The National Supercomputing Center in Shenzhen has just unveiled LineShine, an exascale-level supercomputer powered by 47,000 entirely Chinese processors. This machine marks the first attempt to reach exascale with a 100% domestic architecture, without resorting to Western GPUs. LineShine embodies Chinese ambitions to permanently free itself from technological sanctions and create an autonomous high-performance computing ecosystem.
The Essentials
- LineShine achieves exascale performance with 47,000 Chinese SW26010-Pro processors, without any foreign GPU technology
- The supercomputer operates with the Chinese ShenWei operating system and the SW-26010 architecture developed by the Institute of Microelectronics of the Chinese Academy of Sciences
- This 100% CPU approach represents a technological gamble against NVIDIA and AMD GPU hegemony in high-performance computing
- LineShine is part of China’s response to American restrictions on advanced semiconductors since 2022
- China thus joins the United States in the exascale club, establishing two parallel technological ecosystems
LineShine Demonstrates the Effectiveness of Chinese Massively Parallel Computing
The LineShine supercomputer relies on 47,000 SW26010-Pro processors, an evolution of the SW26010 processor that already equipped Sunway TaihuLight, a former world champion in high-performance computing. This 64-bit chip uses RISC architecture developed by the Institute of Microelectronics of the Chinese Academy of Sciences. Each SW26010-Pro processor integrates 6 groups of cores with 64 elementary cores, totaling 384 cores per processor.
The machine totals more than 18 million computing cores, distributed across 47,000 nodes interconnected by a proprietary Chinese network. This massively parallel approach favors multiplying computing units rather than optimizing individual ultra-powerful processors. The ShenWei operating system, developed specifically for this architecture, manages coordination among the millions of simultaneous computing threads.
LineShine achieves significant exascale performance, computing power comparable to American supercomputers Frontier and Aurora. This computational power enables simulation of complex physical phenomena requiring billions of differential equations, such as fluid dynamics in space mechanics or molecular interactions in pharmaceutical research.
The Anti-GPU Architecture Reveals a Philosophy of Technological Autonomy
LineShine distinguishes itself by its complete rejection of graphics processors, a technology dominated by American companies NVIDIA and AMD. This 100% CPU approach reflects a strategy of technological autonomy in the face of Western restrictions on advanced semiconductors. Since October 2022, the United States has prohibited the export to China of high-performance GPUs such as NVIDIA H100 or advanced manufacturing process chips.
China responds with radical architectural innovation. Where Western supercomputers combine CPU and GPU to optimize different types of calculations, LineShine bets on uniformity and specialization of its SW26010-Pro processors. This approach simplifies programming and avoids bottlenecks between CPU and GPU that sometimes limit the performance of hybrid architectures.
The Chinese technological gamble relies on complete control of the production chain. SW26010-Pro processors are manufactured using 14-nanometer technology by SMIC, the Chinese semiconductor foundry. This manufacturing process, less advanced than the 5 or 3 nanometers of the most cutting-edge chips, compensates through software optimization and energy efficiency. LineShine consumes approximately 15 megawatts, roughly half what the first generation of exascale supercomputers consumed.
China Accelerates the Emergence of Two Parallel Technological Ecosystems
LineShine marks the culmination of fifteen years of Chinese investment in domestic high-performance computing. The national program launched in 2009 explicitly aimed at technological independence in the face of potential Western restrictions. This anticipatory strategy is paying dividends as geopolitical tensions fragment the global technological ecosystem.
The impact extends beyond pure scientific computing. LineShine will serve as a testing ground for Chinese artificial intelligence algorithms, notably language models and recognition systems developed by Baidu, Alibaba, and Tencent. These applications require massive computing power, traditionally supplied by NVIDIA GPUs. The SW26010-Pro architecture offers a credible alternative for training AI models without Western technological dependency.
This technological bifurcation has accelerated since 2022. The United States is strengthening its restrictions on advanced semiconductors, war returns to economic calculation and weighs on vulnerable countries forced to choose their technological camp. China responds with massive investments in its national champions: SMIC for semiconductors, Sunway for high-performance computing, and dozens of start-ups specialized in AI chips.
LineShine’s Performance Repositions China in the Global Exascale Race
LineShine places China on par with American supercomputers Frontier (1.1 exaflops) and Aurora (2 exaflops), confirming Chinese mastery of high-performance computing technology. This performance arrives four years after the first American restrictions on advanced chips, demonstrating the adaptive capacity of the Chinese technological ecosystem.
The machine equips the National Supercomputing Center in Shenzhen, a research hub for Chinese universities and technology companies in the region. LineShine will support national research programs in climate, energy, and materials science. These domains require intensive numerical simulations, traditionally monopolized by Western supercomputers.
The SW26010-Pro architecture specifically optimizes double-precision calculations necessary for scientific simulations. This specialization contrasts with the Western approach that prioritizes flexibility between scientific computing and commercial applications. LineShine sacrifices versatility to maximize efficiency in its target domains, a strategy consistent with Chinese centralized planning of research priorities.
LineShine’s performance also validates the Chinese approach to technological development through leaps. Rather than gradually improving existing architectures, China invests in technological breakthroughs that circumvent Western bottlenecks. This method draws inspiration from the Chinese space strategy that developed independent launch vehicles and space stations in just a few decades.
The West Loses Its Monopoly on Strategic High-Performance Computing
LineShine redefines the geopolitical balance of high-performance computing. The United States can no longer control access to its rivals’ supercomputers through semiconductor restrictions. This loss of diplomatic leverage forces the West to rethink its technology containment strategy toward China.
The European Union is closely observing this bifurcation. The EuroHPC initiative develops its own supercomputers with European ARM processors, but remains dependent on American technologies for critical components. LineShine proves that a completely autonomous alternative remains possible, questioning the European strategy of technological alliance with the United States.
This fragmentation of high-performance computing announces the emergence of two parallel ecosystems: American-Western and Chinese. Each bloc develops its own standards, architectures, and software. This bipartition contrasts with the technological unification of the 1990-2010 decades, when American standards dominated global computing.
LineShine also illustrates the limits of unilateral technological sanctions. American restrictions have certainly slowed Chinese access to the most advanced technologies, but have stimulated Chinese domestic innovation. This dialectic repeats historical lessons: technological isolation forces indigenous innovation, sometimes with results superior to the original technologies.
China demonstrates that a determined technological power can free itself from critical dependencies in a single decade. This lesson resonates beyond high-performance computing, across all strategic sectors where the West thought it could maintain a definitive advantage through control of key technologies.