CXL Overcomes Hierarchical Routing Limits
By Gary Hilson, EETimes (January 9, 2024)
The latest incremental update to the Compute Express Link (CXL) protocol aims to improve disaggregation and keep pace with high-performance computational workloads.
The feature updates in CXL 3.1 were already in play when the standard was last updated, but members of the CXL Consortium needed a little more time to make sure they could work across different architectures and further finesse security features that were introduced in 2.0, Kurtis Bowman, the Consortium’s working group co-chair, told EE Times.
Moving to a fabric is not a small task, he added. “That took some time to get worked out.”
With hierarchical-based routing, memory devices at the bottom must go up to the host if they want to communicate with each other, even if they sit right beside each other. Bowman said port-based routing (PBR) in CXL 3.1 allows for much more efficient switching. Reducing the number of hops necessary by staying in the memory domain dramatically reduces latency, he said, and it also enables more complex topologies.
E-mail This Article | Printer-Friendly Page |
Related News
- CXL Consortium Announces Compute Express Link 3.2 Specification Release
- CXL Fabless Startup Panmnesia Secures Over $60M in Series A Funding, Aiming to Lead the CXL Switch Silicon Chip and CXL IP
- Alphawave Semi Partners with PCISig, CXL Consortium, UCIe Consortium, Samtec and Lessengers to Showcase Advances in AI Connectivity at Supercomputing 2024
- JEDEC® Adds to Suite of Standards Supporting Compute Express Link® (CXL®) Memory Technology with Publication of Two New Documents
- CXL Fulfills AI's Need for Open Industry Standard Interconnect
Breaking News
- QuickLogic Announces $6.575 Million Contract Award for its Strategic Radiation Hardened Program
- Micon Global and Silvaco Announce New Partnership
- Arm loses out in Qualcomm court case, wants a re-trial
- Jury is out in the Arm vs Qualcomm trial
- Ceva Seeks To Exploit Synergies in Portfolio with Nano NPU