No matter how the naming changes, Qualcomm's love for 8 has not changed. From the previous generation of Snapdragon 888 to the new generation of Snapdragon 8 mobile platform released yesterday, the 8 series has always been the signature series of Snapdragon chips.
As before, we can see the progress of the new generation of Snapdragon 8 mobile platforms in various aspects:
- Compared with the previous generation, Snapdragon 8 has a 20% performance improvement in the CPU part and a 30% reduction in power consumption.
- Compared with the previous generation, the graphics rendering speed of the Snapdragon 8 is increased by 30%, and the power consumption is reduced by 25%
- The new Snapdragon X65 5G Modem can provide download speeds of up to 10Gbps, the previous generation X60 5G Modem is up to 7.5Gbps
- Snapdragon 8 is the first mobile platform that supports 8K HDR video shooting, with 3 image processors with 18bit color depth
- Snapdragon 8 has used the 7th generation AI engine, and the AI performance can be up to 4 times that of the previous generation
What's the progress of Snapdragon's AI engine this time?
Wait, the progress of Snapdragon 8 in CPU, GPU, network, and imaging is one step at a time, steady progress, why does it play a long jump in AI performance, and move forward so much all at once?
The answer is that in addition to advances in hardware capabilities, there are also optimizations in software: In addition to increasing the tensor accelerator performance and shared memory of the Hexagon processor in the Snapdragon 8 mobile platform to 2 times, Qualcomm has also increased its AI software The performance is increased to 2 times.
However, for consumers who eventually buy the Snapdragon 8 mobile platform terminal, how strong the AI performance is, and how strong the AI performance is, is not so important. The important thing is, what can such a strong AI performance do? If it's just for a long face on PPT, wouldn't adding this AI engine increase power consumption?
Qualcomm said that the consumer's idea is very reasonable, we first reduce the power consumption, and then consider what it is useful for. As a result, Qualcomm went on to state that the AI energy efficiency of the new generation of Snapdragon 8 is 1.7 times higher than that of the previous generation, and the performance per watt is the highest among all processors.
Then, Qualcomm spent a lot of space telling us what can the AI of the Snapdragon 8 mobile platform do?
Snapdragon AI is actually everywhere
In fact, in the schedule of the first day of this technology summit, they talked about AI very fragmented. This fragmentation is because AI is indeed ubiquitous. Many functional features do not exist independently, but require the assistance of AI to be realized.
For example, this "virtual shoe test" function completed with partners: as long as you point the camera at your feet, you can see the effects of various shoes on your feet. Obviously, computer vision is used here.
For another example, in the basic technology of the meta universe, Qualcomm believes that AI must also have a place, even if the meta universe is just a concept now.
On the new generation of Snapdragon 8 mobile platform, in addition to the three main ISPs (image processors) that ensure that the three cameras are running at the same time, there is also an always-on ISP to realize the always-on face detection of the front camera.
The first thing to emphasize is that this independent ISP is located in the third-generation sensor hub of the Snapdragon 8 mobile platform. It does not record or upload user information. The data only remains in the sensor hub. It only provides functions such as fast facial detection, so No need to worry about it leaking privacy. This third-generation sensor hub is also an important part of the AI engine of the Snapdragon 8 mobile platform. The new low-power AI subsystem consists of a dedicated AI processor, DSP (digital signal processor), and its own memory for processing Situational data streams, including voice, audio, sensors, and connections.
▲ When you can’t use your hands for cooking, you can unlock the bright screen and read the recipe by just looking at your phone
The intelligence of the independent ISP in the sensor hub is that the front camera does not need to be manually woken up, but works at low power consumption all the time, just like the Bluetooth or 5G signal that is always on standby. When it detects the user’s facial information, When the user can't make a move, the screen can be unlocked directly.
This function also has an anti-peeping effect. For example, on the subway, we often encounter screen peeping monsters behind. At this time, the 4th ISP and the front camera that are always working can become the eyes behind us and detect the behind Whether the eyes are peeking at the screen, as to whether to lock the screen directly, remind the user, or ignore it, the mobile phone manufacturers and users can choose the rest of the work.
Furthermore, it is precisely because of this feature that the front camera of the mobile phone can always know whether the person staring at the mobile phone is the owner. Therefore, the mobile phone can even make it impossible for other people to make the mobile phone bright, completely protecting the user's privacy. After all, sometimes the notification information is quite sensitive. This new hardware function closely related to AI technology can provide mobile phone manufacturers and application developers with a variety of new scenarios and new functions.
Let’s talk a little bit more practically. Taking pictures to beautify is not a new technology anymore. The background blur, gesture recognition and gesture recognition are also very simple now, but the performance requirements of using these technologies in the video are even greater.
