NVIDIA Joins Hands with Global Computer Industry to Build AI Factories and Data Centers

NVIDIA News Flash for Computex 2024:

Leading computer manufacturers to launch a range of Blackwell-powered systems featuring Grace CPUs, NVIDIA networking and infrastructure

Rich product portfolio covers cloud, dedicated systems, embedded and edge AI systems, etc.

Rich product configurations, from single GPU to multi-GPU, from x86 to Grace, from air cooling to liquid cooling, etc.

COMPUTEX — June 2, 2024 — NVIDIA and several of the world’s leading computer manufacturers today announced a range of systems based on the NVIDIA Blackwell architecture, equipped with Grace CPUs and NVIDIA networking and infrastructure to help companies build AI factories and data centers and drive a new round of generative AI breakthroughs.

In his COMPUTEX keynote, NVIDIA founder and CEO Jensen Huang announced that Aeon, ASUS, GIGABYTE, Inventec, Pegatron, QCT, Supermicro, Wistron, and Wiwynn will use NVIDIA GPUs and networking to build cloud, dedicated, embedded, and edge AI systems.

“A new round of industrial revolution has begun. Many companies and regions are working with NVIDIA to drive the transformation of traditional data centers worth trillions of dollars to accelerated computing and build a new type of data center - AI factory to produce a new commodity: artificial intelligence,” said Huang Renxun. “From server, network and infrastructure manufacturers to software developers, the entire industry is preparing to use Blackwell to accelerate AI-driven innovation in all fields.”

In order to meet the needs of various applications, the released products have rich configurations, covering from single GPU to multi-GPU, from x86 to Grace, from air cooling to liquid cooling technology, etc.

Additionally, to accelerate development of systems of varying sizes and configurations, the NVIDIA MGX™ modular reference design platform adds support for NVIDIA Blackwell products, including the new NVIDIA GB200 NVL2 platform, which is designed to deliver exceptional performance for mainstream large language model inference, retrieval-enhanced generation, and data processing.

GB200 NVL2 is ideal for emerging areas such as data analytics, where enterprises spend tens of billions of dollars each year. With the high-bandwidth memory performance of NVLink®-C2C interconnect technology and the proprietary decompression engine in the Blackwell architecture, data can be processed up to 18 times faster and 8 times more energy efficiently than using x86 CPUs.

Modular Reference Architecture for Accelerated Computing

To meet the diverse accelerated computing needs of data centers around the world, NVIDIA MGX provides computer manufacturers with a reference architecture that enables them to quickly and cost-effectively build more than 100 system design configurations.

制造商首先可根据其服务器机箱选择一种基本的系统架构，然后再根据不同的工作负载选择 GPU、DPU 和 CPU。目前，已有超过 25 家合作伙伴的 90 多套已发布或正在开发中的系统使用了 MGX 参考架构，较去年来自 6 家合作伙伴的 14 套系统有了显著的增加。通过采用 MGX，开发成本大幅降低，较之前最多降低了四分之三，开发时间缩短到仅六个月，较之前减少了三分之二。

AMD and Intel both support the MGX architecture and plan to create modular designs based on their own CPU host processors for the first time, including the next-generation AMD Turin platform and the P-core-based 6th Generation Intel® Xeon® processor (formerly Granite Rapids). Any server system manufacturer can use these reference designs to save development time and ensure consistency in design and performance.

NVIDIA's latest platform, GB200 NVL2, also uses MGX and Blackwell. In terms of horizontal expansion capabilities, a single-node design can support a variety of system configurations and network options, seamlessly integrating accelerated computing into existing data center infrastructure.

The GB200 NVL2 joins the Blackwell lineup, which also includes NVIDIA Blackwell Tensor Core GPUs, the GB200 Grace Blackwell superchip, and the GB200 NVL72.

Unified ecosystem

NVIDIA has a comprehensive partner ecosystem, including TSMC, the world's leading semiconductor manufacturer and NVIDIA foundry partner, and global electronics manufacturers that provide key components for building AI factories, including many innovations in manufacturing areas such as server racks, power supply and cooling solutions from companies such as Amphenol, AVC, Cooler Master, Colder Products Company (CPC), Danfoss, Delta and Lite-On Technology.

As a result, new data center infrastructure can now be rapidly developed and deployed to meet the needs of global enterprises, further accelerated by Blackwell technology, NVIDIA Quantum-2 or Quantum-X800 InfiniBand networking platforms, NVIDIA Spectrum™-X Ethernet platforms, and NVIDIA BlueField®-3 DPUs through server systems from leading system manufacturers such as Dell Technologies, HPE, and Lenovo.

Enterprise users can also use the NVIDIA AI Enterprise software platform, which includes the NVIDIA NIM™ inference microservice, to create and run production-grade generative AI applications.

