Qualcomm launches AI-optimized WiFi 7 chip

A new wireless communication solution from Qualcomm will integrate WiFi 7, Bluetooth and ultra-wideband technologies into a single chip that will have performance optimized by Artificial Intelligence (AI). The FastConnect 7900 will have the ability to adapt to different use cases and environments, seeking to provide significant savings in energy consumption and optimize latency and data transfer rates.

“WiFi is a bit like the Wild West. There are all kinds of devices out there, congestion, devices coming and going, access points doing this or that – it’s very difficult to guarantee service. And AI is the ‘perfect tool’ to change that,” said Javier del Prado, vice president of mobile connectivity at Qualcomm, in an interview with the IEEE Spectrum website.

WiFi 7, also known as IEEE 802.11be and an evolution of WiFi 6E, is the new generation of wireless technology that promises faster speeds, lower latency and the ability to handle more connected devices simultaneously, the latter an important feature for Internet of Things applications.

Qualcomm uses trained AI models to classify and optimize quality of service and prioritize various types of data flow. The FastConnect 7900 doesn’t inspect packets in real time, it just uses these AI models to optimize latency and bandwidth during data transfer.

According to Qualcomm’s tests, the AI-enhanced WiFi system can make relevant adjustments depending on the type of use, for example, minimizing latency during a video call and consuming less energy during periods of inactivity. With this behavior, Qualcomm hopes to offer energy savings of 30% compared to the FastConnect 7800 present in the Snapdragon 8 Gen 3.

The FastConnect 7900 integrates WiFi 7, Bluetooth and Ultra Wideband (UWB) technologies to provide a set of technologies that allow devices to be discovered, accessed and controlled securely. It also uses a new class of radio frequency front-end modules and the new generation of High Band Simultaneous technology for WiFi 7, which uses two WiFi radios simultaneously to reduce latency and minimize interference. It includes Bluetooth Channel Sounding to easily find nearby devices such as headsets, smart tags or other portable equipment.

According to Roger Sands, CEO and co-founder of Wyebot, a company specializing in Artificial Intelligence-driven WiFi automation, network environments are becoming increasingly complex, so companies need a new way of managing this complexity. “AI in WiFi has already started to take shape in the last year or two. By 2024 and beyond, it will be a widespread initiative all over the world because of these dynamic networks,” says the executive in a Techtarget article.

He explains that AI can analyze large volumes of data to make predictions about network performance. For example, AI in WiFi can detect potential problems affecting the user experience before they occur.

AI hub

Qualcomm also recently unveiled a hub aimed at developers who want to harness the power of AI. The main feature of the new Qualcomm AI Hub is a library of pre-optimized AI models for deployment in devices with Snapdragon and Qualcomm platforms. There are more than 75 AI and Generative AI models optimized to make better use of memory and offer better energy efficiency in different form factors and packaged in various runtimes.

According to Qualcomm, each model is also optimized to take advantage of hardware acceleration on all Qualcomm AI Engine cores (NPU, CPU and GPU), resulting in 4X faster inference.

Developers can integrate these models into their applications, speeding up market launch and reaping the benefits of implementing AI in the devices themselves, such as reliability, privacy, customization and cost savings.

The optimized models are already available on Qualcomm AI Hub, GitHub amd Hugging Face. Qualcomm’s intention is to continually add new models to these sites, as well as ensuring compatibility for more platforms and operating systems.