Lisa Su landed in New York City in May armed with her plans for the future of Advanced Micro Devices.
Four years into her time with AMD, Su was in her seventh month as president and CEO of a company known for its good technology but an uncertain future.
Over the previous several years, some poor business decisions combined with execution issues, a contracting global PC market, corporate restructuring and fierce competition from larger rival Intel conspired to leave AMD with dwindling market share in PCs and in data center systems, a difficult position in the graphics space against Nvidia and—except for a handful of times—losses in financial quarters.
But there were exceptions to this litany of troubles. The vendor’s semi-custom silicon business got its processors into the latest generations of the top three gaming consoles from Microsoft (Xbox One), Sony (PlayStation 4) and Nintendo (Wii U), which helped fuel a brief return to profitability in 2013 and continues to bring in revenue for the company.
Furthermore, through an effort that began under previous CEO Rory Read, AMD continues to reduce its exposure to the volatile PC industry by growing the percentage of revenue its gets from other businesses, such as semi-custom chips and server products.
In addition, the company has attracted strong chip engineers who were spending their time in the lab developing technologies that will play key roles in AMD’s future.
It was with this backdrop that Su and other executives were in New York City standing before a room of financial analysts and journalists outlining an ambitious road map that includes new chip and memory architectures, new focus on such areas as the data center, high-end PCs and graphics, and a list of markets—including gaming and immersive computing—that the company has in its sights. AMD needed to put its efforts and money into areas that could best take advantage of the chip maker’s expertise in high-performance CPUs and GPUs, energy efficiency and visualization.
“We don’t need to do everything,” Su said at the time. “We need to pick the things we can do very well.”
Seven months later, as 2015 winds down and AMD prepares to enter a very important 2016, the CEO is confident in the direction the company is going. Whether that confidence can translate into market share gains, revenue growth and sustained profitability over the next couple of years remains to be seen, but Su said the strategy is the right one for the company.
“We are focusing on markets where I know we can win, where we know we have the technology to win,” she told eWEEK. “I feel good where we are with our product road map.”
Patrick Moorhead, principal analyst with Moor Insights and Strategy, sees a vendor that is facing challenges but is capable of rebounding if everything falls right.
“Right now, they are a very troubled company with a bright future,” Moorhead told eWEEK. “Revenue, market share and profits are all down, and you’ve got to hit on at least one of those to drive success at a company. [But] I think they have a brighter future than where they are right now.”
However, it won’t be easy for AMD, according to Bob O’Donnell, principal analyst with TECHnalysis Research. The company has strong IP and some good products, but it also faces significant competition from the likes of Intel and Nvidia—not to mention ARM and OpenPower in the data center—along with a PC industry that may not see much of an upturn until later next year or early 2017.
“They know what they’re up against,” O’Donnell told eWEEK. “They’re not blind to it.”
AMD has been aggressive in building on the plans that Su laid out, particularly in its graphics business.
AMD Enters Pivotal 2016 With Product Road Map That Must Lead to Growth
Company executives see its GPUs—both discrete graphics as well as those integrated onto the same silicon as the CPU—as key assets in many of its target areas, such as gaming and immersive computing, which includes virtual reality (VR) and augmented reality (AR). They’re also important in the data center as accelerators in servers.
The company in June rolled out its Fury GPUs, which were the first chips to feature AMD’s new High-Bandwidth Memory (HBM) technology. HBM, developed as part of a consortium, essentially stacks memory chips for greater performance, power efficiency and density. It provides 60 percent more bandwidth than GDDR5, and a 4,096-bit memory interface. With HBM, the Fury GPUs deliver more than three times the performance-per-watt of GDDR5 while taking up significantly less surface area on the printed-circuit board.
AMD officials expect to bring HBM to a wide range of products, including when the next generation of GPUs are released next year.
In addition to the Fury GPUs, the company in September launched a graphics business unit, the Radeon Technologies Group (RTG), and over the past several weeks, has pushed to make developing applications on its GPU architecture easier.
The company’s Boltzmann Initiative offers new compilers for C++ and CUDA to help expand the use of its FirePro graphics technologies in high-performance computing (HPC) space, while RTG this month unveiled its GPUOpen program, which will give developers access to its graphics technology. The move will enable AMD to better compete against Nvidia in such markets as gaming and HPC.
AMD next year also will introduce discrete graphics to FinFET devices, a move that officials expect will drive power and performance improvements.
“Graphics is very strategic to us,” AMD’s Su said. “The combination of graphics and CPUs gives us very strong technologies” that can be applied to multiple markets.
That includes chips used as accelerators to improve the parallel-processing performance of servers while keeping a lid on power consumption, an important capability in HPC and hyperscale data centers. It’s an area that AMD can capitalize on, TECHnalysis’ O’Donnell said.
“Graphics is still a good opportunity for them,” he said, including in the data center. “They’re still neck-and-neck with Nvidia.”
AMD also is pursuing premium and corporate PCs and the data center with an aggressive game plan for its x86 processors. In the short term, that means rolling out new “Carrizo” chips that leverage the company’s “Excavator” core. In September, the company unveiled the latest Pro A-Series accelerated processing units (APUs) aimed at the commercial client segment, which AMD had essentially abandoned until a couple of years ago. It’s now core to the chip maker’s corporate strategy.
Officials are expecting steady adoption of Microsoft’s Windows 10 operating system next year to accelerate the refresh of commercial PCs, though—like many industry analysts—they don’t expect the overall PC market to see an upturn until 2017. That said, they expect global shipment declines next year to be less than in 2015, in which shipments are forecast to fall more than 10 percent.
Hewlett-Packard at the same time announced it is using the new APUs in its new EliteBook 705 3 business notebook. HP’s embrace of the chips is another example of AMD’s growing presence in the space, according to John Taylor, corporate vice president of worldwide marketing at AMD.
“AMD is in some compelling designs,” Taylor told eWEEK. “Companies are making nine- and 10-year bets on AMD.”
However, what’s drawing more attention is what comes next. Late next year, AMD will roll out the first of its Zen processors, which will include an entirely new x86 core that has been more than two years in the making. AMD officials expect Zen to carve the path back into high-end PCs and servers.
AMD Enters Pivotal 2016 With Product Road Map That Must Lead to Growth
The architecture will include simultaneous multi-threading (SMT)—a technology similar to Intel’s Hyper-Threading—for improved performance, support for DDR4 memory and a 3D FinFET transistor design for 14-nanometer and 16nm chips. AMD’s current processors are 28nm. Zen will offer a 40 percent performance improvement over current chips.
The Zen architecture will first appear in desktop PCs, and then will find its way into servers, according to CEO Su. While AMD will release the first Zen chips later next year, the company won’t see its first full year of revenue from the architecture until 2017. Company officials expect the Zen architecture to be a cornerstone in a broad range of products. In August, AMD revealed that it was developing a high-end chip aimed at supercomputers that will include as many as 32 Zen cores and an unknown number of its upcoming “Greenland” GPUs.
The company has taped out several Zen products. Officials also have said that engineers already are working on the next generation of Zen, dubbed “Zen+.”
AMD made a significant jump in the data center more than a decade ago when it released its first Opteron chip, which enables 64-bit x86 processing, at a time when Intel was pushing its Itanium architecture for 64-bit systems. AMD saw its market share go up to about 30 percent, but it eroded over time due to Intel’s innovation around its x86 Xeon processor, missteps by AMD and changes in the marketplace.
With Zen, AMD is serving notice that it plans to become a player again in the server market, and that the data center is “not an all-[Intel] Xeon play,” Su said.
“It’s a huge market that has a lot of opportunity,” she said. “It’s also a market you have to be ready to invest in for the next five years. Zen is the key to our data center play.”
Moor Insights’ Moorhead said Zen represents a shift in AMD’s thinking back to focusing on CPU integer-performance-per-core—similar to Intel’s efforts—after several years of targeting acceleration performance. With previous architectures, including its previous disappointing “Bulldozer” core, AMD made a significant bet on parallelism that didn’t work out. With Zen, AMD “is getting back to the basics … of the CPU,” Moorhead said. “That is what makes Zen so incredibly important.”
Nathan Brookwood, principal analyst with Insight 64, said Zen will be important to AMD’s plans going forward, but that it will be important for the company to execute on its plans.
