CMM-H (CXL Memory Module, H: Hybrid)

Industry 1st CXL-based Storage
Optimized for AI/ML

SSD that combines the benefit of
Storage and DRAM memory

Leveraging Compute Express Link(CXL) interconnect technology & offering  CMM-H PMEM(Persistent Memory) option, CMM-H helps us to achieve high bandwidth read/write and low latency. CMM-H PMEM (Persistent Memory) option supports existing software written for NVDIMM (Non-volatile Dual In-line Memory Module) with built-in DRAM and NAND flash.  

CMM-H Architecture

Dual Mode Support

Persistent Memory Mode

20X improvement in speed and latency

 When used in AI & ML applications, CMM-H can have up to 20X* performance improvement by increasing a random read speed and decreasing a latency. This can be achieve by implementation of  Samsung’s industry first Compute Express Link (CXL) interconnection technology & built-in DRAM cache technology. 

 * 20x better performance compared to using PCIe Gen4 NVMe SSD. MS-SSD

Ideal memory solution for AI & ML Workload

Efficient AI recommendation System

* Results based on publicly available DLRM workload traces from Meta and FPGA based PoC CMM-H

**PCIe Gen4 NVMe SSD (Random 4KB read(0.9MIOPS)  for 100% hit rate vs MS-SSD 64B read Performance (42.9MIOPS)

Webinar: CMM-H Samsung's CXL-based SSD for the Memory-Centric Computing Era

Better System TCO*

Small granularity data access, Large memory capacity

*Total Cost of Ownership

Persistent Memory

Battery-backed DRAM with Speed. Supports flush-on fail with CXL 2.0 GPF Feature

Would like to get connected for this project?

Memory Solutions Lab

Transforming Memory Performance

Go back to Main Page

Contact Form

Do you have question to Samsung MSL?

Once the form is submitted, MSL will contact you with details