There have been some interesting developments in the Arm-as-a-server processor field, from Cavium’s success to Amazon offering much cheaper instances on its home-brew Arm processors. But now Arm Holdings itself is getting into the fray, and it's offering is a whopper.
Last October, Arm announced the Neoverse platform designed specifically for cloud computing and edge network environments. This week it revealed the Neoverse N1 and E1 platforms, and they are impressive. Usually when Intel and AMD introduce new server chips, they are basically the same chips with faster clocks and more cores. But these two chips are very different in design and meant for different use cases.
“We are moving from an internet model that distributes video (Netflix and YouTube) to one that consumes, manages and processes information produced at the edge,” wrote Drew Henry, head of Arm’s infrastructure business unit, in a blog post.
“When you consider these factors, and the end of Moore’s Law, the days of one-size-fits-all computing are over. We are entering a new era driven by a greater diversity of specialized compute and the need for flexible architectures tailored to meet application specific needs,” he said.
The Neoverse N1 platform — for servers
The Neoverse N1 platform is built on the 7nm “Ares” core, scales up to 128 cores, and delivers a 2.5x the performance on key cloud workloads over prior generations of Arm silicon. Arm also claims a 30 percent improvement in power efficiency over the Cortex-A72. Getting that many cores in a chip will require “chiplets,” or packages with a bunch of cores all tied together by high-speed interconnects. That’s how AMD does it with Epyc and Ryzen, and it is what Intel plans to do with future processors.
The N1 is an Armv8.2-A 64-bit-only CPU with 32KB or 64KB of four-way VIPT L1 instruction cache, 64K to 256KB of private L2 cache, and cryptographic engines for accelerating encryption, decryption and hashing algorithms in hardware. Up to eight E1 CPUs can sit together in a cluster with up to 4MB of L3 cache.
The new N1 platform is the successor to Arm’s 16nm Cosmos platform, which covers the Cortex-A72, A75 and A53 CPU cores. One major licensee of Cosmos is Amazon Web Services (AWS), which based its Graviton processor, announced at AWS re:Invent last November, on Cosmos.
The Neoverse E1 platform — for 5G
The Neoverse E1 platform is designed as a high-efficiency throughput platform, promising a 2.7x throughput improvement in over previous generations. A key differentiator over the N1 is that the E1 is an out-of-order execution CPU with simultaneous multithreading (SMT). That means each core can effectively run two software threads at once, like Intel and AMD both offer.
The E1 also has a Low-latency Accelerator Coherency Port (ACP) for closely coupled accelerator integration. All told, it is meant as a throughput processor rather than data processor like the N1. Arm is targeting 4G/5G transport, software-defined networking (SDN), software-defined storage, and SD-WAN with this processor.
So, N1 will likely be used in servers, while E1 will be used in routers.
The company expects the first silicon to come to market by the end of this year and ramping up into 2020.
Thanks to Andy Patrizio (see source)