MSI AMD EPYC 9005 Series NVIDIA MGX 4U Barebone GPU Server with ConnectX-8

Scale your enterprise AI infrastructure with the MSI S6053G481RAU8, a high-performance 4U barebone GPU server expertly engineered on the highly adaptable NVIDIA MGX modular architecture. Designed to tackle the most demanding data centre workloads, including large language model (LLM) inference, AI fine-tuning, and complex simulations within NVIDIA Omniverse, this powerhouse chassis supports dual AMD EPYC 9005 series processors operating at a TDP of up to 500W. It ensures bottleneck-free data processing with 24 DDR5 RDIMM slots capable of memory speeds up to 6400MT/s across twelve channels per CPU. Built for extreme computational density, the server features nine PCIe 6.0 and 5.0 x16 expansion slots, specifically configured to accommodate up to eight high-end enterprise accelerators such as the NVIDIA H200 NVL or RTX PRO 6000 Blackwell Server Edition GPUs. For ultrafast, carrier-grade network interconnectivity, the system is fully equipped with eight 400G QSFP ports powered by advanced NVIDIA ConnectX-8 chipsets, drastically reducing data transfer latency across multi-node clusters. Storage versatility is maximised through eight front-accessible, hot-swappable U.2 PCIe 5.0 x4 NVMe drive bays alongside dual onboard M.2 NVMe slots for rapid boot configurations. Complete with hardware-level security via an onboard TPM 2.0 module and streamlined remote administration through a dedicated ASPEED AST2600 management port with IPMI 2.0 and DMTF Redfish support, this MSI AMD EPYC server provides the scalable, mission-critical foundation required for tomorrow's advanced computing environments.

For business enquiries, contact our B2B sales team here.

This is a Barebone Server, requires CPU and RAM to function. Sold seperately.

Note: Power cords are not included, sold seperately.

Features

---

S6053G481RAU8 Barebone GPU Server

• 4U (NVIDIA MGX™ architecture)
• Dual AMD EPYC™ 9005 series processors, TDP up to 500W
• (24) DDR5 RDIMM, 12 channels per CPU(1DPC), up to 6400MT/s
• (8) Front hot-swap 2.5" U.2 PCIe 5.0 x4 NVMe drive bays
• (2) 2280/22110 PCIe 3.0 x2 NVMe M.2 ports
• Max (9) PCIe 6.0/5.0 x16 expansion slots, supporting up to (8) NVIDIA H200 NVL or RTX
• PRO 6000 Blackwell Server Edition PCIe GPU cards
• (8) 400G QSFP ports with NVIDIA ConnectX-8 chipsets
• (1) 1000Base-T Dedicated Server Management Port with:
• ASPEED AST2600 with IPMI 2.0 & DMTF Redfish Support
• Dual BIOS & BMC
• TPM2.0 on board
• Hardware Root of Trust available as an optional feature

Application

• Training, Fine Tuning
• LLM Inference
• NVIDIA Omniverse™