Gigabyte G492-PD0 6NVMe
- Condition
- New
- Manufacturer
- GIGABYTE
- Form factor
- 4U
- RAM
- 4096GB (DDR4 ECC REG)
- NIC card
- 2x 1GbE (Intel I350-AM2)
- GPU
- 8 x NVIDIA A100
- Accelerator manufacturer
- NVIDIA
- Accelerator model
- NVIDIA A100 (HGX A100)
- Installed GPU/APU count
- 8
- Maximum GPU/APU count
- 8
- Memory per accelerator
- 80GB
- Total accelerator memory
- 640GB
- GPU memory type
- HBM2e
- Accelerator memory bandwidth
- 2039GB/s per GPU
- Accelerator form factor
- SXM4
- Accelerator interconnect
- Third-generation NVIDIA NVLink with NVSwitch
- Accelerator-to-CPU interface
- PCIe 4.0 x16
- Supported compute data types
- FP64, FP32, TF32, BF16, FP16 and INT8
- Accelerator virtualization
- NVIDIA vGPU, SR-IOV and GPU passthrough
- MIG support
- Up to 7 isolated GPU instances per GPU
- Software ecosystem
- NVIDIA CUDA, NVIDIA AI Enterprise and NVIDIA GPU Operator
- Maximum system power
- 12kW
- Cooling
- Air cooling with passive SXM4 accelerator modules and hot-swap high-performance fan modules
- Recommended workloads
- Deep learning training and inference, model fine-tuning, HPC, scientific simulation, analytics and data science
- Dimensions
- 448 x 175.2 x 900 mm (W x H x D)
Need a technical consultation on choosing a server?
Submit a requestGigabyte G492-PD0 6NVMe is a 4U enterprise GPU system storage configuration with six front 2.5-inch U.2 PCIe Gen4 NVMe hot-swap bays. It is designed for AI and HPC environments that need dense HGX A100 compute with a fast NVMe-only local-storage tier.
Key features:
-
Storage basket and service access: The chassis provides six front-accessible 2.5-inch U.2 PCIe Gen4 NVMe hot-swap bays. Supported drives can be replaced without opening the chassis.
-
Exact drive topology: All six positions are NVMe-only; SATA and SAS are not part of this basket.
-
Backplane and storage integration: The front backplane exposes six NVMe ports, while four internal M.2 PCIe Gen4 x4 positions provide separate boot and utility storage. 4 x internal M.2 PCIe Gen4 x4 slots.
-
Enterprise compute platform: The compute subsystem combines 1 x Ampere Altra or Altra Max processor, up to 128 cores, 4,096GB DDR4 ECC registered memory; 16 DIMM platform, and 8 x NVIDIA A100 SXM4 80GB accelerators with 640GB (8 x 80GB) of aggregate HBM2e memory. This provides CPU, memory and accelerator resources for sustained production workloads.
-
Management and availability: ASPEED AST2500 BMC, dedicated management LAN and GIGABYTE Management Console; 3+1 x 3000W 80 PLUS Platinum redundant power supplies. These facilities support remote administration, health monitoring and resilient operation.
Use cases:
-
AI model training and fine-tuning: Uses the installed NVIDIA A100 SXM4 (HGX A100) subsystem for model training, fine-tuning and accelerated development.
-
Inference and production AI services: 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 A100 SXM4 (HGX A100).
-
Data-intensive pipelines: Combines six front U.2 NVMe bays with the HGX A100 data path for datasets, checkpoints, staging and local scratch space.
Gigabyte G492-PD0 6NVMe is a purpose-built 4U platform for organizations that need six front U.2 PCIe Gen4 NVMe hot-swap bays, accelerated compute and enterprise-class 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.