UALink Alliance: A New Open GPU Communication Standards

In the rapidly evolving field of technology, open standards and technological innovation are key to driving industry growth. However, monopolistic practices in the market can hinder the healthy development of the industry. In response to this challenge, several tech giants have taken action to jointly develop new open standards to promote fair competition and technological innovation.

 UALink Alliance.jpg

UALink: A New Attempt at Open Standards

Recently, tech giants such as Google, Meta, Microsoft, AMD, Intel, Broadcom, Cisco, and HP announced the formation of an alliance called the Ultra Accelerator Link (UALink). The goal of this alliance is to develop a new industry open standard for artificial intelligence data center GPU network communication systems. Through this initiative, they aim to break the monopoly in the market and promote competition and innovation within the industry.


The establishment of the UALink alliance is based on a profound understanding of the current market situation. Existing GPU network communication technologies in the market, such as NVIDIA's NVLink, perform exceptionally well in terms of performance, but their proprietary nature limits the development space for other manufacturers. NVLink technology significantly improves data transfer efficiency and processing performance between GPUs and CPUs through high-bandwidth direct connections. However, the closed nature of this technology makes it difficult for other manufacturers to compete.


Technical Advantages of UALink

The UALink 1.0 version planned by the UALink alliance aims to achieve direct data transfer between AMD's Instinct GPUs and Intel's Gaudi processors. This design not only significantly enhances the performance and efficiency of AI computing but also demonstrates the determination of alliance members to challenge market leaders with its high-speed I/O communication design and protocol.


Potential Impact of UALink

The launch of UALink will provide more opportunities for other manufacturers to compete in the market. As AI computing demands continue to grow, a robust, low-latency, and highly scalable network is crucial for improving computing performance. The formulation of UALink and industry standards will help create an open high-performance environment for AI workloads, promoting the continuous progress of the industry.

UALink Alliance.jpg

Future Prospects of UALink

Although the UALink alliance has not excluded NVIDIA from its cooperation scope, considering NVIDIA's business interests, it is expected that they will not actively join the UALink alliance. However, the emergence of UALink will undoubtedly weaken NVIDIA's market advantage. The first batch of UALink products is expected to be launched in the coming years. Despite facing many challenges, the UALink alliance firmly believes that this step is a necessary one.



The establishment of the UALink alliance is not only a challenge to the existing market monopoly but also an advocacy for open standards and technological innovation. Through the efforts of this alliance, we look forward to a more fair and innovative technology industry. This is not only a self-renewal for the tech giants themselves but also a positive contribution to the future development of the entire industry.


Recommended IC Models

In the field of exploring electronics and high technology, especially in the field of integrated circuits (ICs), some high-performance and efficient models are highly respected for their excellent performance. Here are some of the popular models that are currently on the market, and they come from well-known semiconductor companies Intel, ADI, and Broadcom.


l Intel N82C55A2 - Known for its 44-pin PLCC package, this CMOS programmable peripheral interface is a specialized interface in the driver and interface category to provide connectivity between the host and peripherals.

l Broadcom BCM56990B0KFLGG - This is a network switching chip manufactured in a cutting-edge 28nm process that features high bandwidth, low power consumption and high reliability in high-speed data transfer and switching applications such as data centers, enterprise networks and cloud computing.

l Analog Devices Inc. ADSP-2185BSTZ-133 - Designed for digital signal processing (DSP) and other high-speed numerical processing applications, this single-chip microcomputer is suitable for a variety of applications requiring high performance processing power. This includes areas such as audio processing, communications systems, medical imaging and industrial control systems.


For design engineers and electronic enthusiasts, Conevo Electronics provides a broad platform, here, you can find all kinds of inspired innovation popular electronic components. Whatever your project requirements, Conevo can provide the necessary support and resources, help you to turn ideas into reality.



Contact Information