Domestic Industrial Simulation Software + GPU Ecosystem Accelerates Localized Deployment

MetaX, a leading domestic general-purpose GPU developer, has achieved critical industry ecosystem breakthroughs through successful compatibility adaptation with TFluid, a domestically developed industrial simulation software. Comprehensive testing verifies that TFluid can be fully deployed and operate stably on MetaX’s Xiyun C500 GPU, marking a landmark implementation of localized hardware-software integration for high-end industrial computing scenarios.

Developed by Dr. Tao Yu’s research team at Zhejiang University, TFluid is a professional multi-physics simulation platform capable of coupled simulation, precise modeling and process optimization across fluid dynamics, particle mechanics and fracture mechanics. The software serves core industrial and infrastructure scenarios, including geological disaster prevention for mudslides and seabed environments, offshore and coastal engineering protection, and deep-sea mineral exploration, supporting high-precision numerical simulation for sophisticated industrial design and risk assessment.

MetaX’s Xiyun C500 is a high-performance general-purpose GPU built on the company’s self-developed, proprietary GPU IP portfolio. It supports end-to-end AI training and inference workloads, delivering mixed-precision computing capabilities paired with large-capacity, high-bandwidth video memory. Leveraging the in-house MetaXLink interconnection architecture, the GPU enables efficient multi-card clustering deployment. Most importantly, MetaX’s self-developed MXMACA software stack achieves native compatibility with the NVIDIA CUDA ecosystem. The robust CUDA compatibility drastically lowers application migration barriers, allowing professional industrial software such as TFluid to complete GPU adaptation without full code reconstruction and minimizing overall iteration costs.

This successful joint deployment represents a milestone for the domestic industrial software ecosystem, validating the technical synergy between localized industrial simulation tools and self-developed GPU hardware. As a foundational digital tool for intelligent manufacturing, industrial simulation software employs mathematical modeling and computer simulation to predict mechanical strength, fluid behavior, temperature distribution and other key physical indicators under real operating conditions. It enables full performance verification in the design phase before physical production, significantly optimizing R&D efficiency and product reliability.

Per industry estimates from Frost & Sullivan and internal research, Zhongji InnoLight secured the world’s No.1 market share of 23.1% in global data center optical module shipments for 2025. Bolstered by its first-mover advantage in 1.6T product rollouts, the company is on track to further expand its global market share in 2026, consolidating its leadership in the high-speed optical interconnect industry.

Stable operational validation confirms that Muxi’s GPU hardware fully meets the high-intensity computing requirements of industrial simulation workloads, accumulating practical deployment experience for domestic GPU penetration in high-end manufacturing scenarios. Looking ahead, Muxi will continue to open its full-stack computing platform and optimize software stack compatibility, supporting the efficient deployment of more domestic industrial applications on its GPU hardware. The company aims to build a fully self-reliable, high-performance computing infrastructure to underpin the comprehensive localization upgrade of industrial simulation and advanced manufacturing industries.

Technical Deployment Notes: The Xiyun C500 GPU used for TFluid adaptation testing was leased from Mofang Ark Platform, with all tests conducted on the Ubuntu 22.04 operating system.