• Huawei FusionServer G5500 V6 – 4 U Dual-Socket AI Chassis for Heterogeneous Compute

    Keywords

    FusionServer G5500 V6,#FusionServer G5500 V6, Huawei AI server,#Huawei AI server, Ice Lake Xeon,#Ice Lake Xeon, DDR4 memory,#DDR4 memory, Optane PMem,#Optane PMem, NVMe SSD,#NVMe SSD, OCP 3.0,#OCP 3.0, FusionDirector,#FusionDirector

    Description

    The Huawei FusionServer G5500 V6 chassis delivers extreme GPU/FPGA density with support for up to 8 dual-slot FHFL or 32 HHHL accelerator cards, leveraging GPUDirect RDMA and peer-to-peer interconnect for ultra-low latency GPU-to-GPU data transfers.

    Equipped with 2 × 3rd Gen Intel® Xeon® Scalable (Ice Lake) CPUs at 270 W TDP each on an Intel C621A chipset, it offers balanced CPU/GPU configurations for AI training, inference, and HPC workloads.

    With 32 DDR4 DIMM slots @ 3200 MT/s plus 16 Optane™ PMem 200 slots @ 3200 MT/s, the chassis supports up to 12 TB of combined volatile and persistent memory, optimizing in-memory analytics and virtualization density

    Key Features

    • Heterogeneous Compute Flexibility: One full-width or two half-width nodes with up to 8 FHFL or 32 HHHL GPUs/accelerators 

    • High-Speed Networking: 3 × OCP 3.0 NIC slots for InfiniBand/OPA/Ethernet; hot-swap support

    • Tiered Local Storage: Supports 36–44 3.5″ SAS/SATA + 4 NVMe SSD bays; dual M.2 SSDs for boot

    • Enterprise RAID: Levels 0/1/10/5/50/6/60; supercap cache backup; RAID migration; drive roaming; web-based config

    • Expansion Capacity: Up to 16 PCIe 4.0 slots (12 standard × x16, 3 OCP, 1 RAID)

    • Redundant Power & Cooling: 4 × 2000–3000 W PSUs (N+N/N+M); 6–8 hot-swap N+1 fans 

    • Intelligent Management: iBMC with Redfish, SNMP, IPMI 2.0, HTML5/VNC KVM; FusionDirector for lifecycle automation 

    • Patented DEMT & FDM: Dynamic Energy Management and Fault Diagnosis & Management for OPEX reduction and rapid fault resolution

    • Enterprise Security: TPM 2.0, Secure Boot, admin & power-on passwords, chassis intrusion detection

    Configuration

    Component Specification
    Model FusionServer G5500 V6 (XFUSION INTERNATIONAL PTE. LTD.)
    Form Factor 4 U AI server chassis (175×447×898 mm)
    Processors 2× Intel® Xeon® Scalable (Ice Lake) up to 270 W TDP; Intel C621A chipset
    Memory 32× DDR4 DIMMs @ 3200 MT/s; 16× Optane™ PMem 200 @ 3200 MT/s
    Local Storage 36–44×3.5″ SAS/SATA + 4×NVMe SSD; dual M.2 SSDs
    RAID 0/1/10/5/50/6/60; supercap cache; migration; drive roaming; web-based config
    Networking 3×OCP 3.0 slots; orderable hot-swap InfiniBand/OPA/Ethernet NICs
    PCIe Expansion 16×PCIe 4.0 slots: 12 standard × x16, 3×OCP, 1 RAID
    GPU Support 8×FHFL dual-slot or 32×HHHL single-slot accelerator cards
    Fan Modules 6–8×hot-swap counter-rotating fans (N+1 redundancy)
    Power Supplies 4×2000 W/2600 W/3000 W Platinum/Ti PSUs (AC/DC), hot-swap, N+N or N+M redundancy
    Management iBMC: Redfish, SNMP, IPMI 2.0, HTML5/VNC KVM; FusionDirector optional
    Security TPM 2.0, Secure Boot, passwords, chassis intrusion detection
    OS Support Windows Server; SUSE Linux Enterprise; RHEL; Ubuntu; Debian; openEuler
    Operating Temp. 5 °C–35 °C (41 °F–95 °F)
    Certifications CCC
    Installation Kit L-shaped, adjustable, and holding rails

    Compatibility

    • AI & Deep Learning: Supports NVIDIA/AMD GPU, Huawei Ascend NPU, and Intel FPGA cards via PCIe and OCP 3.0 slots

    • Network Fabrics: IB, OPA, 10/25/100 GbE via modular NICs

    • Management Platforms: Huawei FusionDirector, eSight, and third-party DCIM via Redfish/SNMP
      (XFUSION INTERNATIONAL PTE. LTD.)

    Usage Scenarios

    1. AI Training & Inference
      Eight dual-slot GPUs and high-speed PCIe 4.0 fabric accelerate large-scale deep learning model training and inference.

    2. High-Performance Computing
      Flexible CPU-to-GPU ratios and GPUDirect RDMA optimize multi-node HPC clusters for simulation and scientific workloads.

    3. Cloud & Virtualization
      Massive memory (DRAM + PMem) and modular compute nodes support dense VM or container orchestration in private clouds.

    4. Video Analytics & Database Acceleration
      Tiered NVMe/SAS storage and FPGA/GPU offload enable real-time video analysis and in-memory database query acceleration.

    Frequently Asked Questions

    Q1: How many GPUs can G5500 V6 host?
    A1: Up to 8 full-height dual-slot FHFL or 32 half-height HHHL GPUs/accelerators, each up to 350 W or 75 W respectively.

    Q2: What memory configurations are supported?
    A2: Supports 32× DDR4 DIMMs at 3200 MT/s alongside 16× Optane™ PMem 200 modules at 3200 MT/s for up to 12 TB combined memory.

    Q3: Which management interfaces are available?
    A3: Integrated iBMC offers Redfish, SNMP, IPMI 2.0, and HTML5/VNC KVM; FusionDirector adds intelligent full-lifecycle O&M with topology switchover and firmware automation.

    Q4: How does DEMT improve efficiency?
    A4: Dynamic Energy Management Technology adjusts power delivery to match workload demands, reducing overall server power consumption by up to 15 % without performance loss.

    Q5: What redundancy features exist?
    A5: The chassis includes hot-swap PSUs (N+N/N+M), hot-swap fans (N+1), chassis intrusion detection, and supercapacitor-backed RAID cache for zero-downtime maintenance and data protection.

与此商品相关的产品
Best Deal for FusionServer G5500 V6: High-performance 4U 2-Socket GPU Server 推荐
Best Price for FusionServer G5500 V6: High-performance 4U 2-Socket GPU Server 推荐
FusionServer G5500 V6: Powerful 4U 2-Socket GPU Server 推荐

FOLLOW US

价格以美元为单位，不含税。 商标和品牌为其各自所有者所有。