National Research and Education Network
Education Cloud Data Center
Multi-Channel HD Telemedicine
Over The Top/Multi-Tenant Data Center (OTT/MTDC)
Internet Exchange Point (IXP)
Internet Access Provider (IAP)
Design & Simulation
Planning & Analytics
Oil & Gas IoT
HPC & Operations Management
Digital Urban Rail
Enterprise Data Center
Server - Intelligent Computing
Enterprise Cloud Communications
Network Management System
Ubiquitous Connectivity Makes Digital Transformation Possible
Economical and Powerful AI Computing
Secure And Resilient Solutions
Leading New ICT
228 Fortune Global 500 Companies Choose Huawei as Digital Transformation Partner
Buy from Huawei
If you want to get more information about your project, you can submit your information and we will contact you as soon as possible.
If your company has signed an eDeal contract with Huawei, please buy your required product/solution via the link below.
Buy from resellers
Search for a nearby reseller and get direct contact information.
Become a Partner
Resources and Support
Huawei Authorized Learning Partner
Huawei Authorized Information and Network Academy
Huawei Atlas G5500 is a heterogeneous server oriented to data center deployment. It features high-density heterogeneous computing capabilities, and supports GPUDirect Remote Direct Memory Access (RDMA) and Peer-to-Peer (P2P), and NVLink interconnect technologies. It also supports one-click topology switching, offering the optimal heterogeneous topology configurations for diverse applications. It uses a fully modular design and supports long-term evolution of CPUs and heterogeneous components. Huawei Atlas G5500 is the best-in-class heterogeneous computing platform for scenarios such as artificial intelligence (AI), HPC, intelligent cloud, video analytics, and database acceleration.
The 4U chassis supports up to 8 full-height, full-length, dual-slot heterogeneous accelerator cards of up to 350W, or 32 half-height, half-length, single-slot heterogeneous accelerator cards of up to 75W. It includes GPUDirect RDMA and Peer-to-Peer (P2P) interconnection technologies for memory access with multiple GPU cards.
Supports 1 full-width or 2 half-width heterogeneous compute nodes, one-click topology switching, and various topologies with different CPU:GPU configuration ratios: 1:2, 1:4, or 1:8.
Leverages decoupled CPU and heterogeneous modules to support long-term evolution of CPUs and heterogeneous components. Power Supply Units (PSUs), hard drives, and fan modules use a fully modular design, supporting hot-swap and redundancy backup.
|Node Support||1 full-width or 2 half-width heterogeneous compute nodes|
|Power Supply Units||Four 80 PLUS Platinum 2,200W AC hot-swappable PSUs; N+N redundancy|
|Fan Modules||6 hot-swappable fan modules with smart speed tuning; N+1 redundancy|
|PCIe Expansion||4 PCIe Gen3 x16 HHHL NICs (IB/OPA/Ethernet)|
|Management||Aggregated management network port|
|Power Supplies||110V/220V AC or 240V HVDC|
|Dimensions (H x W x D)||175 mm x 447 mm x 790 mm|