AMD is keen to discuss the future of its HSA-based APU design, but is a bit more cagey on its FX family of chips.
AMD held its latest Financial Analyst Day in the US yesterday, offering a glimpse of the company's plans for the future as it continues its fight against bitter rival Intel.
Speaking at the event, chief technology officer Mark Papermaster declared that the Heterogeneous System Architecture (HSA) formerly known as Fusion is to get an accelerated roll-out schedule. While the company had previously planned to release fully-integrated HSA processors by 2015, Papermaster claimed at the event that the first fully-functional HSA GPU cores will be available in 2014.
The first Fusion APUs, which combine GPU and CPU technology on a single chip, have been available for some time, but Papermaster's vision goes further. In 2012, he explained, chips will be released that give the GPU access to the CPU memory for increased efficiency and data sharing between the two.
In 2013, this will be further enhanced by doing away with the concept of independent memory altogether. Instead, both the GPU and CPU will share a single unified memory space. The benefit: data can be processed by either subsystem without the delays associated with moving it to different areas of system memory.
Finally, 2014 will see the launch of the first HSA-compatible GPUs. Unlike current general purpose GPU technology, Papermaster claims that the HSA-compatible GPU will be cable of switching its compute context on the fly, executing code on whichever process makes most sense from a performance perspective.
AMD chief executive Rory Read repeated a claim from late last year that his company had reached an inflection point, claiming in his keynote speech that AMD is primed to 'capture
' that point. To do so, Read declared a renewed focus on the entry-level and low-power markets while referring to his company's ongoing battle with Intel at the high end as an 'unhealthy duopoly.
The new focus for AMD, Read explained, will be on mobility - tablets and ultra-portable laptop systems - highly-parallel cloud servers and embedded systems, along with emerging markets.
That's not to say that AMD is giving up on the high-end market, though. In an updated roadmap shared with analysts, the company promised to release its second-generation Trinity APU series this year with a third-generation 28nm HSA-equipped Kaveri family due in 2013. Kaveri, the slide explained, will include improved x86 cores based on the Steamroller architecture for increased instructions-per-clock performance with power saving enhancements along with Graphics Core Next (GCN) GPU capabilities.
The FX series isn't going away, either. While adoption of the Bulldozer-architecture chips has been slow in the consumer market, the company is still on track to release a successor in 2013. Dubbed 'Vishera,' the second generation of FX processors will include between four and eight Piledriver CPU cores. Interestingly, however, these will still be built on a 32nm process size despite the rest of the company's products having transitioned to 28nm by that time.
At the lower power end of the market, AMD revealed that Brazos will give way to Kabini, a second-generation low-power APU based on Bobcat-successor Jaguar cores and again featuring HSA-based technology while skipping the GCN upgrade of its more powerful brother.
Finally, at the ultra-low power end of the market, 2012's Hondo will give way to Temash. Again based on a 28nm process size, the Temash APU will include Jaguar cores on a system-on-chip (SoC) design featuring an integrated Fusion Controller Hub (FCH). This is AMD's tablet and embedded offering, designed to compete directly with the likes of ARM in passively-cooled systems.
On the graphics front, AMD promises to continue its twelve-monthly release cycle with a new GPU family dubbed Sea Islands. Based on the same 28nm process as the current Southern Islands family, Sea Islands will be based on a new GPU architecture and include a subset of HSA functionality. The result, AMD claims, will be significant performance boosts for both graphics and general-purpose compute purposes.
Does AMD's updated roadmap fill you with confidence, or has Read's talk of focusing on tablet and server markets got you worried that it's giving up on chasing Intel at the performance end of the market? Share your thoughts over in the forums