Wolley Introduces CXL® Memory Expansion Controller for AI Workstation at FMS2025

As Large-Language Models (LLMs) shifts focus to support larger context windows such as 128 K tokens, the need of higher memory-bandwidth goes beyond the 100 to 200 GB/s ceiling of today's CPU and NPU platforms. GPUs, while popular for AI acceleration, are limited by on-package memory and rising costs, making them an increasingly expensive way to meet these requirements.

"We believe Wolley's client/workstation centric CXL expansion memory in the range of 32GB-64GB is the sweet spot for AI workstation running Generative AI workload", said Bernard Shung, CEO of Wolley.  Wolley's CXL expansion memory is a dual-purpose device offering "memory and storage" interface simultaneous. Both CXL memory expansion and NVMe SSD share the same PCIe ports from the host.  "This is particularly useful for client/workstations where the number of PCIe lanes is limited. We are actively engaging ecosystem partners, CPU, NPU, and memory suppliers, to validate and promote this CXL memory expansion module for AI workstation.", said Bernard Shung.

Wolley Inc., a controller ASIC company with headquarters in San Jose, CA and an R&D center in Taiwan. The company develops CXL® storage solutions and provides CXL® controller IP licensing to datacenter and edge AI/ML applications. Wolley also provides end-to-end SoC design and integration services for CXL® applications. Wolley has demonstrated its CXL® 3.0 controller core IP in FMS24, CXL® 2.0 controller in a 4-port fabric switch using FPGA in FMS23, and SCM controller using FPGA in FMS19.

For more information, email to [email protected] or visit www.wolleytech.com

SOURCE Wolley Inc.