AMD to Begin Sampling 4 and 8-Core ARM-based ‘Seattle’ Processors in Q2
Posted on January 29, 2014 1:09 PM by Rob Williams
At the Open Compute Summit, held in San Jose, AMD’s General Manager of its Server Business Unit Andrew Feldman took the veil off of the company’s upcoming Opteron A1100 series processors – one of the most important server announcements to come out of AMD in perhaps the past five years.
The biggest thing to set the Opteron A1100 series apart from the others is that they’re based on ARM’s ARMv8 architecture, rather than the well-established x86 architecture. A couple of years ago, even thinking of using an ARM CPU in a server might have had you sent to a doctor for evaluation, but this is 2014, and ARM’s current flagship – believe it or not – is capable of handling typical server workloads.
The exact core to be used in AMD’s A1100 is ARM’s Cortex A57, which sports a 64-bit design. With AMD’s magic sprinkled on top, the A1100 will include a server-class memory controller, 8x PCIe Gen 3 lanes, dual 10Gbps Ethernet ports, and 8x SATA 6Gbps ports. A1100 will be available in both 4 and 8-core designs.
Thanks both to the fact that the A1100 will have a 64-bit heart and AMD’s memory controller will be implemented, the processors will be able to support up to 128GB of DDR3 or DDR4 at speeds of 1866MHz. Regular unbuffered DIMMs can be used, as well as registered (ECC) and even SODIMMs. Overall, the A1100 is going to be very flexible where memory is concerned.
A quick rundown (and recap) of key features:
4 or 8-core ARM Cortex-A57
Up to 4MB of shared L2, and 8MB of shared L3 cache
DDR3 or DDR4 support for up to 1866MHz speeds
8x PCIe Gen 3 lanes
8x SATA 6Gbps ports
2x 10Gbps Ethernet ports
ARM’s TrustZone technology for enhanced security
Crypto and data compression co-processors
If it seems a little odd that a CPU not too dissimilar from those in our smartphones could be used to power a server, consider the fact that AMD’s A1100 will be like an advanced smartphone (or tablet) SoC but tuned for server use and with a bunch of extra features piled on. The reason ARM as a base is useful for such a purpose is because of its low power requirements. With the A1100, AMD is aiming to deliver the best of both worlds here, blending both performance and power efficiency in an attractive overall package.
According to AMD’s press deck, it’s estimated that 25% of the server market in 2019 will be based around ARM’s processors. If true, then that’s an amazing thing for ARM, and a clear reason why AMD wants to make a big impression now.
In addition to the hardware itself, AMD stresses that its commitment to the software side of things is also strong. It’s a gold member of the Linux Foundation, for starters, and is also a member of Linaro, a significant kernel contributor. Further, key compilers have already been ported to 64-bit ARM, so developers won’t have to look too hard to begin supporting the platform on day one.
Not surprisingly, AMD is anxious to get its A1100s out the door, and it’s anticipating that samples will begin shipping in just a couple of weeks, with the final launch to occur in Q2.