Samsung 128GB DDR5-5600 RDIMM 4Rx4 ECC

The Samsung 128GB DDR5-5600 RDIMM (M321RAGA0B40-CWM) is a high-density registered dual in-line memory module engineered for the demands of modern enterprise server platforms. Built on Samsung's advanced DDR5 architecture, this module operates at a rated speed of 5600 MT/s with a quad-rank, 4Rx4 configuration across a 288-pin interface, delivering the raw memory capacity required to support large in-memory datasets. On-die ECC (ODECC) is supplemented by the module-level ECC that is standard to RDIMM designs, providing multiple layers of data integrity protection suited to mission-critical deployments.

As a Registered DIMM, the M321RAGA0B40-CWM incorporates a register (RCD) that buffers command and address signals, enabling higher rank counts and greater per-channel capacities than unbuffered alternatives allow. Operating at 1.1V, the module balances power efficiency with the throughput headroom that memory-bound workloads require. The 4Rx4 topology utilises x4 DRAM devices across four ranks, maximising density per module and making it particularly well suited to platforms where slot counts are limited relative to total memory requirements.

Launched in Q1 2024, this module targets enterprise environments running next-generation Intel Xeon Scalable (Sapphire Rapids and successors) and AMD EPYC (Genoa and successors) server platforms that support DDR5 RDIMM. Its 128GB capacity per slot allows system architects to populate high core-count dual-socket servers with multi-terabyte memory configurations, directly addressing the working-set demands of in-memory databases, large-scale virtualisation, and AI inference serving. Lead times for this high-density module should be factored into procurement planning.

Ideal for

• In-memory database deployments (SAP HANA, Redis, VoltDB) requiring multi-terabyte working sets within a constrained slot footprint
• Large-scale virtualisation hosts running hundreds of concurrent VMs where per-VM memory reservations demand maximum physical capacity per server
• High-core-count AI and machine learning inference servers where model parameters and activation data must reside entirely in DRAM for low-latency response
• ERP and business intelligence platforms with memory-bound analytical queries that benefit from expanded buffer pool and in-memory caching
• High-performance computing (HPC) nodes executing memory-intensive simulation, genomics, or computational fluid dynamics workloads
• Enterprise Kubernetes and container orchestration infrastructure requiring dense memory allocation across a reduced physical server count

Technical specifications

Available from Omnixon Global. Submit an RFQ and our team will confirm configuration and availability for your order.