OpenCAPI standard will be incorporated into the CXL design specification

by ·

The OpenCAPI standard, which was supported by AMD, Xilinx, and IBM in the past, announced earlier that it will merge with the current mainstream CXL (Compute Express Link) standard. In the future, all OpenCAPI standards and assets will be merged into CXL (Compute Express Link) within the standard specification.

The OpenCAPI standard is derived from IBM’s existing accelerator component and CPU communication protocol CAPI. As the overall usage rate is not high, AMD, which originally supported this standard, turned to its own Infinity Fabric protocol to allow the CPU to communicate directly with the GPU used as an accelerator. At the same time, Xilinx is also currently incorporated into the AMD system, so the OpenCAPI Alliance announced that it will incorporate its standard-setting and other assets into the CXL protocol.

At present, the CXL protocol has been supported by many chip manufacturers, including Intel, NVIDIA, AMD, Micron, Samsung, Arm, IBM, Google, Huawei, Cisco, etc. All support this protocol, and it is applied to its chips, memories, accelerators, and other products design. In this way, the communication efficiency between the CPU, accelerator components, and memory modules can be faster, thereby corresponding to faster and larger computing efficiency.
After the OpenCAPI-related standard resources are incorporated into the CXL protocol, it means that the OpenCAPI alliance will no longer exist and develop in the future, and related technologies will continue to be applied to more chip designs through the CXL standard specification.

Tags: