Since the official commercial use of 5G, the industry has used the word "marathon" to describe its development from time to time. This technology, which is characterized by its fast nature, has been named the attribute of "long-term continuous evolution". With the recent freezing of Release 18, the first standard version of 5G Advanced, this "marathon" of technology evolution has officially run into the second half. Compared with the excitement and high-speed propulsion in the starting stage, in order to avoid "hitting the wall" and maintain the speed advantage of 5G in the second half of the race, it needs more involvement of the brain, thinking about the allocation of physical energy and formulating strategies. Today, many of the technological achievements of 5G Advanced have shown the characteristics of fast and intelligent.
快中有智的5G Advanced
As a long-distance race, maintaining the pace and rhythm while not blindly "running stupidly" is the key to achieving the ultimate victory. With the verification and application of 5G Advanced in multiple scenarios in China, this technology has demonstrated the ability to continuously improve the speed experience.
During the F1 China Grand Prix, which focused on speed competition, Qualcomm and Shanghai Unicom completed the 5G Advanced high- and low-frequency collaborative continuous networking, achieving the milestone of breaking 5Gbps for the first time in the continuous network coverage experience.
On the eve of MWC Shanghai, China Unicom Research Institute and Qualcomm completed the first on-site verification of 5G-A high and low frequency NR-CA new technology, using the 800MHz bandwidth of the high-frequency band and the 100MHz bandwidth of the 3.5GHz low-frequency band for carrier aggregation, and achieving a single-user downlink peak rate of more than 8.5Gbps on the smartphone form factor terminal equipped with Snapdragon X80 modem and RF system. New 5G-A high-bandwidth services, such as glasses-free 3D, provide an immersive experience.
From the above verification, it is not difficult to find that the Snapdragon X80 has become a key part of the steady breakthrough of speed on the terminal side, and the hardware specifications and characteristics of this product precisely reflect the characteristics of "fast and intelligent". In addition to maintaining a peak rate of up to 10Gbps, the Snapdragon X80 combines AI with unmatched spectrum flexibility, power efficiency and performance, integrating a dedicated 5G AI processor and 5G Advanced-ready architecture, achieving a number of world-first milestones, including the first integration of NB-NTN satellite communications in a 5G modem, the first support for 6Rx for smartphones, The first downlink six-carrier aggregation and the first support for AI-enabled mmWave extended range communication for CPEs.
"We believe that while 5G is unleashing the potential of applications, 5G Advanced is also ushering in the era of 'wireless AI'. The convergence of wireless communication technology and AI brings a unique opportunity for the industry to transform the future of wireless technology. Meng Feng, chairman of Qualcomm China, specifically mentioned at the GTI International Industry Conference during MWC Shanghai that the Snapdragon X80 is Qualcomm's latest achievement in integrating cutting-edge AI and wireless connectivity technology.
With the explosion of generative AI and the emergence of AI demand in a wide range of industries, the huge potential unleashed by the convergence of wireless technology and AI is bringing unprecedented opportunities for change to the industry.
The era of "wireless AI" that complements each other
It is undeniable that many people's current use and cognition of generative AI are still at the level of cloud AI, believing that whether it is through a client or a web page, it needs to rely on the cloud thousands of miles away to get the desired results.
"Qualcomm thinks the same way about AI as it does about computing: a lot of computing happens in the cloud, while the rest happens on the end side. We think the same is true for AI. There will be AI running in the cloud in the future, and we've seen a lot of examples of that. The next step is to bring the use cases of AI and generative AI to the device side. Akash Palkhiwala, CFO and COO of Qualcomm, talked about the future of AI during MWC Shanghai, and he believes that on-device AI will fully demonstrate the advantages of low latency and privacy protection, and that computing power on the device side will be more economical than that of cloud-based generative AI, which is very resource-intensive.
Although the industry has become accustomed to comparing cloud AI to the "brain", there is a world of difference between cloud AI and device-side AI in terms of energy consumption and latency. Compared with cloud-based AI, generative AI that can run on the device side with very low power consumption is more like the "brain" on the shoulders of all things—with the advantages of low power consumption and efficient intelligence.
"The key to driving generative AI on the device side is the ability to compute at very low power consumption. Everything we do on our mobile is tied to the battery. Today, with high expectations for battery life, we can't afford to sacrifice battery life for the sake of certain use cases," says Akash Palkhiwal. In this regard, NPU is a key technology module newly designed by Qualcomm for AI, which can drive AI applications with extremely low power consumption.
Taking the third-generation Snapdragon 8, which is currently being adopted by a large number of flagship smartphones, as an example, Qualcomm has specially equipped the Hexagon NPU with a dedicated accelerator power supply in order to achieve better power consumption control, and at the same time has carried out microarchitecture upgrades to improve performance. During MWC Barcelona in February this year, Qualcomm ran a large language and visual assistant model technology demonstration based on image and text input and more than 7 billion parameters on the Android platform for the first time through a device equipped with the third-generation Snapdragon 8 reference design, once again proving the potential of AI on the device side.
