Marvell Adds Cores to Structera CXL

When the Compute Express Link (CXL) Consortium introduced the CXL as an interconnect standard in 2019, there were doubts about the need for another option. Today, CXL stands out from a mix of interconnect standards, providing a way to connect memory to various processing elements and share memory resources within a data center. Marvell, with its recently unveiled Structera product line, is using the CXL 2.0 standard.

Marvell’s CXL roadmap was announced in 2022 at the back of its Tanzanite acquisition. Among its first products are near-memory accelerators, called the Structera A, and memory-expansion controllers, the Structera X. The Structera A is the first CXL accelerator to feature server-class processor cores, and it boasts four DDR5 memory channels with capacity for two DIMMs each and 200 GB/s of memory bandwidth. The Structera X, in contrast, doesn’t include cores but can accommodate up to 3 DIMMs of DDR4 memory per channel or 12 DIMMs total. Integrated inline compression adds more capacity.