A GPU Accelerated Era
Breaking Barriers in Accelerated Computing and Generative AI
Blackwell-architecture GPUs
208 billion transistors with TSMC 4NP process
2nd Gen Transformer Engine
Doubling the performance with FP4 enablement
5th Gen NVLink & NVLink Switch
1.8 TB/s GPU-GPU interconnect
RAS Engine
100% In-system self-test
Secure AI
Full performance encryption & TEE
Decompression Engine
800 GB/s
GIGABYTE's Commitment to Flexible and Scalable Solutions
Short TTM for Agile Deployment
Flexible Scalability for Diverse Scenarios
Comprehensive One-Stop Service for Optimized Configuration

NVIDIA GB300 NVL72: Built for the Age of AI Reasoning
The NVIDIA GB300 NVL72 features a fully liquid-cooled, rack-scale design that unifies 72 NVIDIA Blackwell Ultra GPUs and 36 Arm®-based NVIDIA Grace™ CPUs in a single platform optimized for test-time scaling inference.
AI factories powered with the GB300 NVL72 using NVIDIA Quantum-X800 InfiniBand or Spectrum™-X Ethernet paired with NVIDIA ConnectX®-8 SuperNIC™ provide a 50x higher output for reasoning model inference compared to the NVIDIA Hopper™ platform.
Fast Memory 60Xvs. NVIDIA H100 HGX | HBM Bandwidth 20Xvs. NVIDIA H100 HGX | Networking Bandwidth 18Xvs. NVIDIA H100 HGX |
![]() | ||
XN15-CB0-LA01 Compute Tray
|
NVIDIA HGX™ B300 / B200: Optimized for AI and High-Performance Computing
- Support for full-height add-in cards, accommodating DPUs and SuperNICs.
- A PCIe cage design and front-access motherboard/GPU trays for streamlined maintenance.
- Hot-swappable, fully redundant PSUs with multiple connector options for enhanced flexibility.

NVIDIA HGX™ B300
- 8 x NVIDIA Blackwell Ultra GPUs
- Up to 2.1TB of GPU memory
- 105 petaFLOPS training performance
- 144 petaFLOPS inference performance
- 1.8TB/s GPU-to-GPU bandwidth with NVIDIA NVLink™ and NVSwitch™

G894-SD3-AAX7
- NVIDIA HGX™ B300
- 8 x 800 Gb/s OSFP InfiniBand XDR or Dual 400 Gb/s Ethernet GPU networking ports via onboard NVIDIA ConnectX®-8 SuperNIC™
- Compatible with NVIDIA® BlueField®-3 DPUs and ConnectX®-7 NICs
- Compatible with NVIDIA® BlueField®-3 DPUs
- Dual Intel® Xeon® 6700/6500-Series Processors
- 2 x 10Gb/s LAN ports
- 8 x 2.5" Gen5 NVMe hot-swap bays
- 4 x FHHL PCIe Gen5 x16 slots
- 12 x 3000W 80 PLUS Titanium redundant power supplies

G4L3 4U HPC/AI Server
- Liquid-cooled NVIDIA HGX™ B200
- Dual 5th/4th Gen Intel® Xeon® Scalable or Dual AMD EPYC™ 9005/9004 Series CPUs
- Compatible with NVIDIA® BlueField®-3 DPUs and ConnectX®-7 NICs
- 2 x 10Gb/s LAN ports
- 8 x 2.5" Gen5 NVMe hot-swap bays
- 12 x FHHL PCIe Gen5 x16 slots
- 4+4 3000W 80 PLUS Titanium redundant PSUs