As Internet archives fill up with CXL tutorials, vaunted industry experts take turns parsing the technology to the bone, in their effort to find answers to some of the most pressing public questions. For example, we know that CXL enables high composability at datacenters by unlocking massive memory for surprisingly low cost – and that is a golden ticket for organizations handling memory-intensive workloads – but what about latency?
Historically, we have seen that putting memory away from CPU does not help lower latency. So what is the memory latency like with CXL? Is it as bad as we have seen in the past, or is it any better?
In his article “Just How Bad IS CXL Memory Latency”, author Tobias Mann compares CXL’s latency with other architectures, and delivers a verdict. He writes,
Conventional wisdom says that trying to attach system memory to the PCI-Express bus is a bad idea if you care at all about latency. The further the memory is from the CPU, the higher the latency gets, which is why memory DIMMs are usually crammed as close to the socket as possible.
“JUST HOW BAD IS CXL MEMORY LATENCY” is published on The Next Platform where Mann is a regular contributor. Give the article a read to learn about the latency numbers of CXL when paired with the most powerful processors in the market. To take a deep dive into CXL, give a listen to the Gestalt IT podcast series on CXL, Utilizing CXL.
