Hardware

ARM Joining CXL Consortium Could Allow Hardware Developers To Build Better CPUs And Accelerators With Higher Memory Semantics

ARM, the company on which several CPU and System on a Chip (SoC) makers rely heavily, has officially decided to join the Compute Express Link (CXL) Consortium. This has been a highly anticipated move that would significantly help computing hardware designers and developers build better CPUs and Accelerators. Close collaboration and official support of ARM towards the CXL participating companies should spur the development of processors that are even closely linked with other components in the SoC, and achieve significantly improved memory semantics between processors and accelerators.

ARM, a highly reputed British multinational semiconductor and software design company, has officially joined the CXL Consortium. This should enable hardware makers or clients of the company to implement the new CPU-to-Device interconnect standards, and better contribute to the specification. It is not immediately clear why ARM had chosen to stay away from official participation for so long, especially when the company has been participating in CXL workgroups. Moreover, ARM has been providing technological and promotional resources to support the development of the technology that CXL Consortium actively supports.

ARM’s Official Participation in the CXL Consortium To Achieve A Big Leap Computing Efficiency Of CPUs and Electronic Devices:

ARM’s Intellectual Property (IP), in addition to the chips that it actively manufactures, is widely used in many electronics, mobile computers, smartphones, wired and wireless networking equipment, and many more. Needless to add, chipsets, as well as the technological developments made by ARM, have made the company the preferred choice for many companies. Still, ARM was among a few major technology companies that were yet to join the CXL consortium.

Interestingly, ARM and CXL have a long history of active collaboration. The company actively participates in the workgroups organized by the CXL. Moreover, ARM has been one of the most active companies that have regularly lent technological and promotional resources to support the development of technology.

As part of the official participation, the company is expected to commit to providing its customers with a full software framework. Since the official participation has just been signed, it is quite unlikely that ARM would provide or add appropriate logic to its upcoming AMBA PCIe Gen 5 PHY implementations.

ARM is a board member of the PCI SIG and the Gen-Z Consortium. Moreover, the company supports its own in-house developed CCIX interface that works for an optimized inter-package chip-to-chip interface. The company claims CCIX supports full cache coherency. ARM has assured that CCIX  will be used as an inter-package chip-to-chip interface for heterogeneous system-on-packages

With the CXL participation, ARM should enable its myriad clients to design, develop and build CPUs or accelerators that support low-latency cache coherency. Additionally, end-products that have the ARM chips embedded within them, should benefit from the memory semantics between processors and accelerators.

It might seem that ARM’s CCIX and CXL could be counterproductive or even competing. However, that’s not the case. CXL’s implementation doesn’t have the exact same functionality as that of CCIX. Simply put, it is out of the scope of CXL at present. Hence the CXL’s implementation of the chip-to-chip interface should not clash with Arm’s version of CCIX.

According to reports, it appears NVIDIA too, is a formal member of the CXL Consortium. Moreover, it joined through the parent company and not just via Mellanox. With ARM’s official participation, the High-Performance Computing or the HPC Industry is sure to benefit in the long run, claim analysts.


Leave a Reply

Your email address will not be published.

Close