NUREMBERG, Germany – Freescale Semiconductor Inc. could use its recently announced Cortex-A50 series processor license from ARM to extend its digital networking activity into server computing, CEO Gregg Lowe told EE Times at the Embedded World exhibition here.
When asked in an interview by EE Times whether Freescale could use the A50 series 64-bit ARM subscription license to expand into the server market Lowe said: "Yes, I think so." Lowe added: "There is a lot happening in that space."
The Cortex-A50 series, currently being developed by ARM and partners, includes the Cortex-A57 and Cortex-A53 as a big-little combination. Freescale announced that it had licensed the A50-series of microprocessors on a multiyear subscription basis for future versions of its i.MX applications processor and QorIQ communications processor product lines.
Freescale also has provides processors based on the PowerPC architecture licensed from IBM.
The server market has begun to fragment in recent years with the emergence of markets for microservers and more intelligence being pushed towards the network. At the same time there is a focus on energy efficiency in large data centers that expected to perform "cloud" services for consumers who are online for most of their professional and private lives.
Gregg Lowe, CEO of Freescale (left) and Steve Wainwright, general manager EMEA for Freescale, on the booth at Embedded World.
Good Question! To my knowledge there about 10 known companies working on ARM based Micro servers, I guess most likely performance per watt will be very close among all, it will be a tough market unless there is a key feature differentiation!
I'm happy to see Freescale adopting the ARM ISA. One of the best outcomes would be to see Freescale merging some of their more valuable "legacy" IP with the newer architectures. IMHO one of their best features was the Time Processor Unit (TPU) which was available on the ColdFire processors and was also brought into the PowerPC uCs. If you've ever tried to write and support a complex application using the generic ARM timer you'd understand how much code development it really takes. If Freescale decides to start designing ARM uCs (and especially in it intends to de-emphasize PowerPC for these applications) it needs to seriously consider integrating TPUs into such an architecture, both for new designs and to support legacy code bases to give them a reasonable "upgrade" path.
Every ARM 50 licensee would have their eyes on this market. Other than AMD all the existing licensees have been embedded players so they are most likely intent on retaining or expanding their market share there. The server market is a huge lucrative space that they can target if that does take off so it appears to be a low risk approach.
The question on differentiation remains. In the embedded market that is in the SoC with the special accelerators etc. In the server space it is primarily about the CPU so while there will be slightly different variations to memory interfaces etc. they will most likely look alike. Which means the only differentiation will be on price - which makes the business folks at Dell/HP salivate at higher margins....
This doesn't really come as a surprise. We can't know whether Freescale will succeed in the ARM server market, but they can't realistically stay *out* of it. There are too many markets Freescale would like to be in where ARM is the architecture of choice, and if they want a piece of that market they offer ARM cores.
David Patterson, known for his pioneering research that led to RAID, clusters and more, is part of a team at UC Berkeley that recently made its RISC-V processor architecture an open source hardware offering. We talk with Patterson and one of his colleagues behind the effort about the opportunities they see, what new kinds of designs they hope to enable and what it means for today’s commercial processor giants such as Intel, ARM and Imagination Technologies.