The Future of AMD: Jim Keller and Chekib Akrout Interviewed



Company: AMD
Author: James Prior
Editor: Sean Ridgeley
Date: April 29th, 2013

Looking forward

On a recent trip to AMD's Austin Campus -- recently sold and two-thirds leased back -- I had the opportunity to talk with AMD's CPU architect and leader of all the 'cores' groups, Jim Keller. Keller was hired from Apple in 2012, and a look through his credentials shows he has massive breadth and depth to his portfolio for CPU work, including the original AMD Athlon64. Keller is notoriously press adverse, making our brief video chat between Austin and Sunnyvale -- where he's based -- all the more special.

AMD Key Moments for Q1 2013

Keller's role at AMD gives him oversight and leadership of the big cores and small cores -- ones that climb on racks with many little ARMs. A few short years ago, it would be easy to say, "IP A goes in this segment; IP B goes in this segment; IP C isn't needed." But now, we have the brave new world of maximizing performance per watt while constraining socket size, transistor count (and thus cost), and overall power envelope.

Big Core strategy

AMD's current stack features two IP lines: the Bulldozer series (also known as big cores) and the cat series (known as little cores). We currently sit at the second iteration of Bulldozer (known as Piledriver), which is found in the second generation FX and A-series APUs. On April 23, AMD announced the embedded Kabini series of APUs which use the Jaguar cores -- the same CPU core architecture the upcoming semi-custom HSA APU features in the Sony Playstation 4.

e-Kabini
e-Kabini

AMD is currently undergoing a very public and intensely scrutinized reorganization, shrinking into profitability. This extends to Jim Keller's cores groups; one of the first tasks he completed was his own internal reorganization and responsibility shuffle to align his group's efforts best with the new efforts being made by the company. As we now know, that includes ARM cores, in the form of both IP for hardware security zone functionality and 64-bit ARMv8 for servers; an AMD Opteron branded ARM processor is expected in the next 12 months.

Big Core Optimization

The ethos of AMD's x86 designs is modular, first placed by design library and then hand-tuned. This makes for a fundamentally portable design between manufacturing foundries, but still allows fine tuning to get the most of the process the chosen product is made on. Manual tuning allows for better clock frequencies, leaving AMD rather uniquely in the middle between the process library designers and 'manual placement of everything guys'.

"AMD are on track to catch up on high performance cores"

- Jim Keller, Corporate Vice President and Chief Architect of AMD's Microprocessor Cores

The best part of interview came in a nice little tidbit about core performance while discussing how much market and application awareness plays a role in core design. Many things are incremental, one of which is legacy performance on new designs. Jim confidently stated AMD are on track to catch up on high performance core, a function of design improvements. We couldn't pin down a timeline for this, but with a time scale of two years core design and one year build and test, it's not going to be immediate. My expectation is 2015.

My second interview was with AMD's Business Transformation Officer, Chekib Akrout. Chekib is another semiconductor industry veteran hired in as part of AMD CEO Rory Read's reset, restructure strategy. Chekib is a great guy to talk to, has a great approach and outlook and not at all the 'hatchet man' I had envisioned his title would make him seem. As AMD transforms, Chekib says this is a 'nice era for engineers who want to innovate.' As AMD's Reset and Restructure period comes to a close with Q1, the critical next step is Accelerate and Execute, a single quarter long drive to return to profitability and deliver on the financial commitments. Once that happens, AMD moves onto the 'transform to win' phase, which is based on server, graphics, and low power IP.

3 Steps to Win

Breaking both the x86 and Windows market dependencies are key for AMD's diversification, something not going unnoticed inside AMD. In fact, it is a driving force for its strategy -- without abandoning x86. While there has been no public commitment to ARM outside of the licensing mentioned above, it seems far-fetched to presume a semiconductor IP design company with a heritage of a decade of semi-custom design in partnership with ARM, IBM, Microsoft, Sony, and more would leave the big new core IP out of their bag of tricks. AMD are uniquely positioned to provide solutions with multiple IP options at multiple power levels and expand their already broad product offering. Embedded systems are currently receiving a lot of attention as this is an area that will see a lot of growth. If you think about what embedded systems are (smart devices, smart grids, data monitoring, converged platform controllers), you see a need for low power, simply programmable, high performing compute platforms.

"This is a nice era for engineers who want to innovate"

- Chekib Akrout, Senior Vice President & Business Transformation Officer

AMD's first steps are to shrink into a self-sustainable footprint, pursuing safe, viable, ongoing revenue sources and then expand into new and innovative markets. Currently that means consumer gaming, dense cloud servers, GPGPU compute accelerators, and embedded platforms. Of these, three are driven by the ongoing development of Heterogeneous System Architecture, which is building momentum and reaching a final stable first state this year with the Kaveri design, launching early 2H 2013. SeaMicro continues to lead for AMD in the dense server space, and serves as a way for them to profit from the success of a competitor and show how they're in tune with customer needs for performance per watt.

Use IP in new markets

AMD customer base is another key point of 'transform to win': new customers will mitigate the effects of large OEM customers changing or cancelling orders to accommodate new third party design fund driven product lines or shrinking markets affecting one IHV who expects sales percentages to trend with whole market volume and not with consumer demand. AMD are working hard in that area, acquiring new partners like Vizio, but also they have made partners of the new competitors in their adjacent markets whereas the two traditional competitors of AMD have not. There's no such thing as gentleman's agreements in global business, but paid memberships to common standard foundations go a long way to smoothing over ruffled feathers as everybody tries to grow.

When the PS4 announcement event happened, I stated the PS4 GPU had some enhancements to it not found in current GCN designs. The publication of Gamasutra's 'Inside the PS4 with Mark Cerny' article shows that to be true, confirming eight Asynchronous Compute Engines and expanding the command list capability of each ACE from two to eight, also. This, and the other major changes AMD and Sony worked together to provide for the semi-custom APU, are all the result of work since Sony put together the feature list in 2010. The Sea Islands codename encompasses the PS4 chip, as well as the new mobile chips announced at CES, and the Bonaire GPU found in the Radeon HD 7790. I bring this up because it shows the people inside AMD continue to work hard, long hours to deliver great products amongst the turmoil shown in the press.

The next generation of graphics cards -- Volcanic Islands -- is coming this year and shaping up nicely. When you name products after places, it leads to interesting thoughts about where to hold events surrounding that namesake product. Typically places that are cheap to get travel to internationally for a worldwide congregation are preferred, so I'm off to renew my passport on the off chance I'm headed to Reykjavik later this summer. But I sure as hell wouldn't complain about being sent to Honolulu, either.