AMD’s ‘Seattle’ ARM Server Chips Set To Ship Next Year

AMD is pinning its hopes for competing with Intel in the server market on the expected growth in popularity of server chips based on designs from British company ARM, the chip maker said on Monday.

The company’s first ARM-based server chip, code-named “Seattle”, will begin shipping in volume in the second half of 2014, marking a departure for the chip maker, which has used the same x86 instructions as Intel since the 1980s.

Roadmap

AMD also announced two x86-based chips: “Berlin”, which will be available both as a conventional CPU and as an “accelerated processing unit” (APU), AMD’s term for a chip integrating a graphics coprocessor, and “Warsaw”, which will compete with Intel’s Xeon in the high-performance computing (HPC) market.

The chip maker’s server roadmap is intended to make headway against Intel, which benefits from the vast majority of server sales – 95.7 percent of all server processors as of the end of 2011, compared to AMD’s 4.3 percent, according to Mercury Research.

For this, AMD is relying on ARM-based chips to begin to take significant server market share starting next year. AMD expects a large data centre to adopt ARM in 2014, leading to significant growth in 2015 and “double-digit” server market share in 2016 and 2017, according to Andrew Feldman, vice president of AMD’s server business unit.

“Our strategy is to differentiate ourselves by using our unique IP to build server processors that are particularly well matched to a target workload and thereby drive down the total cost of owning servers,” Feldman stated. “This strategy unfolds across both the enterprise and data centers and includes leveraging our graphics processing capabilities and embracing both x86 and ARM instruction sets.”

Energy efficiency

The appeal of ARM’s technology, which already powers most smartphones and tablets, is largely down to its energy efficiency, which could dramatically reduce costs associated with server power consumption – particularly in the large server farms operated by companies such as Google, Facebook or Amazon.

Chip makers including Nvidia and Applied Micro have already begun making ARM-based chips for the server market, and AMD said it expects Samsung and Qualcomm to join in. However, AMD argued none of these rivals has its experience in making enterprise-grade server chips.

AMD’s “Seattle”, based on the ARM Cortex A57 core, will be available in 8- and 16-core versions and is expected to run at 2 GHz or greater, offering two to four times the performance of the recently announced “Kyoto” Opteron X-Series processor, with significant improvements in compute-per-watt.

It will offer 128GB DRAM support, offload engines for power efficiency, dedicated encryption and compression blocks and integrated 10GB Ethernet. The chip will be AMD’s first to integrate “Freedom Fabric”, a high-speed networking technology for dense compute systems acquired from Sea Micro.

x86 chips

Also announced on Monday was “Berlin”, an x86 processor which will include four “Steamroller” cores and offer nearly eight times the gigaflops per-watt compared to the current AMD Opteron 6386SE chip. It will be based on AMD’s Heterogeneous System Architecture, which enables uniform memory access for the CPU and GPU – a technology intended to make the GPU co-processor easier to program. “Berlin” is expected to be available in the first half of 2014.

Finally, “Warsaw”, also x86-based, is an enterprise server chip designed for high performance in two- and four-socket servers. It will improve performance-per-watt over existing AMD Opteron 6300 server chips, driving down costs while offering socket compatibility. AMD expects to ship the processor in the first quarter of 2014.

Boston Limited, said to be the first company to make ARM-based servers, demonstrated its products in London last October; Dell has also begun making ARM servers aimed at “big data” analytics.

Do you know all about UK tech leader ARM Holdings? Take our quiz!