According to the latest quarterly tracking report by IDC, the shipment volume of wearable devices in China in the third quarter of 2023 was 34.7 million units, a year-on-year increase of 7.5%. The overall market continues to grow and is entering a stable recovery state. The shipment volume of the smartwatch market was 1.14 million units, a year-on-year increase of 5.5%. Among them, adult smartwatches were 5.59 million units, a year-on-year increase of 3.9%; children's smartwatches were 5.8 million units, a year-on-year increase of 7.2%.
The development of chips and the upgrade of computational power are important factors driving the development of wearable devices. Wearable devices are another application scenario suitable for edge computing. Wearable devices usually need to process and transmit a large amount of data in real time, such as health monitoring, smartwatches, etc.
On the one hand, wearable devices are located at the edge, and for small devices such as wearables at the edge, the requirements for computational power are much lower than those for intelligent driving and cloud computing devices, but they are very sensitive to cost, power consumption, delay, and development difficulty. On the other hand, because it is closest to humans and the real world, wearable devices are also the place closest to data. In the future, to achieve end-to-end neural networks, it is necessary to enhance the computational power in this field.
Perhaps one day in the future, wearable devices will also become products with complete functions that integrate data collection, processing, and analysis. We look forward to this day, so we start to think, is it too early to talk about computational power for wearable devices? Are there already companies doing this? What is the significance of the increase in computational power of wearable devices for edge computing? Who can pay for the energy and resource consumption brought about by high computational power? In this article, we will reveal these questions for everyone.
Is it too early to talk about computational power for wearable devices?
Overview of Wearable Devices
Wearable devices, as the name suggests, mainly carry devices on the body, just like wearing clothes and jewelry. According to the product form and purpose, we divide the current wearable devices into the following categories:
- Head-mounted category, with Apple Meta as the main representative. Through the display of head-mounted devices, it provides multi-level user experience such as audio, video, and virtual reality. Google Glass, which has stopped research and development, is also a typical representative among them. The head-mounted craze brought by Apple VisionPro, even accompanied by the development of Meta, strongly pushed a wave of Meta's stock price.
- Watch and bracelet category, representative companies include Apple, Google, Amazon, Huawei, Xiaomi, Little Genius, etc., can provide functions such as making calls, positioning, taking pictures, alarm clock reminders, sports health monitoring, etc., and can achieve more interactive operations when connected with mobile devices.
- Medical device category, representative companies include Fish Leap, Lexin, Jiuan, etc., and tech giants like Apple have also begun to apply for medical device qualifications in their focus areas. Of course, some classifications also include products that can be connected to the Internet and have IoT characteristics in the category of wearable devices.
- Human implant category, a new type of wearable device represented by brain-computer interface, represents a cutting-edge device technology. In fact, in the precision instrument majors of various universities, there are almost similar brainwave control product demonstrations. What Tesla does is closer to science fiction plots, such as controlling humans or bodies through chips. From this perspective, Meta CEO Zuckerberg has repeatedly stated that he hopes to use it after the technology matures, because he does not want to experience the "update" operation of products that have been implanted in the body.
Jitesh Ubrani, research manager of IDC's mobile and consumer device tracker, said. International Data Center. "Since the first Fitbit and Pebble watches came out, health and fitness tracking has made great progress, but the biggest driving force for wearable devices is the emergence of smaller, more fashionable designs. Oura, Noise, BoAT, Circular and other new brand smart rings are expected to start a new shape design in the next few quarters, while also putting pressure on existing brands to innovate in health tracking."
"Smart watches and earphones still dominate the wearable device market," added Ramon Llamas, research director of mobile devices and AR/VR. "These still resonate with consumers and continue to attract first-time users, especially those who are most cautious and price-sensitive. Here, we can still see that the shipments of emerging suppliers are enough to rank among the leading brands. Combined with a strong update cycle (including those who recently purchased wearable devices in 2020), the wearable device market has established a strong flywheel to maintain sales growth."
We have sorted out some typical products in terms of performance/computational power for your reference:
At present, the computational power of wearable device chips is usually not measured in TOPS (trillions of operations per second), because their computational power is relatively low, and it is more suitable to measure with DMIPS (millions of instructions per second) or other measurement standards more suitable for low power consumption and low computational power applications. However, with the development of technology, some chips designed specifically for wearable devices, such as Google's Wear OS chip, have begun to provide higher performance, including integrated AI functions and improved connectivity.
Industry Views
When choosing the core (module/chip) of wearable devices, manufacturers usually have two solutions: one is to adopt the module solution, although it may occupy more space, but the integrated and standardized design helps to reduce user costs; the second is the chip CoB solution, which is particularly common in wearable products such as children's watches, and about 90% of companies adopt this solution. Key considerations include whether to support Bluetooth, WiFi technology, volume, cost, and the balance of battery life and heat dissipation capabilities.
In communication with industry insiders, we found that for module companies or wearable device manufacturers, because products such as watches and bracelets have less application of computational power, and can even only be described by performance and power consumption, plus these products have certain requirements for battery life, Therefore, "talking about computational power" in traditional wearable scenarios such as personal watches and bracelets, there is no market demand and R&D motivation. China Mobile IoT believes that products that need to realize AI services or virtual reality functions, such as AR, VR and other new wearable devices, will increase the demand for computational power in the future, and as a company that provides modules, China Mobile IoT will continue to pay attention to customer needs and technology development. , Continuously optimize products.
