4U AI server with dual Intel® Xeon® CPUs (up to 385 W each)
32 DDR5 DIMMs, up to 256 GB per DIMM
Supports 10 dual-width GPUs for AI/HPC workloads
Storage: Up to 24 SAS/SATA or 12 NVMe drives + M.2 SSDs
14 PCIe 5.0 slots, including 10 GPU-dedicated slots
3 x OCP 3.0 NICs, high-speed networking
Redundant cooling and 4 hot-swappable PSUs (up to 3000 W)
iBMC management with optional FusionDirector
Security: TPM 2.0, secure boot, chassis intrusion detection
Compatible with Windows Server and major Linux OS
FusionServer G5500 V7 - Balanced and Flexible Configuration, Preferred Option for AI
AI Server
FusionServer G5500 V7 (G5500 V7) is a next-generation 4U 2-socket AI server. It supports up to 10 x dual-width GPU cards and has the capability to accommodate 4 x standard PCIe cards and 3 x OCP cards for extensive expansion. With 24 x 2.5/3.5" drives or 12 x NVMe SSDs, it provides remarkable storage options with ultra-large capacity or ultra-fast storage. G5500 V7 features high performance, flexible architecture, solid reliability, easy deployment, and simplified management. It is designed for accelerating applications such as AI training, AI inference, high-performance computing (HPC), image and video analysis, and databases, and supports enterprise and public cloud deployment.
.jpg)
- General computing: 2 x 4th or 5th Gen Intel® Xeon® Scalable processors with 385 W TDP per processor, and 32 x DDR5 DIMMs, improving overall performance by up to 50%
- Intelligent computing: up to 10 x dual-width GPU cards, improving AI computing by 25% compared with the server with 8 x GPU cards
- Large storage: up to 24 x 2.5/3.5" SAS/SATA drives or 12 x NVMe SSDs + 8 x SATA/SAS drives
- Numerous I/Os: typical configuration of 8 x dual-width GPUs, another 6 x standard PCIe cards and 3 x OCP NICs, leading the industry
- One-click switchover between cascaded and balanced topologies, flexibly adapting to multiple scenarios
- PCIe Retimer-free design, reducing the PCIe lane latency and system power consumption
- GPUDirect Storage/RDMA/P2P, adaptive to large-scale cluster deployment
- Innovative drive design with vibration resistance and noise reduction, diminishing the failure rate by 60%
compared to the industry average
- Enhanced system robustness design and automatic firmware recovery
- Reliable fan redundancy: 6 or 8 x customized 8080+ fans in N+1 redundancy
- PSU redundancy and high efficiency: 4 x 3000 W Titanium PSUs in N+N/N+M redundancy
