Wiwynn Achieves Record-Breaking MLPerf® Training Results with Llama 2 70B at YTL Malaysia

Wiwynn Achieves Unprecedented MLPerf® Training Results in Malaysia

In an impressive showcase of technological capabilities, Wiwynn, a frontrunner in cloud IT infrastructure solutions, has announced its remarkable achievements in the MLPerf® Training v5.1 benchmark. The company garnered top scores in the Closed division using the Llama 2 70B LoRA model, asserting its technological prowess by successfully deploying systems at the YTL Malaysia Data Center.

Utilizing both single and multi-rack setups of NVIDIA GB200 NVL72 configurations, Wiwynn delivered exceptional results. The benchmarks featured systems comprising 1x NVIDIA GB200 NVL72, equipped with 72 NVIDIA Blackwell GPUs, and a scalable 8x configuration that harnessed 576 GPUs. These achievements establish a new benchmark for AI training performance across the Asia-Pacific region.

"Wiwynn is dedicated to optimizing workloads for real-world deployment," stated William Lin, the President and CEO of Wiwynn. Underpinning these results is a robust partnership with YTL AI Cloud, the collaboration emphasizing a shared commitment to advancing AI infrastructure. This partnership showcases significant Industry advancements, from system integration to software optimization, further enabling the capabilities of high-performance GPU clusters.

Philip Lin, CEO of YTL AI Cloud, echoed these sentiments, stating that their focus is to create the region's most sophisticated AI infrastructure. The partnership with Wiwynn is poised to transform the AI landscape in APAC, offering premier hubs for training and inferencing on complex models like Llama 2 70B.

The verified results from MLPerf reinforce Wiwynn's design and manufacturing efficiencies, as well as its innovative approaches to system design, including liquid cooling and multi-rack integration. This collaboration exemplifies how strategic partnerships between hardware manufacturers and data center operators can lead to significant performance improvements in AI workloads.

Wiwynn’s success in the MLPerf® Training v5.1 showcases a commitment to enhancing the AI ecosystem, benefiting overall accuracy, speed, and efficiency in AI systems. Each component of their infrastructure, including advanced GPUs in liquid-cooled high-density racks, contributes to the region's increasing demand for AI solutions optimized for both energy efficiency and performance.

The combination of Wiwynn's high-quality IT infrastructure and YTL's advanced AI operations creates an exemplary model in the AI domain. As YTL AI Cloud continues to position itself as a critical AI hub, it paves the way for global customers to efficiently train and deploy frontier AI models. Such innovations are essential for nurturing sustainable and scalable AI technologies in today's ever-expanding digital landscape.

With MLPerf® Training v5.1 results validated by MLCommons Association, Wiwynn is proud to contribute to the constant evolution of AI technology, demonstrating the transformative impact of collaborative efforts in the industry. By continually pushing the limits of technological advancements, Wiwynn and YTL AI Cloud are setting a new standard for what can be achieved in AI training, ensuring that they stay at the forefront of this rapidly advancing field.

These achievements reflect not only Wiwynn’s dedication to excellence but also the critical nature of effective teamwork in realizing innovative solutions that will shape the future of AI technology in the Asia-Pacific region. The synergy between innovation and operational efficiency is key to unlocking the full potential of AI systems, and Wiwynn is exemplifying this through its exemplary benchmarks and partnerships.