At present, the chips and modules of wearable devices on the market are mostly based on frequency and power consumption as the main performance indicators, and have not yet made computational power a core selling point. However, to realize the intelligence of wearable devices and enhance the computational power of chips is an area that cannot be bypassed. For the new wearable market, this is undoubtedly a huge opportunity. Some companies have begun to enhance computational power on wearable devices, and this field is gradually becoming a new blue ocean.
So, which companies have started to pile up computational power on wearable devices?
Who is piling up computational power on wearable devices?
With the overnight fame of ChatGPT, Nvidia's market value has also gradually risen, and the importance of computational power in people's eyes is self-evident. From the large model at the end, people always want to make the big computational power product smaller, and smaller, preferably the kind that can be realized in wearable devices. Whether it is an exploration for future landing or a pursuit of terminal intelligence, these companies have indeed begun to give computational power to wearable devices, in other words, the chips on their products have initially taken shape in terms of computational power.
Apple
The first-generation head-mounted device caused a sensation among global tech enthusiasts. Apple has invested a lot of manpower and financial resources in this device. Of course, there is nothing to say about the heap on VisionPro, and the M2+R1 chip combination is directly adopted. The integrated graphics card in the M2 provides 8-10 cores and a peak computational power of 3.6 TFLOPs.
And equipped with 12 cameras, 5 sensors (including 1 lidar sensor) and 6 microphones, it can also be externally connected to a battery that can be used for up to 2 hours. Not only that, Apple has also specially designed an operating system for it, which is not to be underestimated. The appearance of VisionPro, a large-scale "wearable device", has also provided a solution to the "battery life crisis" to a certain extent, allowing some companies that want to provide computational power at the wearable device end to see hope.
On Apple's watch device side, it also uses its own 64-bit S9 SiP, which is said to contain 5.6 billion transistors and a new four-core engine, and carries almost a complete operating system. Such a device, without a bit of computational power, really can't hold it down.
As one of the earliest companies to bring VR devices and the producer of classic products Google Glass, Google's voice in the chips and computational power of wearable devices cannot be underestimated. On March 15, 2023, Google posted a notice on the Google Glass product page stating: Thank you for your innovation and companionship over the past decade. As of March 15, 2023, we will no longer sell Glass Enterprise Edition (Google Glass Enterprise Edition, referred to as Glass Enterprise Edition below). We will continue to provide technical support for Glass Enterprise Edition before September 15, 2023.
Subsequently, Google proposed similar plans, such as Project Iris, and there are even rumors that Google may have just temporarily shelved it, and with Apple officially entering the game through Vision Pro, Google may have resumed work on Project Iris. The project has gone through a turbulent development stage, with Google executives constantly changing their vision for Iris. To promote its efforts, the company also acquired North and Raxium. However, despite Google's various promotions of its actions in computational power and AI, such as self-developed chips, etc., the AI capability of Pixel phone photography still comes from the cloud.
Meta/Qualcomm
In recent years, with the development of large models, Qualcomm has gradually adapted to the demand for computational power of terminal devices and achieved rapid development. One of the representative products is Stable Diffusion running on the mobile end, and another hidden product is to provide chips and computational power platforms for Meta's head-mounted devices. Qualcomm has also played an important role in enhancing the computational power of wearable devices. A very typical example is providing chip platforms for Meta's AR/VR devices. Quest3 is equipped with Qualcomm's XR2 gen2 platform, which has significantly improved performance compared to the previous generation. Qualcomm promises that the new second-generation chip has a 2.5 times increase in GPU performance, an 8 times increase in AI performance, and a "50% increase in GPU energy efficiency", but it may not be achieved at the same time.
Neuralink & Tesla
In recent years, Musk has not only made Tesla "show muscles" in the field of intelligent driving and embodied intelligence, but also continued to follow up in the seemingly "sci-fi" track of brain-computer interface. Neuralink is committed to helping patients with mobility difficulties or paralysis of the limbs to control their bodies again through brain surgery to implant chips. Patients only need to convert thoughts into computer language and transmit them to the nerves to issue instructions to the body. Last May, Neuralink obtained approval from the U.S. Food and Drug Administration (FDA) to conduct the first human clinical trial; at the end of last year, Neuralink opened recruitment for trial participants who were paralyzed due to spinal cord injury or amyotrophic lateral sclerosis.
Musk's brain-computer interface company Neuralink introduced the situation of the first patient to receive brain-machine implantation in a live broadcast event on March 20, local time. This patient named Noland Arbaugh was paralyzed below the neck due to a diving accident. In the live broadcast, Arbaugh moved the computer cursor on the screen to play chess. Arbaugh said, "It's like using force on the cursor." Subsequently, Musk posted on X that Neuralink's next product will be "Blindsight" that can help blind people restore vision.
In conclusion, the future of wearable devices is promising, with computational power playing a crucial role in their development. As technology advances, we can expect to see more powerful, smarter wearable devices that can process and analyze data in real-time. This will not only enhance user experience but also open up new possibilities in fields like health monitoring, virtual reality, and more. The race is on to see who can harness this power most effectively and bring about the next big breakthrough in wearable technology.