Understanding AMD's Roadmap & New Direction

We've been providing live coverage of AMD's 2012 Financial Analyst Day from Santa Clara today, but if you want a summary of the company's strategy under new CEO Rory Read you've come to the right place. Below you'll find links to everything we've published from AMD's FAD 2012:

AMD's Rory Read Outlines AMD's Future Strategy
AMD Outlines HSA Roadmap: Unified Memory for CPU/GPU in 2013, HSA GPUs in 2014 
AMD is Open to Integrating 3rd Party IP in Future SoCs
AMD's Financial Analyst Day 2012 - Mark Papermaster, SVP & CTO Presentation 
AMD: Flexible Around ISA
AMD Nods at Shorter Design Cycles, More Synthesized Designs
What AMD Views as Important: Tablets, Servers, Notebooks & GPUs
AMD & Compal Show Off 18mm Trinity Notebook 
AMD's 2012 - 2013 Client CPU/GPU/APU Roadmap Revealed 
AMD's 2012 - 2013 Server Roadmap: Abu Dhabi, Seoul & Delhi CPUs
AMD is Ambidextrous, Not Married to Any One Architecture, ARM in the Datacenter? 
AMD's Tablet Architectures: Hondo at 4.5W, Future Sub-2W SoC

The name of the game for AMD in 2012 is execution. Far too often at previous AMD events we'd see a roadmap with no indication of whether or not AMD would actually stick to it. Plans were always aggressive and made sense, but execution rarely mirrored what was promised on paper. AMD in turn ended up delivering CPUs sometimes years after they were promised, which obviously didn't help grow the company.

Over the past few months Rory and his new management team at AMD have been restructuring the company to improve execution. As a much smaller company than Intel, AMD had an almost impossible task competing in the x86 space, but AMD should have also been far more agile than it was given its size. AMD's restructuring is supposed to fix these agility and execution problems. The proof will be in how well AMD is able to introduce products over the next 24 months.

As far as the roadmap goes, AMD already laid out what it hopes to accomplish by 2013. The best way to summarize AMD's next two years is: APUs and servers. On the APU side, we'll get updates to all of the current lines this year (Trinity & Brazos 2.0), but we'll also get a new 4.5W APU (Hondo) aimed at the Windows 8 tablet market. Now 4.5W is a bit high for a tablet but AMD has plans to bring even lower power architectures to market in the future. Next year we'll also see the first single-chip solutions from AMD: Kabini and Temash will integrate the I/O controller (SATA, USB ports, etc...) on-die. Intel is doing something similar with Haswell. 

The big transition will happen next year, as AMD moves its entire APU stack from 32nm SOI to a bulk 28nm process at Global Foundries. This is an important move as it signifies the use of more easily synthesized designs, which enables AMD to bring out APUs in a quicker fashion and with lower design costs. Effectively everything follows the path laid by Brazos/Bobcat at this point. Note that AMD will be shipping 28nm APUs while Intel is well transitioned to 22nm, a gap that AMD has no intentions on narrowing. The process technology gap has almost always existed between AMD and Intel, but now AMD is taking a firm stance in saying that it has no intentions to blindly pursue the closing of that gap.

What once was a 3+ year design cycle for AMD CPUs and APUs now shrinks to less than 24 months (maybe even as short as 18 months) as a result of this migration to more easily synthesized designs. If this sounds a lot to you like the old GPU design cadence don't be surprised. AMD's APUs are, after all, largely made of GPU transistors that have always strayed from custom logic where possible. The big change is simply doing more non-custom x86 design.

Obviously AMD's ability to execute on this roadmap will depend heavily on Global Foundries delivering good yields at 28nm, however AMD does seem fairly confident at this point. That being said, 2013 is a very broad timeframe. Executing on the above roadmap sooner rather than later in 2013 will be the difference between a competitive AMD and one that's quickly written off.

I don't believe a move towards easily synthesized architectures is necessarily a bad thing. As long as AMD can deliver good performance at competitive prices the process technology really doesn't matter as much. Typically process technology was used as an enabler of good performance but as a newly fabless semiconductor, that strategy doesn't necessarily apply to AMD - at least not across all market segments. In other words, what works for Intel may not work for AMD.