In 2024, Microsoft will use an in-house successor to its Ampere Altra Arm Neoverse N1 chips. The Microsoft Azure Cobalt 100 is a 128-core Arm Neoverse N2 design that’s utilizing Arm’s accelerated path to chip design to create a lower-power chip.
Microsoft Azure Cobalt 100 128 Core Arm Neoverse N2 CPU Launched
A couple of weeks in the past we coated the Arm CSS N2 announcement. The essential concept is that Arm is offering not simply the Core and material IP, but additionally ready-built IP blocks of many cores. The important thing constructing block Arm confirmed throughout that announcement was a 64-core IP block organized in twin core compute tiles.

It then confirmed the potential to hyperlink a number of 64-core CSS N2 blocks collectively utilizing chiplet interconnects.

Arm additionally confirmed a 128-core theoretical gadget in every socket of a dual-socket server.

Its proof level Arm additionally mentioned that it helped a hyper-scale buyer go from venture begin to silicon in 13 months. Our greatest guess is that that is Microsoft the place a number of ex-Arm of us now work.

Deployment is seemingly taking place in the present day, nevertheless it looks as if GA goes to be in 2024. It’s fairly fascinating that Microsoft is utilizing the 2021-era Arm Neoverse N2 for this. We first began seeing merchandise just like the Marvell Octeon 10 Arm Neoverse N2 DPU that rivaled 2017-era Intel Xeons. A 12 months later we get a Microsoft announcement, and a 12 months after that we must always get GA of elements.
Last Phrases
The Microsoft Cobalt 100 is admittedly fascinating. From what we have now seen from the Arm Neoverse N2, it’s roughly like a Skylake-era Xeon when it comes to Integer efficiency at a decrease energy level and a considerably greater density. Microsoft can even get new options like DDR5 reminiscence. From a efficiency perspective, AMD’s 2023 Bergamo will probably be considerably quicker than Microsoft’s 2024 Cobalt 100, however our large query has been whether or not Bergamo is simply too quick for cloud suppliers on the lookout for low-cost and lower-power cores.
It is usually fascinating that Microsoft goes with a {custom} N2 design because it additionally has AmpereOne utilizing custom-designed Arm cores. That begs the query is Microsoft merely chopping out the middle-man or as a substitute planning to supply Arm at a number of efficiency and functionality factors? An Arm Neoverse N2 design in 2024 just isn’t even going to be the best efficiency per core cloud Arm CPU when it’s accessible, and perhaps not even within the high half.
For just a few years now at STH, we have now been speaking about simply how large the cloud-native processor will probably be. Microsoft apparently thinks this phase will probably be large enough that it warrants constructing its personal CPUs.