Intel Announces Compute Express Link (CXL) Interconnect Open Standard

As the datacenter market continues to increase in capability, we need to worry about a lot more than just the processors and memory. Interconnects are an integral ingredient here. As we learned just earlier, NVIDIA acquired networking giant Mellanox to ensure that it has its foot planted as firmly in the market as possible (not that it’s struggling there right now). Another datacenter giant, and also the biggest one, has its own news to share this week.

Introducing: CXL, the Compute Express Link. This is Intel’s solution to the whole interconnect problem, and it’s helping to prove that it cares greatly about it by opening the standard up to the industry. Whether you’re a company that wants to get on board and contribute to the spec, or make use of it, it’s going to be available.

Today, version 1.0 of the CXL specification has been released, and seems to represent Intel’s efforts only. 2.0 is already in the company’s eye sights, as well as a new consortium which will be created later this year. While CXL is obviously very important to Intel, it’s already become likewise important to some other giants, as well. Those include Google, Alibaba, Cisco, Dell EMC, Microsoft, Huawei, HPE, and a little social networking company called Facebook. Now that is some day zero support.

The ultimate goal here is to maximize bandwidth and reduce latency between the CPU and other devices. Those can include such things as GPUs, FPGAs, and networking – all things Intel itself is intimately familiar with. Intel says that CXL differs by offering CPU/device memory coherency, and even an interface that allows plug-and-play functionality.

It’s going to be quite some time before CXL solidifies as a commercial product, but as it stands today, it already has immense support, and it definitely comes at an interesting time with NVIDIA having just acquired Mellanox. Isn’t technology exciting?