AMD, Globalfoundries Reach Milestone in 14nm FinFET Development

The chip foundry says it has hit 14nm validation, which is an important step for both Globalfoundries and AMD's future processors plans.

AMD logo

Chip foundry Globalfoundries has successfully fabbed the first sample 14-nanometer chips for Advanced Micro Devices, an announcement that is good news for both the manufacturer and the chip vendor.

Officials with Globalfoundries said Nov. 5 that the company had "demonstrated silicon success" on the first AMD products using the 14nm FinFET process technology, and that AMD is now planning to integrate the high-performance, low-power process into multiple products spanning compute and graphics for everything from PCs and data center systems to emerging immersive computing devices.

"Through our close design-technology partnership with AMD, we can help them deliver products with a performance boost over 28nm technology, while maintaining a superior power footprint and providing a true cost advantage due to significant area scaling," Mike Cadigan, senior vice president of the product management group at Globalfoundries, said in a statement.

It's a big deal for the chip foundry, which was spun out of AMD in 2009 and became fully independent in 2012. Globalfoundries had been struggling to develop its own 14XM FinFET process, eventually cancelling it and opting instead to license Samsung's 14nm FinFET process. Now officials are able to say they have successfully fabbed the first 14nm sample chip.

The FinFET process enables significant improvements in performance and power efficiency by leveraging a 3D transistor architecture. Other chip makers, including Samsung and Intel (with its 3D Tri-Gate technology), also are taking advantage of the FinFET process. Proponents have said it can offer as much as 20 percent better performance and more than 30 percent better power efficiency than traditional planar transistor architectures.

The 14nm FinFET process is an important part of AMD's plans around such upcoming products as the "Zen" CPU. The company's current chips are manufactured via a 28nm process, but for the next generation of processors, AMD is skipping the 20nm step and going right to 14nm as it looks to catch up with rival Intel, which already is rolling out 14nm processors. The chip maker will use Globalfoundries' 14nm Low Power Plus (14LLP) process across its CPU, GPU and accelerated-processing units (APUs)—which integrate the compute and graphics on the same piece of silicon—according to Mark Papermaster, senior vice president and CTO at AMD.

"FinFET technology is expected to play a critical foundational role across multiple AMD product lines, starting in 2016," Papermaster said in a statement.

Globalfoundries' 14LPP FinFET is being developed at a fab in New York. The early-access version of the technology was qualified for volume production in January, while the performance-enhanced version was qualified in the third quarter, officials said. Early ramp is occurring this quarter, with full-scale production set for 2016, they said.

AMD's Zen CPU core design, announced in May, is an architecture that the vendor has spent two years developing. It will include simultaneous multi-threading (SMT)—a technology similar to Intel's Hyper-Threading—for improved performance, support for DDR4 memory and the FinFET transistor design. Zen is designed to offer significant performance and power-energy improvements over current APUs, and will appear in an array of chips for everything from high-end PCs to servers, according to company officials.

During a conference call last month to discuss third-quarter financial numbers, AMD President and CEO Lisa Su said that the company had taped out several products using the FinFET technologies, which will help AMD keep on schedule with Zen.

"Zen remains on-track for availability in 2016, and we believe will return AMD to the mainstream server and high-end client market in a significant way in 2017 and beyond," the CEO said."As we stated in the Financial Analyst Day [in May], we had a target of 40 percent IPC [instructions-per-cycle] performance [improvement in] Zen over our previous generation. We believe we're on track for that."