The development of on-device AI has made the future of hybrid AI possible, which means that the device side will be able to run some AI workloads, while the cloud will be responsible for the work with higher computing requirements. The need for private data and responsiveness can stay on-premises, allowing the cloud to complement each other. Qualcomm believes that hybrid AI is the future of AI, and running AI in the cloud, edge cloud, and device side together will help achieve more powerful, efficient, and ubiquitous applications.
This also means that hybrid AI will have extremely high requirements for connectivity, such as extremely low latency, extremely high bandwidth, and connection reliability. Just as the Snapdragon X80 played an important role in a series of 5G Advanced verification tests, when connectivity technology has become one of the core driving forces to support the large-scale expansion of generative AI, Qualcomm, which has been deeply involved in the field of wireless and device-side AI for many years, has become more and more prominent in this regard.
"Qualcomm believes that 5G + AI will truly revolutionize all industries, not only the mobile communications industry we are in now, but also PCs and many other industries." Hou Mingjuan, Global Vice President of Qualcomm, spoke on "5G Business: Where is the Next Wave of Growth?" during MWC Shanghai. In the topic discussion, "5G + AI" was chosen as the answer.
The relationship between the two is even more than 1+1>2, and Meng Feng expressed at the beginning of his speech that "the integration and innovation of 5G and AI will release the 'multiplier effect' for the development of the digital economy". In the process of two-way empowerment of 5G and AI, many industries will usher in new upgrades.
Unleashing the "multiplier effect" of digital intelligence in the future
Qualcomm formalized its business diversification strategy many years ago, with a focus on extending advanced technologies to multiple types of devices. Now, as Qualcomm President and CEO Ammon recently said at COMPUTEX, Qualcomm is transforming from a communications company to a connected computing company that makes intelligent computing ubiquitous. In order to enable AI to empower thousands of industries, Qualcomm is actively exploring the role of AI in various terminals, including mobile phones, PCs, automobiles, industrial equipment, etc.
In addition to providing leading-edge AI hardware, Qualcomm can also provide a flexible cross-device, cross-OS, and cross-platform Qualcomm AI software stack. "Our partners and developers only need to complete the deployment of the model on one platform of Qualcomm, and it is very convenient to migrate to other product lines of Qualcomm." Wan Weixing, head of Qualcomm's AI product technology in China, said that Qualcomm's AI software stack has been able to empower the vast majority of platforms in almost all of Qualcomm's product lines.
With excellent connectivity and the ability to run AI on the device side anytime, anywhere, the next question that the industry is concerned about is what kind of digital and intelligent future can "5G + AI" open up for us.
In this regard, Hou Mingjuan used an AI table tennis training robot as a good example. Powered by Qualcomm technology, the robot can help coaches tailor training plans for each student and help participants who enjoy the national sport to further improve their skills, demonstrating the potential of AI to transform many industries, including education and sports.
The transformation XR is facing is also a great example of this. At the beginning of this year, Qualcomm joined hands with China Mobile, ZTE and the popular Qitian to complete the industry's first 5G Advanced multi-concurrency large-space XR competitive game service pilot, using terminals equipped with Qualcomm solutions, in a large space of nearly 1,000 square meters, when 12 XR services are connected at the same time, the picture is clear and smooth without stuttering, and the average air interface delay is less than 10ms, bringing a multi-concurrent and large-space XR competitive game experience without backpacks and cables.
Through the above use cases, it is not difficult to see that 5G Advanced has made XR say goodbye to heavy equipment and backpacks, and AI is the next key to making XR experience leap. AI will promote the evolution of the interactive interface of many terminal devices, and XR may become the most obvious industry, such devices in the past are often cumbersome in the initial wearing and debugging and daily control, in the future, through AI assistants, users can perform more tasks that could not be done in the past, get rid of the limitations of size, display and other aspects; Users can also talk directly to the AI assistant and handle and perform more tasks through the cloud.
As an annual telecommunications industry event, many people would never have imagined that in the various pavilions of this year's MWC Shanghai, cars have become mobile intelligent terminals that can be seen everywhere. Meng said: "'AI on the car' is becoming a reality. "At present, many automobile manufacturers such as Ideal, Xpeng, and Jiyue have released their vehicle-end large-scale model functions based on the Snapdragon 8295 cockpit platform. With the development of multimodal technology, generative AI is expected to open up new application spaces for intelligent cockpits, autonomous driving, and other fields.
Qualcomm is also focusing on the development of ambient IoT, and at the same time flexibly adopts NB-loT, RedCap and other technologies to achieve all-round coverage of IoT application scenarios through technologies for IoT terminals and services at different levels. Regarding the combination of industrial and AI, Akash Palkhiwal pointed out that many industrial applications need to be enabled by connectivity, processing and AI, and the market will shift from microcontrollers to microprocessors with AI capabilities.
As a marathon, 5G, which is halfway to the marathon, not only continues to accelerate steadily, but is also becoming smarter, helping to unleash the multiplier effect of the "wireless AI" era in the process of synergistic development with AI.