Applied Micro is to use the ARM v8 CPU architecture it developed for data centers to keep hold of embedded-systems customers who have decided to move away from its existing MIPS and PowerPC-based processors. In a panel at ARM TechCon, customers said the broad base of ARM developers and the promise of lower cost through competition convinced them to migrate from the PowerPC architecture.
Paramesh Gopi, Applied Micro CEO, claimed: “We are now in the process of shipping a derivative based on the same server class technology [as X-Gene] and bringing server-class computing, I/O security, and reliability to the embedded space. Helix is sampling today and will ship [for production] in 2015.
“Since 2004, the company was involved in shipping IBM PowerPC-based embedded products. Apart from Freescale, Applied was the biggest company shipping PowerPC. We had a smorgasbord of architectures: MIPS and PowerPC,” Gopi said, adding that they are dying out. “Given the large developer base it’s clear that ARM is going to be the de facto architecture in the embedded space. What we are essentially doing is taking our PowerPC heritage and using that as our growth plank, to change our embedded business base to be an ARM-based business. This is moving ARM into the key markets that are already being serviced by us.”
Gopi said the main target markets for the Helix multicore processors lie in networking, document imaging, storage, and robotics for industrial control. Although initial versions of the Helix will have a power consumption of up to 45W, later dual-core versions based on the 28nm X-Gene should hit 16W or less. “We don’t need to get to finFET to give you a fabless product,” Gopi claimed.
Pradeep Kathail, chief software architect at Cisco, said moving to ARM gives the company a wider choice of vendors “and we can reduce the number of different architectures we are using”.
Kathail added: “There is now very little going on in PowerPC. And with MIPS there is less going on. But we need 64bit processing: routing table sizes are becoming huge. For us, embedded boxes are supercomputers: we need 64bit processing for those.”
Keith Shea, head of alliances and business development at Wind River, said: “64bit is becoming more important because compute workloads being pushed out to the edge [of networks] are becoming more intensive. Why ARM? It’s the ecosystem. As we see that ecosystem grow, that migration [to ARM] is going to be exponential.”
Jeff Capone, CTO at Netgear, said a shift to ARM as standard architecture will help its storage products business: “For storage there are two important things: I/O and cost. Even on lower-end products we need to introduce support for 10Gbit/s ethernet. And to deliver the ability to handle the processing at line rates, that’s where [ARM] v8 plays for us. Also, internally, the important thing for us is to be able to use the same code base [across products].”
Wider talent pool
Gerald Kleyn, director of server R&D at HP, added: “With ARM we have a great ecosystem of software developers. There is a great talent pool out there.”
Although ARM provides a common software platform across a variety of SoCs, Kathail said the increasing requirement for coprocessors and hardware accelerators will drive the need for common APIs there. “From a networking perspective, the SoC approach is very important. We need features such as cryptography and the ability to distribute packet handling across multiple cores and use offload engines. ARM provides for us a single architecture, but we need to do the same thing around these acceleration features.”
Kathail said in the world of software defined networking (SDN), standards groups such as OpenDataPlane are helping. “They are trying to define APIs so the software looks the same [across different SoC implementations]. We want to make sure we have this common API: software that makes all the chips look the same.”