The phone can no longer accommodate Snapdragon 8

Yuyang Mingmin from Aofei Temple
Quantum Bit Report | Public Account QbitAI

What kind of fresh experience can the latest generation of Android flagship bring?

Qualcomm has already given an advance preview in the past two days.

If the 20 to 30 percent improvement in CPU and GPU is not strong enough, then how do you feel about the AI ​​super-resolution function that was previously only available in PC games being moved entirely to mobile games?

With the launch of the new generation Snapdragon 8 mobile platform , shooting 8K HDR video with mobile phones is expected to become standard in the future.

Or the kind that can reach 33 million pixels per frame and can take up to 64 million pixel photos at the same time.

Some netizens have already heard the news and are delighted: from now on, I won’t have to worry about taking bad photos of my girlfriend anymore. I can just shoot a video and then slowly take screenshots and I’ll be done (dog head) .

Although Xiaomi CEO Lei's first flagship is still on the way, there is no doubt that Snapdragon 8 has once again sparked people's infinite imagination of mobile phone performance with the eye-catching displays at the press conference.

However, if you are eager to try it out at the end of the year, I have to remind you: the above is not the whole picture.

How to say?

If you take a closer look at the content released at the Snapdragon Technology Summit, you will find that the word AI is everywhere .

And this time, it is no longer limited to images.

Now, on a mobile device equipped with Snapdragon 8, AI can be said to have penetrated into every detail you can see or not, and is “ always on ”.

What kind of AI capabilities does Snapdragon 8 have?

First, let’s take a look at the specific parameters of the new generation Snapdragon 8.

Whether it is the CPU, GPU or various modules, it can be said to be a comprehensive upgrade.

In these upgrades, AI technology is also fully demonstrated. Let’s start with the specific functions.

This year, if you ask what details Snapdragon 8 has surprised people in, you must talk about ISP.

Based on the three ISPs of the previous generation, Snapdragon 8 adds a fourth ISP.

This smart sensing ISP is very different, it can make your phone anti-spying .

That is, if it detects someone is looking at your screen, it will automatically lock the phone .

This ISP supports the camera to operate with extremely low power consumption, which means it is "always online".

Not only does it prevent peeping, it is also on standby at all times to recognize some special images, such as QR codes.

Face unlocking anytime and anywhere is also achieved based on this ISP.

On top of this, with the help of face recognition algorithms, Snapdragon 8 increases the number of facial feature recognition points to 300 , which can capture more subtle changes in expressions, while also increasing the face detection speed by 300% .

This also solves the problem of faces being hard to recognize after being shrunk in ultra-wide-angle lenses .

The latest AI algorithm can handle even the side profile or facial obstructions.

In terms of photography , the AI ​​support brought by Snapdragon 8 has also been further improved.

Qualcomm announced that it will cooperate with Leica to realize Leitz-style filters on mobile phones through AI technology, allowing ordinary users to easily take blockbuster films.

In addition, Snapdragon 8 also combines intelligent voice and health applications.

They have partnered with health company Sonde Health to use AI to recognize the user's voice and determine his health condition.

(Is this the “smell” in “look, smell, ask, and feel”?)

It can make health judgments by recognizing the user's voice and inferring whether the user is depressed, suffers from asthma, etc.

Of course, we can’t forget about games . As mentioned at the beginning, Snapdragon 8 brings AI super resolution function to mobile games.

This technology is similar to the Deep Learning Super Sampling (DLSS) technology previously used in PC games . It can use deep neural networks to extract multi-dimensional features of the rendered scene and intelligently combine details from multiple frames to construct a high-quality final image.

This method can achieve the same effect as traditional rendering, but with higher performance.

Qualcomm said that as long as the platform is equipped with Snapdragon 8, this function will be automatically turned on when playing games .

At the same time, Qualcomm also took into account the signal problem that is a headache for gamers.

Through AI technology, Snapdragon 8 can achieve more stable signal transmission and take into account battery life issues, while also improving hand-holding accuracy by 30% .

The key to realizing the above multiple functions is the AI ​​engine .

The new generation Snapdragon 8 is equipped with the 7th generation Qualcomm AI engine .

Compared with the previous generation, the speed of the tensor accelerator and shared memory have both increased by 2 times .

Large shared memory is essential for running deeper neural networks, so that they can run in full without having to be processed in chunks.

In terms of the processing performance of each model, the new generation Snapdragon 8 has also been comprehensively upgraded based on Snapdragon 888.

The overall performance has been improved by 4 times .

It is worth mentioning that all the above AI functions can be enabled in real time , relying on the third-generation Qualcomm sensor hub on the AI ​​engine , which allows AI to process massive amounts of data information with extremely low power consumption.

For example, the camera can detect images online in real time and the voice assistant can be called up at any time, all relying on its functions.

