공시 • Mar 17
Wiwynn Showcases NVIDIA Vera Rubin NVL72 AI Factory Infrastructure At NVIDIA GTC 2026 Wiwynn is showcasing the latest NVIDIA-powered AI solutions, developed in collaboration with Wistron, at GTC 2026 (Booth #1121). From board-level innovation to rack-scale integration and validation, Wiwynn's end-to-end expertise in accelerated computing, storage, and liquid cooling delivers future-ready AI factories with breakthrough performance, exceptional energy efficiency, and faster time-to-value. Technology highlights NVIDIA Vera Rubin NVL72: Wiwynn and Wistron are among the first in line with readiness for the fully liquid-cooled, rack-scale platform unifying 72 NVIDIA Rubin GPUs and 36 NVIDIA Vera CPUs. Optimized for frontier AI model training, inference and reasoning, the platform delivers up to 10X higher performance per watt through extreme co-design —delivering breakthrough performance and efficiency for AI factories. NVIDIA HGX Rubin NVL8: A top-tier modular accelerated computing platform—a fanless, high-density 2U system with eight NVIDIA Rubin GPUs and 100% liquid cooling to achieve remarkable PUE. Scales from 8 to 16 systems per rack over NVIDIA Spectrum-X Ethernet or NVIDIA Quantum-X800 InfiniBand, adapting to evolving compute needs at scale. NVIDIA RTX PRO Server: A compact 2U Arm-based platform that pairs dual NVIDIA Vera CPUs with two NVIDIA RTX PRO 4500 Blackwell Server Edition GPUs (32GB GDDR7 each) to power neural rendering and AI-driven design. Storage-Next: Part of NVIDIA's Storage-Next initiative, the GPU-initiated storage architecture leverages NVIDIA SCADA to directly orchestrate I/O across a 96-drive NVMe array via the GPU, delivering ultra-high IOPS, sub-millisecond tail latencies and petabyte-class density for GNN, LLM inference, and RAG. Direct liquid cooling, integrated per-drive telemetry and multi-zone leak detection enable high efficiency and hot-serviceability. Session at GTC 2026 Topic: Build an AI-Ready Data Center: Practical Insights from YTL, Wiwynn, and NVIDIA March 19, 2026, 10:00–10:40 a.m. PT A deep dive into L11–L12 co-engineering—from rack delivery to cluster acceptance—accelerating deployments and optimizing thermals, power, and serviceability. 공시 • Mar 02
Wiwynn Corporation, Annual General Meeting, May 25, 2026 Wiwynn Corporation, Annual General Meeting, May 25, 2026. Location: 2 floor building. c no,94, sec.1 hsin t`ai 5th rd., sijhih district, new taipei city Taiwan 공시 • Nov 18
Wiwynn Delivers Best MLPerf Training v5.1 (Closed) Llama 2 70B LoRA Results at YTL Malaysia Data Center Wiwynn announced best results in the MLPerf®? Training v5.1 Llama 2 70B LoRA benchmark (Closed division), earning best performance on both 1x and 8x NVIDIA GB200 NVL72 configurations. The submissions were executed on production systems already deployed by YTL AI Cloud, spanning a 1-rack NVIDIA GB200 NVL72 (with 72 NVIDIA BlackwellGPUs) and an 8-rack NVIDIA GB200NVL72 integrating 576 GPUs-- demonstrating leadership from single-rack to multi-rack scale. The verified MLPerf scores highlight Wiwynn's strengths in system design, manufacturing, liquid cooling, multi-rack integration, and hardware/software co-optimization, combined with YTL's excellence in AI infrastructure integration and operations. Together, the partners demonstrate how close collaboration between system manufacturers and data center operators can deliver production-grade, benchmark-verified AI training performance. 공시 • Oct 14
Wiwynn Debuts Double-Wide Rack Architecture for Next-Generation AI at OCP Global Summit 2025 Wiwynn unveiled its double-wide rack architecture to address the industry shift toward larger, higher-power AI chips at the OCP Global Summit 2025 (Booth B13, San Jose Convention Center). The architecture brings together an HVDC power rack and busbar, advanced liquid cooling, and a signal-integ integrity-optimized compute/switch layout to tackle data transmission, power delivery, thermal, and mechanical challenges in high-density computing--propelling the future of AI and turning ideas into impact. Double-Wide Rack: Doubles the width of traditional rack to accommodate growing chip/module footprints as compute demand rises. The double-wide compute tray places four large AI accelerators side by side, supporting chips scaling to 9.5-reticle with 12x HBM and substrates up to 120 x 150 mm. The design reserves on-board high-speed interconnects and signal integrity while reducing cable routing and signal loss. It is co-designed with the switch tray configuration to ensure high-speed signal quality in scale-up scenarios. The compute and switch trays use an advanced mechanical design with a reinforced chassis to prevent sagging. The new architecture is engineered to accommodate next-generation accelerators such as AMD Instinct™? MI400 Series GPUs, enabling rack scale performance for frontier AI training and inference. HVDC power: To meet surging rack-level power density, the architecture adopts HVDC power delivery. A HVDC power rack converts AC to +-400/800 VDC and distributes it to IT racks next to it via vertical and horizontal busbars and connectors, significantly reducing distribution losses while boosting energy efficiency and scalability. With ecosystem partners, Wiwynn offers +-400/800 V DC and 50 VDC busbar options and the required power delivery boards (PDBs) to flexibly support data center needs. Liquid-cooled Busbar: Engineered for high-current busbars, integrating liquid cooling to control temperature and reliability, improving safety and stability. Supports Wiwynn's in-house design with flexible inlet/outlet configurations to fit diverse data center liquid-cooling loops, while also supporting multi-vendor solutions. Advanced Liquid Cooling Solutions: Targeting 4 kW class chip thermal capability, the portfolio includes high-efficiency cold plates, AALC, in-row CDUs, and liquid-cooling management. It balances reliability, serviceability, and scalability to deliver efficient, sustainable cooling for AI/HPC workloads. 
