SK Hynix 96GB DDR5-6400 RDIMM PC5-51200 2Rx4 ECC Registered

The SK Hynix 96GB DDR5-6400 RDIMM (HMCG94AGBRA123N) is a high-capacity, high-bandwidth registered memory module built on SK Hynix's advanced DDR5 DRAM architecture. Operating at 6400 MT/s with a PC5-51200 designation, it delivers a substantial leap in memory bandwidth over prior-generation DDR4 platforms, making it well-suited for compute-dense server environments. The 2Rx4 (dual rank, x4 organization) configuration maximizes capacity and signal integrity across the memory bus, while the registered buffer architecture ensures stable operation in multi-DIMM server configurations by reducing electrical load on the memory controller.

This module incorporates on-die ECC (Error Correction Code) inherent to DDR5, supplemented by the registered DIMM's own ECC logic, providing robust data integrity protection essential for mission-critical enterprise workloads. The 96GB capacity per DIMM enables extremely high per-socket memory populations when deployed across the full DIMM slot complement of platforms such as Intel Xeon Scalable (Granite Rapids) and AMD EPYC (Turin) processors, both of which natively support DDR5-6400 speeds and benefit directly from the bandwidth this module provides. The module operates at the DDR5-standard 1.1V supply voltage, contributing to improved power efficiency relative to equivalent DDR4 capacity points.

Designed for next-generation two-socket and single-socket server platforms, the SK Hynix HMCG94AGBRA123N addresses the growing memory bandwidth and capacity demands of in-memory databases, large language model inference and training, high-performance computing, and virtualization consolidation workloads. Its JEDEC-compliant design ensures broad platform compatibility within the DDR5 RDIMM ecosystem, and the ECC registered architecture meets the reliability, availability, and serviceability (RAS) requirements of enterprise and hyperscale data center deployments.

Ideal for

• In-memory database acceleration (SAP HANA, Oracle Database In-Memory) requiring large contiguous memory footprints and high sustained bandwidth per socket
• Large language model (LLM) inference serving, where high per-socket memory capacity reduces model sharding requirements and high bandwidth accelerates tensor operations
• High-performance computing (HPC) workloads such as computational fluid dynamics, molecular dynamics simulation, and seismic processing that are latency- and bandwidth-sensitive
• Dense virtualization and cloud host infrastructure, maximizing VM or container density per server node through high per-DIMM capacity
• Real-time analytics and streaming data processing platforms (Apache Spark, Flink) that benefit from large working-set memory and high read bandwidth
• AI/ML training cluster CPU nodes where large dataset pre-processing and data pipeline staging demand both capacity and sustained memory throughput

Technical specifications

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