Skip to content

Search the site


Arm beefs up Neoverse N2 with more complete CSS platform

Arm is adding additional layers to its CPU designs with CSS, a platform that will ease the burden on chipmakers to handle external functions

UK chip designer Arm has unveiled what it says is a more complete and capable addition to its Neoverse N2 CPU line.

The company this week unveiled the Neoverse N2 CSS, a chip design that expands on the Neoverse platform by adding additional functions and capabilities. Whereas the Neoverse N2 would only include a license for the CPU itself, N2 CSS also includes the higher-level functions such as memory I/O and acceleration.

The N2 CSS will support designs for 24 to 62 cores and will run anywhere from 2.1-3.6ghz clock speed. The design supports both DDR5 and PCIe Gen5. Arm claims that in its high-density form, the N2CSS will allow for 256 cores to packed into a 2-socket server design.

For those unfamiliar, Arm does not actually fabricate or sell any of its own hardware. Rather, the company designs chips and then licenses those designs to vendors such as Broadcom and Qualcomm, who sell the actual silicon- traditionally for mobile or embedded systems, but increasingly for cloud datacenter and AI systems as well.

Steve Demski, Arm senior marketing manager for hyperscale and HPC, explained that the Neoverse N2 CSS is being rolled out in response to feedback from customers who wanted to obtain more than just a bare-bones CPU design.

"In our time engaging with Cloud-to-Edge customers, and more recently with silicon development teams targeting AI and ML solutions, we’ve heard a consistent theme," Demski explained.

"Customers love the scalable efficiency Arm Neoverse platforms deliver, but they don’t want to reinvent the wheel – IP selection, system configuration, floorplanning, verification, validation, 3rd party IP and fab integration – that comes with building a CPU compute subsystem."

Arm has also been pitching the Neoverse CSS as a way for hardware vendors to cut down on their time to delivery and save money on engineering costs, claiming one customer was able to go from drawing board to running operational hardware in 13 months, while others have saved decades worth of engineering time.

Arm dropped news of the new platform at Stanford University's Hot Chips conference, the same show where Intel unveiled its own plans for the next generation of Xeon server chips.

Though x86 and Arm hardware have traditionally run in very different use cases, the two architectures have begun to see some overlap in emerging fields such as cloud and machine learning.