AMD on AI and 5G Innovations at IMC 2024

In an interview with AMD on the sidelines of the India Mobile Congress (IMC) 2024, we spoke with Gilles Garcia, Senior Director of Business Lead Data Centre and Communications Group, AMD. The conversation delved into AMD's cutting-edge AI portfolio, its role in shaping the future of 5G and 6G networks, and how the company is driving innovation in edge computing and telecommunications. 

With a focus on AI integration across multiple sectors, Gilles shared insights into AMD's recent developments, collaborations with Indian companies, and the strategic impact of its solutions on India's rapidly evolving digital ecosystem. This interview sheds light on how AMD is positioning itself as a key player in the next wave of digital transformation.

Excerpts:

DQ: How does AMD’s AI portfolio stand out in the competitive AI market?

Gilles Garcia: Our AI portfolio spans from data centers to the edge, covering a wide range of products. We do not force our customers to use a GPU on the edge if it’s unnecessary, as some edge environments don’t need that level of processing power, or they cannot afford to consume that much energy. Our broad portfolio allows us to offer the right solution based on the customer’s needs, whether they are working with inference or training models. This flexibility is a key differentiator for AMD in the AI space.

DQ: How is AMD integrating software to unify its AI infrastructure across different hardware?

Gilles Garcia: For the last two years, we have made significant investments to unify our AI infrastructure from a software standpoint. Our goal is to ensure that, whether you are using a GPU, CPU, FPGA, or adaptive SoC, the software experience is consistent. This means that users—whether they are data scientists or algorithm developers—can use the same software ecosystem regardless of the hardware being employed.

This approach ensures that our customers can seamlessly transition between different types of hardware while still working within a unified software framework. This is particularly important given our extensive portfolio, which includes products for everything from large data centers to edge IoT deployments. Our software ecosystem supports these diverse environments, making it easier for businesses to adopt AI technology tailored to their specific needs.

DQ: As AMD is participating in the India Mobile Congress, can you talk about how 5G and AI are being integrated within your technologies?

Gilles Garcia: I lead the wired and wireless division at AMD, and we have a 6G initiative within our group. We know that AI will play a significant role in 6G, so we’re already working with customers who are trialing their algorithms for 6G networks, focusing on AI-based improvements. These improvements can help the network become more efficient in power management, user mobility, and other applications.

At our booth today, we’re demonstrating two types of AI applications: one related to 5G and 6G AI, and another showcasing how AI can be applied on top of 5G or 6G networks. We currently have the technology to integrate AI into the far edge and radio, and this technology—called AI Engine—can be embedded in radios or base stations. This allows customers to prepare for 6G by integrating AI into network operations like baseband processing and radio functions, improving efficiency with spectrum management and interference cancellation.

DQ: We often hear about generative AI and large language models (LLM). How is AMD addressing broader AI use cases beyond these popular technologies?

Gilles Garcia: While generative AI and large language models (LLMs) are the most popular aspects of AI today, AI is much bigger than that. At AMD, we certainly support generative AI and LLMs, but we also work on AI across many different use cases. For example, AI is transforming networks, making them smarter and more efficient through optimization. AI plays a role in IoT at the edge, where it can automate tasks and improve decision-making in real time.

One area where we see significant growth for AI is in security. Security threats are constantly evolving, and AI is instrumental in learning from these threats and creating predictive models. For instance, we’ve already demonstrated AI-based DDoS detection, using machine learning to identify and prevent massive server attacks in real time.

DQ: Beyond GPUs, how does AMD envision its Edge AI portfolio, and what are your thoughts on the high cost and shortage of GPUs in the market?

Gilles Garcia: When it comes to AI at the edge, we’re seeing three main types of applications: those that run on CPUs, those that need accelerators (like PCIe cards or AI engines in switches and routers), and specialized use cases like gaming or AI-enhanced features in laptops. GPUs are generally more suitable for data centers and large-scale computing environments due to their power consumption, which makes them less viable for edge use cases.

That’s why, for edge AI, we focus on solutions that don’t require the heavy processing power of GPUs. Instead, we offer CPUs, embedded AI technologies, FPGAs, SoCs, and AI PCs, which provide sufficient processing power without the high energy consumption of GPUs. Of course, some edge data centers are large enough to support GPUs, but for most edge applications, our alternative solutions are more appropriate.

DQ: Could you explain AMD’s Doctor as a Service initiative, and how it is being implemented in India?

Gilles Garcia: One of the demos we’re showcasing at our booth is the eSanjeevani application, which is a virtual ophthalmology solution powered by AI. The government of India runs this application on its telemedicine platform, and we’ve partnered with them to demonstrate how our AI models can be used for healthcare services.

The solution runs on our latest Turin Gen 5 CPUs, hosted on a Make in India server produced by our partner Velankani in Bangalore. It supports multiple Indian languages and allows users to consult virtually with an ophthalmologist. This is particularly important for rural areas where access to healthcare is limited. The system can conduct virtual eye exams, provide diagnoses, and even prescribe medication, all while using AI to enhance the accuracy and efficiency of the process.

DQ: You mentioned a partnership with a Bangalore-based startup. Are there other significant collaborations that AMD has in India?

Gilles Garcia: Yes, we have many collaborations in India, and we’re showcasing several Make in India products at our booth. Some of the radio units on display come from companies like L&T, Tejas Networks, Bharat Broadband Network Limited (BBNL), NEC India, CDOT, and Lekha Wireless. These products represent a broad range of AMD-powered technologies in the telecom space.

We’re also showcasing servers from NetWeb India, Velankani, and HP, all of which are designed and manufactured in India. One key differentiator is that these servers are not just assembled here but also feature motherboards designed in India, which is rare in the industry. This underscores our commitment to building a robust tech ecosystem in India, supporting local manufacturing and innovation.

DQ: How is AMD contributing to the integration of AI in 5G, and what kind of technological developments are you working on in this space?

Gilles Garcia: We’ve already deployed AI-enabled radios in the field, working with major players like Samsung and NEC. These companies are using AMD technology in their 5G radios, with integrated AI engines that enhance functions like beamforming. This allows for more precise and efficient network performance in 5G environments.

The AI engines integrated into these radios are based on our 7-nanometer Versal technology. This technology is also used in our 4th and 5th generation CPUs, which can run AI workloads in base stations. We’re actively working on future technologies, but the products we have today are already helping our customers deploy AI in their 5G networks. These solutions will also be crucial as we move toward 5G Advanced and eventually 6G.

DQ: Is AMD launching any new products at this event?

Gilles Garcia: We didn’t plan specific product launches for IMC 2024, but we did make several major announcements just two days ago during our Advancing AI event. These included the launch of our 5th generation CPUs, the MI250X GPU, and previews of future GPU technologies. We also announced new networking products designed for GPU connectivity.

At IMC, we’re showcasing some of these innovations, including live demos of the Turing 5th generation processors and the Ryzen AI processors. While we aren’t officially launching new products here, we’re leveraging the event to demonstrate how our latest technology can support the telecom sector.