Supermicro 8U AMD EPYC MI300X GPU Server

• AI and deep-learning model training with multi-GPU accelerator infrastructure
• High-performance computing for scientific, engineering and research workloads
• Enterprise AI platforms for large-scale machine learning and analytics
• Industrial automation using accelerated compute for complex workloads
• Retail analytics and computer-vision processing
• Climate and weather modeling requiring sustained HPC resources
• Large-scale simulations and computational research
• Private-cloud AI infrastructure for controlled enterprise workloads
• Hybrid-cloud environments requiring dedicated on-premises accelerator capacity
• Scale-out GPU clusters using high-speed networking and direct RDMA connectivity

• 8U high-density rackmount GPU architecture for enterprise AI and HPC
• Dual AMD EPYC 9004/9005 Series processor support
• Up to eight onboard AMD Instinct MI300X OAM accelerators
• PCIe 5.0 x16 CPU-to-GPU interconnect
• AMD Infinity Fabric Link for GPU-to-GPU communication
• Eight PCIe 5.0 x16 low-profile expansion slots
• Up to four PCIe 5.0 x16 full-height half-length expansion slots with optional configuration
• 8x NIC architecture for GPU-direct RDMA with 1:1 GPU ratio
• 24 DDR5 ECC RDIMM slots
• Up to 6TB DDR5 memory depending on processor generation
• Up to 18 front hot-swap 2.5-inch NVMe/SATA/SAS drive bays
• Two onboard M.2 NVMe slots for boot storage
• Six 3000W redundant Titanium-level power supplies in 3+3 configuration
• Ten heavy-duty fans with optimal fan-speed control
• TPM 2.0, Secure Boot and Silicon Root of Trust security capabilities
• Technical Specifications

• 
  
• Brand: Supermicro
• Part Number: AS-8125GS-TNMR2
• Product Family: A+ Server
• System Type: GPU Server
• Form Factor: 8U Rackmount
• Motherboard: Super H13DSG-OM
• Processor: Dual AMD EPYC 9004/9005 Series
• Processor Socket: Socket SP5
• Maximum GPU Count: 8
• Supported GPU: AMD Instinct MI300X OAM Accelerator
• CPU-GPU Interconnect: PCIe 5.0 x16
• GPU-GPU Interconnect: AMD Infinity Fabric Link
• Memory Slots: 24 DIMM Slots
• Memory Type: ECC DDR5 RDIMM
• Maximum Memory: Up to 6TB
• Memory Speed: Up to 6000 MT/s with AMD EPYC 9005 Series
• Expansion Slots: 8 PCIe 5.0 x16 LP Slots
• Additional Expansion: Up to 4 PCIe 5.0 x16 FHFL Slots
• Storage Bays: Up to 16 PCIe 5.0 x4 NVMe U.2 Bays, plus 2 SATA 2.5-inch Bays
• Default Storage Configuration: 12 Front Hot-Swap 2.5-inch NVMe Bays, 2 Front Hot-Swap 2.5-inch SATA Bays and 4 Additional NVMe Bays
• M.2 Storage: 2 M.2 NVMe Slots
• Management LAN: 1 RJ45 Dedicated BMC LAN Port
• USB: 2 USB 3.0 Ports
• Video: 1 VGA Port
• Security: TPM 2.0, Secure Boot, Silicon Root of Trust, NIST 800-193 Compliance
• Power Supply: 6 x 3000W Redundant 3+3 Titanium-Level Power Supplies
• Cooling: 10 Heavy-Duty Fans with Optimal Fan-Speed Control
• Chassis: CSE-GP801TS
• Chassis Dimensions: 437 x 355.6 x 843.28 mmWhy Choose This Product?
• Operating Temperature: 10°C to 35°C

Why Choose This Product?

• AS-8125GS-TNMR2 brings eight AMD Instinct MI300X accelerators, dual AMD EPYC processing, up to 6TB DDR5 memory and PCIe 5.0 expansion into one 8U enterprise platform. Its GPU-direct RDMA architecture, Infinity Fabric connectivity, flexible storage and 3+3 redundant Titanium-level power make it well suited to large AI and HPC clusters. For organizations expanding or replacing GPU infrastructure, the system offers high accelerator density and room for networking, storage and memory growth while supporting enterprise data-center requirements.