AI and High-Performance Computing Solutions
Covers the SuperAI deep learning platform, SuperHPC high-performance computing and SuperHCI hyper-converged infrastructure. From GPU clusters and parallel file systems to hyper-converged cloud data centers, it delivers a software-hardware integrated compute foundation for AI training, engineering simulation and scientific research.
AI & HPC Product Family
From deep learning training platforms and full-stack supercomputing to hyper-converged cloud data centers
SuperAI Deep Learning Platform
Built on a high-performance GPU computing hardware platform, SuperAI integrates Docker and Kubernetes with an AI management platform to deliver a deep learning solution purpose-built for AI analytics. It uses GPU servers designed for AI workloads, supports the latest 5th Gen Intel Xeon Scalable processors, and supports up to 10 GPUs per node, with mainstream frameworks such as TensorFlow and PyTorch.
SuperHPC High-Performance Computing
A full-stack HPC platform built on a five-tier architecture: hardware (X86/ARM compute nodes), system (parallel file system), cluster (unified deployment and monitoring), tooling (parallel computing libraries) and application (simulation design, fluid dynamics). Its core cluster scheduling software integrates job scheduling, resource management, cluster monitoring and application templates.
SuperHCI Hyper-Converged
Designed for the future intelligent data center, SuperHCI provides flexible, comprehensive APIs to manage computing, storage and network resources, combined with hyper-converged server hardware, so organizations can rapidly build their own smart cloud data center. Guided by the 4S principles โ Simple, Strong, Scalable, Smart โ it supports KVM and VMware virtualization and works with DAS/NAS/SAN/DFS storage and VLAN/VXLAN network models.
Core Capabilities of AI & HPC
GPU clusters with containerized scheduling
Built on high-performance GPU hardware and integrated with Docker and Kubernetes, the AI management platform centrally schedules compute resources to support concurrent training jobs across teams.
Five-tier software-hardware integration
Hardware, system, cluster, tooling and application layers work together, delivered as one system from compute nodes to parallel file systems and parallel computing libraries, reducing integration risk.
Job scheduling and resource management
Cluster scheduling software integrates job scheduling, resource management, cluster monitoring and application templates for efficient job scheduling and resource utilization.
Parallel file system with high-I/O optimization
The system layer adopts a parallel file system with high-I/O optimization for massive small-file workloads such as meteorology and oceanography, relieving storage bottlenecks.
Hyper-converged resource pooling
APIs centrally manage compute, storage and network resources, supporting KVM/VMware virtualization and multiple storage and network models for elastic, on-demand scaling.
Turnkey delivery, ready to run
HPC and AI platforms are delivered as complete systems, significantly reducing deployment complexity and shortening the path from purchase to production.
Typical Deployment Scenarios
CAE Engineering Simulation
Structural mechanics and fluid dynamics simulation with integrated environment software packages to accelerate R&D iteration.
Basic Research & Meteorology
Multi-user parallel program support and high-I/O optimization for massive small files, meeting research and weather forecasting needs.
AI Model Training & Fine-Tuning
Large-scale deep learning training, computer vision, NLP and recommendation system development.
Smart Cloud Data Center
Rapidly build enterprise private cloud and intelligent data centers on a hyper-converged architecture to host multiple workloads.
The above describes typical application scenarios and does not represent specific customer cases.
AI & HPC Selection Guide
| Dimension | SuperAI Deep Learning | SuperHPC | SuperHCI Hyper-Converged |
|---|---|---|---|
| Positioning | GPU deep learning platform | Full-stack HPC platform | Hyper-converged cloud data center |
| Architecture | GPU servers + Docker/K8S + AI management platform | Five-tier architecture + cluster scheduler | Compute/storage/network pool + APIs |
| Typical workload | Deep learning training and inference | CAE simulation, research computing, meteorology | Virtualization, databases, private cloud |
| Key features | Up to 10 GPUs per node | Job scheduling / resource management / monitoring / templates | 4S principles, KVM and VMware, multiple storage and network models |
| Delivery | Turnkey system | Turnkey system | Turnkey system |
| Scenarios | AI R&D teams | Research and engineering simulation | Enterprise cloud data center |
Frequently Asked Questions
Which AI development scenarios is the SuperAI deep learning platform suited for?
SuperAI is built on a high-performance GPU computing hardware platform and integrates Docker, Kubernetes and an AI management platform. It suits large-scale deep learning model training, computer vision, natural language processing and recommendation systems. A single node supports up to 10 GPUs, with mainstream frameworks such as TensorFlow and PyTorch supported, providing AI R&D teams with a robust and reliable computing platform.
How does SuperHPC differ from a conventional server cluster?
SuperHPC is a full-stack HPC platform. Through a five-tier architecture โ hardware (X86/ARM compute nodes), system (parallel file system), cluster (unified deployment and monitoring), tooling (parallel computing libraries) and application (simulation design, fluid dynamics) โ it builds a software-hardware integrated HPC environment. Its core cluster scheduling software integrates job scheduling, resource management, cluster monitoring and application templates, and it is delivered as a turnkey system that significantly reduces deployment complexity.
How does SuperHCI hyper-convergence help build a smart cloud data center?
SuperHCI provides flexible, comprehensive APIs to manage computing, storage and network resources across the data center, combined with hyper-converged server hardware, so organizations can quickly build their own smart cloud data center. Guided by the 4S principles โ Simple, Strong, Scalable and Smart โ it supports KVM and VMware virtualization, storage types including DAS/NAS/SAN/DFS, and network models such as VLAN/VXLAN.
Further Reading
AI Computing All-in-One
RoycomONE platform, DeepSeek all-in-one and the three-tier collaborative full-stack AI solution.
SuperStorage
RS8000 all-flash, RS6000 massive scale, RS4000 unified storage and Weka.
AI Servers
AI, general-purpose, storage, domestic, liquid-cooled and customized servers.
Product Center
Explore the full VISBAT enterprise network and security product portfolio.