“Our first partners are starting to get their first results,” Pete Hutton executive v-p and president of ARM’s product group told Electronics Weekly.
Hutton sees the server market as a 50 million unit a year opportunity and ARM aspires to a 5-10% market share by 2017.
This year it sees itself gaining a ‘single digit percentage’ market share.
Asked if the collapse of Calxeda has had any effect on ARM’s momentum in servers, Hutton replied: “It’s unfortunate for them. We’ve hired a few of their guys. They tried to go with 32-bit but the market really wants 64-bit and those guys are getting traction.”
A major plus for ARM’s move into servers is the launch of AMD’s ARM-based server offerings. “Their experience in this space is a real advantage,” said Hutton.
However the biggest advantage ARM has in the server space is the opportunity it gives customers to customise their ARM-based SoCs for servers.
Intel just provides a ready-made chip and peripherals and the customer can take it or leave it – but if an ARM core is used a customer can tailor the server SoC precisely to his requirement.
With the big datacentre operators – Microsoft, Amazon, Facebook, Google et al – increasingly drawing up their own server specs which they then bung over to the Taiwan OEMs for implementation, the customisation opportunity is becoming increasingly valuable.
Asked what he thought of the characterisation that the heavy-lifting in servers needs to be done by Intel processors while lighter tasks can be left to ARM processors, Hutton replied:
“I’d say that’s an Intel characterisation. ARM-based solutions can use multiple high performance cores for heavy-lifting.”
ARM partner Applied Micro has a server chip which uses eight 64-bit ARM cores each running at 3GHz.
Finfet processes will add quite a bit to the performance as ARM has found from running its processors on TSMC’s 16nm finfet process. When will we see 16nm ARM-based SoCs on the market.
“We’re squinting at this year,” said Hutton.