Revolutionizing AI and HPC: The CXL Memory Expansion Server Set to Debut at Computex 2024
At Computex 2024, MSI rolled out a lineup of servers powered by the 4th Gen AMD EPYC processors, complete with CXL (Compute Express Link) technology and DC-MHS architecture, aimed at enhancing scalable cloud applications.
S2301 CXL Memory Expansion Server
Designed for data-heavy tasks and in-memory databases, the S2301 is a 2U server equipped with dual-socket AMD EPYC CPUs. It adheres to CXL 2.0 standards and offers 24 DDR5 DIMM slots alongside two PCIe 5.0 x16 slots. For storage, it includes eight E3.S 2T drive bays for CXL memory and eight E3.S 1T bays for NVMe storage, allowing for memory expansion up to 8-TB.
G4101 GPU-Accelerated Rack-Mounted Workstation
MSI also unveiled the 4U G4101 workstation, optimized for AI tasks. This unit features a single AMD EPYC 9004 processor with liquid cooling, 12 DDR5 RDIMM slots, four PCIe 5.0 x16 slots which accommodate triple-slot GPUs, and 12 2.5-inch U.2 NVMe/SATA drive bays. Its design ensures efficient thermal management through airflow spacing and a liquid closed-loop cooling system.
CX271-S4056 All-Flash Server
The CX271-S4056 is a 2U all-flash server featuring a single-socket AMD EPYC 9004 processor. It includes two double-wide slots for GPU cards, two OCP NIC 3.0 mezzanine slots, and up to 24 2.5-inch PCIe 5.0 U.2 NVMe drive bays, making it well-suited for AI inferencing applications.
Multi-Node Servers for Data Centers
MSI introduced two multi-node servers: the CD360-S4051-X2 and CD260-S4051-X4.
CD360-S4051-X2
This 3U, two-node server is designed for high TDP server processors. Each node houses one AMD EPYC processor, 12 DDR5 RDIMM slots, two optional double-wide PCIe 5.0 x16 slots for GPU cards, one OCP NIC 3.0 mezzanine slot, two NVMe M.2 ports, and three 2.5-inch U.2 NVMe drive bays.
CD260-S4051-X4
The CD260-S4051-X4 is a 2U, four-node server which shares the same specifications as the CD360 but lacks the PCIe 5.0 slots.
These new offerings from MSI aim to address a wide range of applications from data-intensive workloads to GPU-heavy AI processing, providing robust and scalable solutions for modern data centers and enterprise needs.