▲ Snapdragon combined with Rainbowsoft to realize 4 neural networks working at the same time
Moreover, if these four technologies (face recognition, background blur, gesture recognition and gesture recognition) are to be used in video shooting at the same time, the requirements for AI performance will be further increased, because four different technologies correspond to four neural networks Operation. With the support of the new generation of Snapdragon 8 mobile platform AI capabilities, the visual algorithm provider Hongsoft has realized that four neural networks can run simultaneously in video shooting, and simultaneously perform facial recognition, background blurring, gesture recognition and posture recognition.
Someone may ask, what use are these four different technologies? It doesn't seem to bring anything. For example, the AI capabilities of Snapdragon here are like the foundation, and Rainbow Soft builds a rough house on the foundation. Eventually, mobile phone terminal manufacturers or application manufacturers will carry out fine decoration (beauty, long legs, blur… ), what the final consumer lives in is a well-decorated house. But if there is no foundation and rough house, then fine decoration is just empty talk.
And at the foundation stage, it has been decided whether the house is a two-bedroom, one-living room or a luxury villa. From here, it can be seen that whether the AI performance is strong or not is closely related to the final experience.
In fact, Qualcomm will not provide hardcover rooms.
Earlier, Sharp and Leica created the Leitz Phone 1 based on the Snapdragon 888 mobile platform. It uses a 1-inch outsole camera with a built-in Leica LeitzLooks mode that can outline Leica-style photos on the phone. On the Snapdragon 8 mobile platform, Qualcomm and Leica’s Ernst Leitz laboratory announced that the future Snapdragon mobile platform will support the Leitz Looks mode powered by Qualcomm’s AI engine. Even more Leitz Looks modes based on other traditional Leica lenses will be launched in the future.
▲ Leica style photos
This means that as long as the mobile phone manufacturer is willing, they can directly use the Leitz Looks mode in the mobile phone to help users take Leica-style German photos. Yes, all Qualcomm and Ernst Leitz laboratories are ready, and the mobile phone manufacturers can use them directly.
Computer vision is of course the most widely used and strongest perceptual field among various AI technologies, but AI has many other magical uses.
AI actually has hearing ability. In the four ways to look, hear, and ask, you can hear and ask about it.
▲ AI assesses emotional status
Qualcomm began to cooperate with the health company Sonde Health. AI voice analysis can be performed on the mobile phone of the Snapdragon 8 mobile platform to track the user's health, such as mental health, and sound-related health, such as whether the user has Asthma and so on.
In the game, many people are actually quite familiar with AI: either the AI player is violently beaten, or the AI player is violently beaten. In addition to such mutual harm, AI also has some practical uses.
The AI engine of the new generation of Snapdragon 8 mobile platform is in the game, and there are 2 functions that allow users to directly perceive: advanced rendering and anti-cheating detection. The former can help the game to achieve a clearer picture, while the latter can help detect whether other players have used cheating software and promote fair play.
In many subtle places, such as the signal level, AI also has interesting and useful uses. One of the headaches for many mobile phone manufacturers is the signal, especially when the user holds the mobile phone horizontally to play games or watch videos, it will block the antenna and interfere with the signal. However, games and videos are the scenes with the highest requirements for network signals. To solve this problem, in addition to being hard-headed, it is to redesign the antenna layout and use the power of AI.
In the Modem (modem) and radio frequency system of the Snapdragon 8 mobile platform, AI antenna tuning technology has been added to the package. It can detect the position of the user’s hand when holding the phone, enhance the antenna tuning ability, ensure stable and fast data transmission, and battery Endurance level (the worse the signal, the higher the antenna receiving power and the power consumption).
It can be seen that on the new generation of Snapdragon 8 mobile platform, Qualcomm provides a variety of AI capabilities at all levels, such as games, imaging, productivity, connectivity, and health monitoring, from the foundation level to the hardcover room bag occupancy level. Many are still working 7×24 hours.
In fact, the ultimate goal is to make AI ubiquitous and make AI a daily companion.
It will know the user's daily running habits, open the health app at the right time, count exercise data, and provide running music; it will also know that the user has started driving to work, automatically turn on the navigation, turn on the voice control mode, and ensure safe driving…
From CPU to SoC to mobile platform, the name change also shows the development history of mobile phone chips and changes in responsibilities, from computing instructions, to image and graphics processing, to receiving and processing network signals, and AI computing, etc., we have experienced SoC has gone from the era of core scoring to the era of AI wrestling. Behind it is the change in the form of smart phones. It is a phone, a camera, a game console, and a close partner.
From this perspective, the development of smart phones is still a long way from the end. After all, one of the biggest improvements in the new generation of Snapdragon 8 mobile platforms includes AI. Intelligence is still advancing rapidly. Of course, smart phones is also like this.
#Welcome to follow Aifaner's official WeChat account: Aifaner (WeChat ID: ifanr), more exciting content will be provided to you as soon as possible.