#Taiwan #Taipei #Compal Electronics #NVIDIA HGX #NVIDIA Vera Rubin

Compal Electronics Showcases High-Density NVIDIA HGX® Rubin NVL8 Integrated Solution at GTC 2026

On March 16, 2026, during the NVIDIA GTC 2026 event, Compal Electronics made a notable entrance by presenting their latest high-density AI server solution, which is built upon the innovative NVIDIA HGX Rubin NVL8 architecture. This development aligns seamlessly with NVIDIA's vision of integrated artificial intelligence supercomputing infrastructure as exemplified by their new architecture, dubbed the Vera Rubin framework.

Advanced Architectural Integration

The NVIDIA Vera Rubin architecture stands at the forefront of modern computing technology, incorporating a suite of advanced hardware components. It integrates the NVIDIA Vera CPU, NVIDIA Rubin GPU, NVLink 6 Switch, BlueField-4 DPU, Spectrum-X Ethernet, and ConnectX-9 SuperNIC into a cohesive computing environment. This comprehensive design enables data centers to maximize their operational efficiencies, delivering a remarkable 260 TB/s of total bandwidth across the system with a GPU-specific capacity of 3.6 TB/s for all-to-all communications. Such specifications are particularly beneficial for managing large-scale AI training and inference workloads.

The SG231-2-L1 Server

Among the standout products presented by Compal was the SG231-2-L1 server, engineered for high density and advanced capabilities. This server features:

• - High-Density Accelerated Architecture: By housing eight NVIDIA Rubin GPUs within a compact 2U chassis, it inherently maximizes compute capability while optimizing space.
• - Incredible Inference Performance: Capable of delivering up to 400 petaFLOPS (NVFP4), it supports extensive functionalities, including large language model (LLM) development, generative AI tasks, and high-performance computing (HPC) workloads.
• - Superior GPU Interconnects: Powered by NVIDIA NVLink 6, it achieves up to 28.8 TB/s of bandwidth between GPUs, enhancing multi-GPU scalability.
• - Robust Memory Architecture: With support for up to 2.3TB of GPU memory and 176 TB/s of memory bandwidth, it is adept at handling memory-heavy AI applications.
• - Liquid Cooling Efficiency: The design facilitates approximately 24kW of system power through a streamlined liquid cooling system, ensuring consistent performance under heavy loads.
• - Seamless Rack-Level Deployment: Designed for easy integration into AI-focused rack systems, it can adapt well from single-node setups to comprehensive data center deployments.

Commitment to Engineering Excellence

Compal also showcased a new module featuring the NVIDIA Vera CPU in their booth, underscoring their engineering readiness and manufacturing capabilities in accordance with the NVIDIA Vera Rubin framework. In an evolving landscape where heterogeneous computing is pivotal, the synchronization of CPU and GPU design is critical for realizing efficiency in data center operations.

In addition to these offerings, Compal presented support for the NVIDIA RTX PRO 4500 Blackwell Server Edition, which touts 32GB GDDR7 memory and up to 800 GB/s bandwidth, tailored for effective acceleration of AI inference and visual computing endeavors.

Insight from Leadership

Alan Chang, Vice President of Compal’s Infrastructure Systems Business Group, articulated the current trajectory of competition within AI infrastructure. He emphasized that it has evolved beyond mere performance comparisons of single nodes, focusing now on the overall efficiency of deployments and their long-term scalability. As the NVIDIA HGX Rubin NVL8 platform continues to develop, he asserted the necessity for data center architectures to adapt promptly, investing not only in computational capacity but in the entire infrastructural framework to enable sustainable models for the next generation of AI workloads.

Conclusion

As workloads related to AI grow in complexity and scale, the competitive edge of data centers increasingly hinges on their ecosystem integration and architectural synergy. Compal’s presentation at GTC 2026 not only demonstrated their formidable capabilities in high-density GPU systems and innovative cooling solutions but also reinforced their strategic role within NVIDIA's technological ecosystem.

About Compal Electronics

Founded in 1984, Compal Electronics has emerged as a premier manufacturer in the notebook and smart device sector, building strong brand equity through collaboration across various industries. Their innovative product designs have garnered numerous international accolades. In 2025, Compal was honored by CommonWealth Magazine as one of Taiwan's top manufacturers and consistently ranks among the esteemed Forbes Global 2000 companies. Compal is also actively venturing into new sectors such as cloud servers, automotive electronics, and smart healthcare, leveraging its integrated RD and manufacturing strengths to deliver relevant and effective solutions. For more details, visit Compal's official website.