 2013 will see 'major' GPU architecture enhancement with the successor to the current Radeon 7000 series 'Southern Islands' currently known as 'Sea Islands'.
The power technology progression outlines how AMD are gradually maximising the efficiency of their processors, by dynamically changing the resources used of the APU to which component is most efficient at the task this will help further optimise power consumption.
Finally we now see why there has been such a rocky road for AMD, we have entered a period of transition in processing design. Originally AMD expected to see 'fully fused' processors in 2015, but we now see that the target is 2014. For those who still scratch their heads over what 'Fusion' means, it is AMD's vision of integrating the massively parallel processing capabilities of GPU compute seamlessly into the traditional CPU to create a new architecture that no longer looks like either a CPU of GPU and can automatically dedicate 'the best resources' to the task at hand. In short 'GPU compute context switching'. The slides display a progression of technologies gradually integrated to achieve heterogeneous computing, lower power while improving performance.
It is a shame that AMD have not yet outlined the successor to Vishera, though it would appear they are hoping the enthusiasts will be patient.
Source 1 2