We take cryptocurrency — so the coins you hold can buy real hardware.Pay in crypto, spend it on real hardware. Free insured shipping worldwide over $399. Plain, unbranded boxes, sent from Germany.Free insured shipping over $399. Your rate is locked for 30 minutes once checkout opens.Rate locked 30 minutes. Every item sealed, serial-checked and covered by a 24-month warranty.Sealed, 24-month warranty.

Saved items Sign in
2U rack server, matte black metal chassis, front drive bays, rear redundant power supplies

Supermicro

Supermicro 821GE-TNHR-OTO-2-Radium rack server, Dual Xeon 8558H, 1TB DDR5, 960GB SSD, 8 H200 141GB

SKU CH-P5GH-WBAJ1 MPN 821GE-TNHR-OTO-2-Radium

Dual Xeon 48-core server with 1 TB DDR5 memory, eight H200 Tensor Core GPUs and 960 GB SSD

  • Dual Xeon 8558H deliver 96 threads at 2.1 GHz with 260 MB cache
  • 1 TB DDR5-5600 memory across 32 modules supports large-model training
  • Eight H200 Tensor Core GPUs provide 141 GB each for accelerated workloads
  • 960 GB SSD rated at 1 DWPD ensures reliable boot and logging storage
  • Redundant 3000 W power supplies with 54 V and 12 V outputs maintain uptime

Supermicro 821GE-TNHR Dual Xeon Server

The Supermicro 821GE-TNHR-OTO-2-Radium is a 2U GPU server system built for large-scale AI training and high-performance computing clusters. It integrates dual Intel Xeon Platinum 8558H processors, each delivering 48 cores and 96 threads at 2.1 GHz with 260 MB cache. Memory capacity reaches 1 TB through 32 DDR5-5600 modules. Eight NVIDIA H200 Tensor Core GPUs provide 141 GB each for massive model workloads.

Suited for AI labs needing maximum GPU density per

Organisations training multi-trillion-parameter models or running inference at scale will use the eight H200 GPUs and 1 TB system memory. Buyers who need more than eight GPUs in a single node, or who require a different GPU architecture, should evaluate alternative platforms. Teams without dedicated 54 V power infrastructure or compliance approval processes will also need a different solution.

Eight H200 GPUs with 141 GB each decide fit

The HGX 8-GPU configuration delivers a fixed 141 GB per accelerator, setting the hard limit for single-model parameter count without pipeline parallelism. Dual Xeon 8558H CPUs supply 96 total threads and 2.1 GHz base frequency for data preprocessing. Dual 3000 W redundant power supplies with 54 V and 12 V outputs power the complete system. A 960 GB 1 DWPD SSD handles boot and local staging.

Highlights

  • Dual Xeon 8558H deliver 96 threads at 2.1 GHz with 260 MB cache
  • 1 TB DDR5-5600 memory across 32 modules supports large-model training
  • Eight H200 Tensor Core GPUs provide 141 GB each for accelerated workloads
  • 960 GB SSD rated at 1 DWPD ensures reliable boot and logging storage
  • Redundant 3000 W power supplies with 54 V and 12 V outputs maintain uptime

Specifications

No published specifications for this item yet.

Questions about this item

Will the Dual Xeon 8558H 48-core CPUs and 1 TB DDR5-5600 memory work with my existing PCIe 4.0 storage and networking cards?

The server uses current-generation Xeon processors and DDR5-5600 memory, which are compatible with standard PCIe interfaces used by existing add-in cards.

How can I verify the eight H200 141 GB Tensor Core GPUs are all recognised and running at full capacity after installation?

Use the NVIDIA system management interface to confirm all eight GPUs appear with 141 GB memory each and check that GPU utilisation scales across all devices during workload runs.

Does the package include the 3000 W 54 V and 12 V redundant power supplies and rail kit, or must I order them separately?

The 3000 W redundant power supplies are integrated into the 2U chassis; rail kits and any additional cabling are typically ordered as separate accessories.

Also in this aisle

People compared these

Same shelf, same checkout — eight coins and a 30-minute rate lock.