“Whenever Zen shows up, if it’s not a home run, it will at least be a triple,” Brookwood told eWEEK. “But if it’s too little, too late, they have no backup plans.”
A driving force behind the development of Zen, Chief Architect Jim Keller, left the company in September, but analysts said they don’t expect the program to be hurt by the departure. “The sense that I got was that a lot of the hard work had been done by then,” TECHnalysis’ O’Donnell said.
To make room for the x86 server push with Zen, AMD officials pulled back on some of the company’s plans for its ARM-based chips. The Opteron A110 “Seattle” is shipping in what Su has called “modest volumes,” but the K12 chip, based on a custom ARM core, will launch in 2017 rather than next year.
There is an opportunity in the data center for AMD, Moorhead said. Intel holds more than 95 percent of the market, and OEMs are looking for a second source of silicon to drive innovation, lower prices and protect them against supply chain issues. Zen could be the technology that enables the company to be that second source, the analyst said.
“Server vendors are reeling from the lack of competition,” Moorhead said, adding that AMD’s products don’t need to match the capabilities of those from Intel, but need to get close.
Su said it’s less about having a choice of vendors and more about a choice of products. AMD is bringing a competitive road map that includes CPUs, GPUs and accelerators that give OEMs and end users a wider range of options for products that can be optimized for workloads.
AMD Enters Pivotal 2016 With Product Road Map That Must Lead to Growth
“It’s not one company that is going to have all of the innovation in their products,” she said.
At this point, any gains for AMD in the data center would help, O’Donnell said.
“They went from having a huge position in servers a few years ago to almost zero, so any gains they get in servers would be good for them,” he said.
A challenge for AMD will be that the company is not the only vendor vying to be that second silicon supplier. ARM and some of its chip-making partners, including Qualcomm, Applied Micro and Cavium, are looking to push the low-power architecture commonly found in mobile devices into servers. In addition, IBM is looking to expand the reach of its Power architecture deeper into the data center through its OpenPower initiative.
AMD officials also expect to see continued gains through the company’s semi-custom efforts. At the annual Credit Suisse technology conference earlier this year, Su noted that the gaming console cycle is five to seven years.
“The peak of game consoles are probably anywhere from the third to fourth year,” she said, according to transcript on Seeking Alpha. “We believe this year game console units are up relative to last year. From our visibility, next year will be up relative to this year.”
The CEO also said there are other semi-custom design wins coming in the second half of 2016. It’s a healthy pipeline, she said.
AMD’s roadmap is broad and touches on a wide array of areas. Yet it still remains to be seen whether they will help the company in efforts to gain market share and return to profitability. The weaker-than-expected PC market this year helped push back AMD’s plans to return to sustainable profitability in the second half of this year, and now analysts are saying it could happen within the next 12 to 24 months.
The company over the past couple of years has undergone a painful restructuring to streamline the business that had included several rounds of job cuts, with the latest being announced in October, when officials said they planned to lay off about 500 people, or 5 percent of the workforce.
In addition, the chip maker continues to reduce its exposure to the PC industry. Last year, 60 percent of its revenues came from the PC industry, while Su said during the Credit Suisse event that she expects that to drop to 50 percent by the end of this year or early in 2016.
The company faces its share of challenges, from whether the PC market rebounds to competition from Intel, Nvidia and others. However, for AMD, much of that will come down to how well the company executes on its roadmap, according to analysts.
In addition, all the milestones in the roadmap mean there are a lot of variables that have to be managed. Next year “will be the most pivotal year that they have had,” Moorhead said. “They have to be successful. … It is a make-or-break year for AMD.”
TECHnalysis’ O’Donnell noted that there have been unconfirmed media reports that some tech vendors—including Google and Microsoft—have been interested in buying AMD.
“It’s not going to be an easy path for sure, and it’s one of those things where it’s not inconceivable that they could go away,” he said.
AMD’s Su noted that the chip maker is entering an important time in its history, but added that it will come down to how the company performs. And that is where she wants the responsibility to lie.
“The next 18 months is our payoff cycle,” she said. “We have to show the industry, our customers, our team that we are executing on what we have committed to. We have great people and great ideas. The difference will be how we executive. In that sense, it’s our story to define.”