| Product SKU | SYS-821GE-TNHR |
|---|---|
| Motherboard | Super X13DEG-OAD |
| Processor | Dual Socket E (LGA-4677) |
| GPU | Maximum onboard GPUs: 8 |
| System Memory | 32 DIMM slots |
| Chipset | Intel C741 |
| Network Connectivity | Optional: 2 x SFP+ 10GbE with Intel X710-BM2 |
| Input / Output | 1 x VGA |
| BIOS | AMI 32MB SPI Flash EEPROM |
| Management Software | Supermicro Server Manager (SSM) |
| Security | Silicon Root of Trust supporting NIST SP 800-193 |
| PC Health Monitoring | Voltage monitoring for CPU cores, chipset and memory |
| Operating Systems | VMware ESXi 8.0 |
| Form Factor | 8U rackmount |
| Dimensions and Weight | System: 14 in (355.6 mm) H x 17.2 in (437 mm) W x 33.2 in (843.28 mm) D |
| Front Panel | Buttons: Power On/Off, Reset |
| Expansion Slots | Default: 8 x PCIe 5.0 x16 low-profile slots |
| Drive Bays / Storage | Default configuration: 15 front hot-swap 2.5-inch bays |
| M.2 | 2 x M.2 NVMe M-key slots |
| System Cooling | Up to 10 heavy-duty fans |
| Power Supply | 6 x 3000W redundant power supplies in 4+2 configuration |
| Operating Environment | Operating temperature: 10°C to 35°C |
Servers
Supermicro SYS-821GE-TNHR 8U GPU Server
The Supermicro SYS-821GE-TNHR is an 8U dual-socket GPU server supporting NVIDIA HGX H100/H200 8-GPU platforms, 4th/5th Gen Intel Xeon Scalable processors, up to 8TB DDR5 memory, and flexible NVMe/SATA storage.
Overview
The Supermicro SYS-821GE-TNHR is an 8U dual-socket GPU server designed for large-scale AI, deep learning and high-performance computing workloads. It supports NVIDIA HGX H100 or H200 8-GPU platforms, 4th and 5th Gen Intel Xeon Scalable processors, up to 8TB of DDR5 memory, and high-density NVMe/SATA storage.
Key Features
- 8U rackmount platform with rear I/O
- NVIDIA HGX H100 8-GPU 80GB or HGX H200 8-GPU 141GB
- Dual Socket E (LGA-4677) for 4th/5th Gen Intel Xeon Scalable processors
- 32 DIMM slots with up to 8TB ECC DDR5 RDIMM
- PCIe 5.0 expansion and NVIDIA NVLink with NVSwitch
- Up to 19 front hot-swap 2.5-inch drive bays depending on configuration
- Six 3000W redundant Titanium-level power supplies in a 4+2 configuration
Applications
- HPC and AI/deep learning training
- Conversational AI and business intelligence
- Industrial automation and healthcare
- Drug discovery and climate or weather modeling
- Finance and economics
Configuration availability depends on selected GPU, networking, storage and expansion options. Review the required configuration before quotation.