Industry embraces Blackwell

Huang also announced in his COMPUTEX keynote that leading companies are rapidly adopting Blackwell to empower their businesses with AI.

Chang Gung Memorial Hospital, a leading medical center in Taiwan, plans to use the NVIDIA Blackwell computing platform to advance biomedical research and accelerate imaging and speech applications to improve clinical workflows and enhance patient outcomes.

Foxconn, one of the world’s largest electronics manufacturers, is planning to use NVIDIA Grace Blackwell to develop intelligent solutions for AI electric vehicle and robotics platforms, as well as growing language-based generative AI services to create more personalized experiences for its customers.

Partner Testimonials

“NVIDIA’s groundbreaking AI systems require advanced interconnect solutions, and Amphenol is pleased to provide key components,” said R. Adam Norwitt, President and CEO of Amphenol. “As a key partner in NVIDIA’s vast ecosystem, we are able to provide highly sophisticated and efficient interconnect products for the Blackwell acceleration platform, helping it deliver cutting-edge performance.”

AVC董事长兼首席执行官沈庆行表示：“AVC 在为 NVIDIA 的产品 - AI 硬件提供高效冷却方案中起着重要作用，如最新的 Grace Blackwell 超级芯片的冷却方案。随着 AI 模型和工作负载的日益增长，可靠的散热管理对于处理密集的 AI 计算至关重要，我们一直与 NVIDIA 携手共进。”

“ASUS is working with NVIDIA to take enterprise AI to new heights with our powerful server lineup at COMPUTEX,” said ASUS Chairman Jonney Shih. “With NVIDIA’s MGX and Blackwell platforms, we will tailor data center solutions to handle users’ training, inference, data analytics and HPC workloads.”

“CPC’s innovative, purpose-built connector technology makes it easier and more reliable to connect liquid-cooled NVIDIA GPU-based AI systems,” said Janel Wittmayer, president of CPC, a Dover Corporation company. “We share NVIDIA’s passion for performance and quality, and CPC has the capabilities and expertise to provide key technology components to meet NVIDIA’s significantly increased and ongoing needs. Our connectors are key to ensuring consistent temperature-sensitive products, which is especially important when AI systems are running compute-intensive tasks. We are excited to be part of the NVIDIA ecosystem, which brings our technology to new applications.”

“As the demand for accelerated computing continues to grow, companies using advanced accelerators also need solutions that can effectively meet energy standards,” said Andy Lin, CEO of Cooler Master. “As a pioneer in thermal management solutions, Cooler Master is helping the NVIDIA Blackwell platform reach its full potential and deliver exceptional performance to users.”

“Danfoss’ commitment to innovative, high-performance Quick Disconnect technology and fluid power design enables our connectivity products to play a key role in enabling efficient, reliable and secure data center operations,” said Kim Fausing, CEO of Danfoss. “As a key member of the NVIDIA AI ecosystem, our collaboration enables data centers to meet the growing demands of AI while minimizing environmental impact.”

“The ubiquitous demand for computing power has ushered in a new era of performance acceleration,” said Ping Cheng, chairman and CEO of Delta. “With our advanced cooling and power systems, Delta has developed innovative solutions that enable the NVIDIA Blackwell platform to achieve the highest performance levels while maintaining energy and thermal efficiency.”

“GIGABYTE has been collaborating with NVIDIA for nearly three decades, and we have been fully supportive of NVIDIA’s technology platforms, including GPUs, CPUs, DPUs, and high-speed networking,” said Yi-tai Li, vice president and general manager of GIGABYTE. “To enable enterprises to achieve higher performance and energy efficiency for compute-intensive workloads, we are launching a series of systems powered by the Blackwell architecture.”

“Generative AI is profoundly changing every industry, and Foxconn is well positioned to provide cutting-edge solutions to meet the demand for diverse and high-performance computing,” said Young-Wei Liu, chairman and CEO of Hon Hai Technology Group. “We not only use the latest Blackwell platform in our own servers, but also provide key components to NVIDIA to help customers get products to market faster.”

“For nearly half a century, Inventec has been designing and manufacturing electronic products and components, which is the foundation of our company,” said Inventec President Tsai Chih-an. “With the NVIDIA MGX rackmount solution powered by the NVIDIA Grace Blackwell superchip, we are helping our customers reach new heights of AI capabilities and performance.”

“Power management and cooling solutions are at the heart of building greener, more sustainable data centers,” said Sen-Pin Chiu, president of Lite-On Technology. “With the launch of the NVIDIA Blackwell platform, Lite-On is introducing multiple liquid cooling solutions to help NVIDIA’s partners build more efficient and environmentally friendly data centers of the future.”