Intel Builds on vRAN Momentum with New AI Development Kit, Future Intel Xeon Processor

[Cristina_Rodriguez]

Intel Builds on vRAN Momentum with New AI

Development Kit, Future Intel Xeon Processor

-- Lab tests with AT&T, Deutsche Telekom, SK Telecom and Vodafone demonstrate AI’s potential to enhance network energy efficiency and service quality

-- Collaborations with Aira, Capgemini, Deepsig, Ericsson, Mavenir, Samsung, and Tech Mahindra highlight broad ecosystem commitment to advancing AI in radio networks

Artificial intelligence (AI) is among the tech industry’s most exciting new frontiers, and the mobile industry has started its AI journey. This includes investigating the benefits of AI for diverse use cases. With scale deployments of vRAN today, and continued momentum for steady growth in the years ahead, the industry can progressively begin to take advantage of its flexibility to add intelligence into the RAN. And let’s face it, the industry and Intel have seen some exciting milestones in recent months.

Last summer, we announced commercial availability of our 4th Gen Intel Xeon processors with Intel vRAN Boost (codenamed Sapphire Rapids EE), featuring fully integrated Layer 1 acceleration. Last year also saw Verizon announce the industry’s first data session over its commercial vRAN network with Samsung. Ericsson also made its first Cloud RAN call on the processor. In addition, Vodafone announced volume deployment of Open RAN in the UK with Samsung, noting that its earlier Open RAN deployment based on our 3rd Gen Intel Xeon processor already demonstrated performance exceeding that of their legacy equipment. Vodafone will deploy 4th Gen Intel Xeon processors with Intel vRAN Boost in the UK from April 2024. Additionally, AT&T and Ericsson unveiled plans to deploy and scale Open RAN services in collaboration with Intel. And in very recent news, Telus and Samsung announced plans to deploy Canada’s first commercial virtualized and open radio access network based on our 4th Gen processor.

These milestones not only highlight the mobile industry’s commitment to vRAN and Open RAN, but also show that Intel is fulfilling its commitment to product roadmap leadership to ensure industry success. And we’re not stopping there! Our future Intel Xeon platform, codenamed Granite Rapids–D with availability in 2025, will deliver significant gains in vRAN performance and power efficiency utilizing integrated vRAN Boost acceleration and AI acceleration, alongside other architectural and feature enhancements. Silicon is currently sampling.  Samsung has demonstrated a first-call at their research and development lab in Suwon, South Korea. Ericsson has also demonstrated a first-call validation in the Ericsson-Intel joint lab in Santa Clara, California. These achievements illustrate our processors’ easy gen-over-gen software portability. Additionally, we’re working with Dell Technologies, Hewlett Packard Enterprise, Lenovo, Mavenir, Red Hat, Wind River, and other leading ecosystem partners to ensure market readiness next year.

In parallel with these advances in Intel’s silicon roadmap, we’re also working with the mobile ecosystem to drive new RAN innovations with artificial intelligence that will help operators maximize the value of their existing Intel Xeon-based general-purpose hardware, which includes built-in AI acceleration.

We believe AI will play a pivotal role in optimizing performance and enabling intelligent management of resources within the RAN environment. Through advanced machine learning algorithms, AI analyzes relevant data generated by the network, identifying patterns and predicting trends to make real-time decisions that optimize radio resource allocation, increase energy efficiency, and improve user experience and overall network performance. With the software flexibility inherent to a virtualized RAN, the RAN can be continuously evolved with AI-based innovations on deployed systems.

Our multi-year research in this area shows that many RAN AI models are not as deep and wide as computer vision, but they may have more stringent latency requirements. These factors, in conjunction with inherent power constraints of the RAN, make it imperative that AI RAN workloads run efficiently on the same processor as the other RAN stack workloads.

Deployed Intel Xeon processors with built-in AI acceleration are capable of processing RAN inference workloads within the CPU. This integrated AI acceleration makes it possible to deploy optimized AI models and achieve great results on existing hardware already deployed in the RAN Intelligent Controller (RIC), the Centralized Unit (CU), or in the Distributed Unit (DU).

 To help operators capitalize on the AI opportunity using general-purpose processors, Intel is announcing early availability of the Intel vRAN AI Development Kit to select partners. The development kit allows independent software vendors (ISVs), telecom equipment manufacturers (TEMs), System Integrators (SIs), and network operators to quickly build intelligence in the RAN, without extensive investments in AI expertise.

 The development kit includes a suite of AI models optimized for vRAN use cases that are built on top of Intel-optimized libraries such as oneAPI, frameworks such as TensorFlow and PyTorch, and tools such as OpenVINO. These commonly used libraries, frameworks and tools make the AI development workflow efficient in generating code optimized for Intel Xeon processors. And leveraging these AI models, along with Intel Xeon processors’ built-in AI acceleration, telemetry and power management, operators can build intelligence into their network to dynamically reconfigure their network to conceivably save costs, maximize the value of their infrastructure investment, and support new revenue streams.