In addition, the performance of the new generation of sensor hubs has been improved by 75% and power consumption has been reduced by 50% compared to the previous generation .

It's not just an improvement in hardware. How detailed Qualcomm's AI application is this time also depends on the software level.

Language processing is one of the details of Qualcomm's AI application this year.

On the new generation Snapdragon 8, they cooperated with Hugging Face, a well-known company in the NLP field , to enable the smart assistant on the mobile phone to prioritize and analyze notifications.

This is a technology company focusing on natural language analysis. In their open source community, there are more than 6,000 models and nearly 800 data sets covering more than 300 languages.

When running their sentiment analysis model on the Qualcomm AI Engine, it can be 30 times faster than on a regular CPU .

Not only that, Qualcomm has also reached a cooperation with Google, a giant in the field of artificial intelligence.

Snapdragon 8 becomes the first SoC platform to provide Google Cloud Vertex AI NAS neural architecture search service.

Vertex AI is a managed machine learning platform released by Google at this year's I/O conference. It can help developers deploy and maintain AI models faster and easier.

According to Google, Vertex AI NAS reduces the number of lines of code required to train models by nearly 80% compared to other platforms.

This time, Qualcomm integrated Vertex AI NAS into the Qualcomm Neural Processing SDK and ran it on the Qualcomm AI Engine. With this, engineers will be able to build and optimize new mobile AI models in a shorter time.

It starts with mobile phones, but it’s more than just mobile phones

The above mentioned seems to be still focused on mobile phones, the most popular mobile terminals.

But in fact, Qualcomm's AI capabilities go beyond this and have already covered a wider range of smart terminal devices.

As Alex Katuzian, senior vice president and general manager of mobile, computing and infrastructure at Qualcomm Technologies, said:

Innovations in the Snapdragon flagship platform are always applied to smartphones first and then quickly expand to other growth areas.

For example, autonomous driving .

Now, you can find Qualcomm AI technology in the fields of in-vehicle networking, digital cockpit and autonomous driving.

After spending $30 billion to acquire Arriver, a supplier of autonomous driving parts, Qualcomm announced its new strategic map for its automotive business this year.

Its Snapdragon automotive digital platform includes the Snapdragon Automotive Digital Cockpit Platform, Snapdragon Ride Autonomous Driving Platform, Snapdragon Car-to-Cloud Service Platform, and Snapdragon Auto Connectivity Platform.

Currently, Qualcomm has reached a cooperation with BMW in the field of autonomous driving. BMW's next-generation assisted driving system and autonomous driving system will adopt Qualcomm's autonomous driving solutions, including the Snapdragon Ride vision system-level chip (SoC) , visual perception, and ADAS central computing chip (SoC) managed by Qualcomm's Car-to-Cloud platform .

In addition, Qualcomm also revealed that it has reached cooperation with more than 25 car companies, and the number of connected cars using Qualcomm solutions has reached 200 million.

For example , XR .

At the Snapdragon Technology Summit, Alex Katuzian performed a live "shoe-changing" scene.

Qualcomm has partnered with Wanna Kicks to bring the capabilities of the Qualcomm AI engine to the AR try-on app.

Just point the Snapdragon-powered camera at your feet to complete the "shoe-trying" experience online.

Earlier this year, Qualcomm also released the Snapdragon Spaces XR development platform, which is designed to help developers create immersive experiences for head-mounted AR glasses.

Qualcomm's New Quadrant

As the saying goes: One minute on stage, ten years of hard work off stage.

As one of the most successful companies in the mobile Internet era, the public's general perception of Qualcomm began with generations of flagship mobile phone chips and mobile platforms.

But in fact, judging from the various capabilities demonstrated today, "a mobile phone chip manufacturer" is obviously not enough to summarize Qualcomm's full picture.

Qualcomm, which has entered the AI ​​era, has already opened up its own new quadrant by continuously increasing its investment in cutting-edge technologies such as 5G and AI.

Looking back from this point in time, Qualcomm's deep involvement in the field of AI has become clearer.

Starting with Snapdragon 820 in 2018, Qualcomm's flagship chip has fully integrated Qualcomm's self-developed artificial intelligence engine AI Engine.

The 2019 Snapdragon 855 uses heterogeneous computing to mobilize different functional cores such as CPU, GPU, Hexagon, etc. to achieve efficient AI processing.

With Snapdragon 888, AI capabilities are not only stronger, but are also increasingly taking charge of the overall situation .

This year, AI is behind almost everything you can and cannot see on your mobile device, whether it is connection, imaging, gaming, or security.

It can be said that starting from the main battlefield of smart devices such as mobile phones, Qualcomm, through a combination of efforts with a number of partners, has actually used AI technology to create more possibilities in people's daily lives.

Finally, a small question:

Are you planning to buy a flagship phone equipped with Snapdragon 8?