Gigabyte H223-V10-AAW1 8NVMe
- Condition
- New
- Manufacturer
- GIGABYTE
- Form factor
- 2U
- RAM
- 960GB (LPDDR5x)
- NIC card
- 4x 10GbE (Intel X550-AT2)
- GPU
- 2 x NVIDIA GH200 (Hopper)
- Accelerator manufacturer
- NVIDIA
- Accelerator model
- NVIDIA GH200 Grace Hopper Superchip
- Installed GPU/APU count
- 2
- Maximum GPU/APU count
- 2
- Memory per accelerator
- 96GB
- Total accelerator memory
- 192GB
- GPU memory type
- HBM3
- Accelerator memory bandwidth
- 4 TB/s per GPU
- Accelerator form factor
- Integrated NVIDIA Grace Hopper Superchip
- Accelerator interconnect
- No chassis-wide GPU-to-GPU fabric; each GPU is integrated with its Grace CPU through NVIDIA NVLink-C2C
- Accelerator-to-CPU interface
- NVIDIA NVLink-C2C coherent interface
- Supported compute data types
- FP64, FP32, TF32, BF16, FP16, FP8 and INT8
- Accelerator virtualization
- GPU passthrough and containerized CUDA workloads; NVIDIA vGPU availability depends on the deployed software stack
- MIG support
- Up to 7 isolated GPU instances per GPU
- Software ecosystem
- NVIDIA CUDA, NVIDIA AI Enterprise and the NVIDIA Grace Hopper software platform
- Maximum system power
- 9kW
- Cooling
- Air cooling with four 80 mm system fan modules
- Recommended workloads
- Large-model AI training and inference, scientific computing, HPC, simulation, analytics and memory-intensive accelerated workloads
- Dimensions
- 87.5 x 440 x 850 mm (H x W x D)
- Weight
- 50.91 kg
Need a technical consultation on choosing a server?
Submit a requestGigabyte H223-V10-AAW1 8NVMe is a 2U two-node enterprise GPU system storage configuration with eight front 2.5-inch PCIe Gen5 NVMe hot-swap positions distributed across two independent nodes. It is designed for AI and HPC clusters, research centers and enterprise data centers that need two Grace Hopper nodes in a compact chassis.
Key features:
-
Storage basket and service access: The chassis provides eight front 2.5-inch PCIe Gen5 NVMe hot-swap positions, four assigned to each node. Front access lets administrators replace supported drives without opening the chassis and preserves node-level service separation.
-
Accelerator subsystem: The installed accelerator subsystem contains 2 x NVIDIA GH200 Grace Hopper Superchip, providing 192GB HBM3 or 288GB HBM3E across two nodes of aggregate GPU memory. NVIDIA NVLink-C2C within each Grace Hopper Superchip; 900 GB/s.
-
Storage topology and data path: Each node receives four direct PCIe Gen5 NVMe bays. 4 internal M.2 PCIe Gen5 x4 slots in total (2 per node). This layout supports fast dataset staging, checkpoints and local scratch storage without mixing the basket with a different drive protocol.
-
Enterprise compute platform: The compute subsystem is configured with 2 x NVIDIA Grace 72-core CPUs, one per node and 960GB LPDDR5X ECC CPU memory in total (480GB per node). It supplies CPU, memory and PCIe resources for the accelerator complex and production services.
-
Management and availability: ASPEED AST2600 BMC and GIGABYTE Management Console for each node; 2+1 x 3000W 80 PLUS Titanium redundant power supplies. These facilities support remote administration, health monitoring and resilient operation.
Use cases:
-
AI model development: Uses the installed NVIDIA GH200 Grace Hopper Superchip subsystem for training, fine-tuning and large-model development.
-
Inference and production AI: Provides parallel compute and accelerator memory for LLM, multimodal and enterprise inference services.
-
HPC, simulation and analytics: Accelerates scientific computing, numerical simulation, analytics and other parallel workloads supported by NVIDIA GH200 Grace Hopper Superchip.
-
Data-intensive pipelines: Combines eight front NVMe bays across two nodes with coherent Grace Hopper compute for datasets, checkpoints, staging and local scratch space.
Gigabyte H223-V10-AAW1 8NVMe is a purpose-built 2U two-node platform for organizations that need eight front hot-swap NVMe bays across two independent nodes, 192GB HBM3 or 288GB HBM3E across two nodes of installed GPU memory and enterprise-class compute and management in one deployable system.
to Buy a Server
from Servermall
warranty
testing
We want to make sure that our equipment is fully reliable for you. That’s why all our equipment (even new) is subject to complete diagnostics, testing, and part replacement.
testing
We fully load our equipment with various utilities and software. Rest assured, it’ll work for a very long time.
Power supply
The components work at full capacity and are tested for several days. If we are not sure about something, we replace it.
of all equipment
Our regulations imply the same testing for both used and new equipment. Oh, and by the way, we complement the official manufacturer’s warranty with our own warranty.