What’s more, the included training code will allow ecosystem partners to customize their AI solutions to specific network scenarios. The end-to-end reference architectures included in the development kit show how to integrate AI models with other RAN applications and components, such as with Intel FlexRAN reference software and RAN intelligent controllers (RICs).

The inference code for AI models in the development kit are deployable anywhere in the network, based on network operators’ specific network architecture, and they can be deployed as virtual network functions in cloud environments. They are optimized for performance on Intel Xeon Scalable processors and are portable across multiple generations of the Intel Xeon platform. This compatibility and extensibility ensures that software developed for current Intel Xeon processors can easily run on future generation processors with just a simple recompile.

The Intel vRAN AI Development Kit is informed by years of AI research and technology development, as well as collaborations with our broad ecosystem of partners. In this early release, the development kit includes AI models for key RAN use cases such as power savings, traffic steering, and network slice radio resource management. We will continue adding AI models for other use cases in the future.

 To provide the industry a sneak peek into the future of AI-enhanced radio networks, Intel is collaborating with AT&T, Deutsche Telekom, SK Telecom, and Vodafone to showcase the benefits AI can bring to vRAN. In testing with these operators using the Intel vRAN AI Development Kit, our latest vRAN processor, and Intel FlexRAN reference software, we‘ve seen results such as improved energy efficiency, enhanced user experience, and guaranteed service level agreement (SLA) targets.

The feedback has been very encouraging.

“Energy efficiency is a high priority for AT&T, as well as the broader Telecommunication industry, and AI has the potential to improve it significantly,” said Adam Loddeke, AVP – RAN Technology at AT&T. “Our joint tests with Intel suggest that AI-assisted CPU frequency scaling applied on a FlexRAN DU deployed on a 20-core 4th Gen Intel Xeon processor server with vRAN Boost can result in a 16% overall savings in server power. Applying this technique alone can lead to meaningful cost savings – for example, up to $4.5M in annual energy savings may be realized in a network with 100,000 DU servers. We are pleased to work with Intel on innovations including AI as we start Open RAN deployments in 2024 with 4th Gen Intel Xeon processors with vRAN Boost.”

“We are constantly exploring new technologies and solutions that can elevate our ability to provide customers a premium, high-quality user experience,” said Petr Ledl, Vice President of Network Trials and Integration Lab and Chief Architect of Access Disaggregation program, Deutsche Telekom. “It’s still early days for RAN AI technology, but our test collaboration with Intel shows that AI on the x86 compute platform, in combination with O-RAN architecture, can improve vO-DU beam selection at the cell edge. It’s great that Intel is helping to drive development of this emerging technology based on general-purpose processors.” 

“In working with Intel, we have been able to show how an inferencing algorithm working on Intel x86 architecture can be used to make better C-state power management decisions, leading to reduced network power consumption,” said Takki Yu, Vice President and Head of Infra Tech at SK Telecom.

“Optimizing software to take advantage of the latest silicon design advances is absolutely critical to meeting Open RAN’s total cost of ownership goals,” said Paco Martin, Group Head of Open RAN at Vodafone. “The vRAN AI Development Kit is a great asset for developing hardware-optimized software leveraging Intel Xeon processors’ built-in AI acceleration. Our joint tests of AI-assisted Network Slice Radio Resource Management showed significant potential for improved SLA guarantees, radio resource allocation, and spectrum efficiency versus without AI.”

 In addition to our operator collaborations, Intel is working across the broader vRAN ecosystem with partners such as Aira, Capgemini, Deepsig, Mavenir, Samsung, Tech Mahindra, and Ericsson via the companies’ joint Ericsson-Intel Tech Hub innovation lab to help drive the advancement and availability of AI in RAN.

As you can see, the transformation to open, virtualized RAN continues its steady march forward. And as it does, Intel will continue to deliver new innovations and solutions that enhance the performance, power efficiency, and value of general-purpose vRAN hardware.

MWC 2024 Attendees: Stop by Intel’s exhibit (Hall 3, #3E31) to see demonstrations of the Intel vRAN AI Development Kit, in collaboration with operator partners.

To learn more about the business benefits of AI in RAN, check out our ebook, “Unleashing the Potential of AI in the RAN.”

Performance varies by use, configuration, and other factors. Learn more on the Performance Index site.

Intel does not control or audit third-party data. You should consult other sources to evaluate accuracy.

Your costs and results may vary.

Intel technologies may require enabled hardware, software, or service activation.

© Intel Corporation. Intel, the Intel logo, and other Intel marks are trademarks of Intel Corporation or its subsidiaries. Other names and brands may be claimed as the property of others.