The Pixel 6 and Pixel 6 Pro are the first two devices to ship with Google's custom processor, Tensor, instead of the conventional Snapdragon 888. The company has claimed that it is the most powerful mobile chip and will be put to the test when reviewers compare it to its top-notch counterpart from Qualcomm. Will it be so important? Here is the comparison between Google Tensor vs Qualcomm Snapdragon 888.

At the launch event, Google devoted most of its effort to detailing its new system-on-chip (SoC). The company said it applied its knowledge of machine learning (ML) to integrate artificial intelligence (AI) capabilities into a mobile phone equipped with the new chipset.

Why Tensor in the first place?

The Google Pixel 6 has never been a well-kept secret. Ahead of launch, plenty of compelling leaks and official certifications revealed key details about the upcoming handsets.

Google even officially announced the Tensor chip more than two months before launch, then gave hints about the design of the Pixel 6 and Pixel 6 Pro in its offline store in New York.

The Pixel, as the name suggests, has been dedicated to not only improving photography on cellphones, but also opening up APIs that other manufacturers can adopt to get better photos on their devices.

While the entire mobile industry has bet on bigger camera sensors and higher megapixel counts in its flagship phone, Google has always emphasized that its computational photography algorithms can outperform advances in camera hardware. throughout the history of the Pixel family.

But, despite the advanced software features, Google's reluctance to upgrade camera sensors on its flagship devices has led to a rapid decline in interest in Pixel phones.

The tech giant is finally making conscious efforts to remedy this problem by opting for improved camera hardware to complement its excellent photography software.

However, all of these efforts would not be as effective as with its new custom processor that allows it to maximize the performance efficiency of the new Pixel handsets.

Hardware Review

The Google Silicon team outlined some details of the new Tensor SoC, including its design, number of cores, and dedicated security features. This confirms many of the leaks and speculation we know about the Tensor processor, which was previously codenamed "Whitechapel".

Eight-core, three-cluster processor.

Like most other chipmakers, Google has licensed intellectual property from ARM to design a custom mobile processor. Google Tensor is equipped with eight cores, including two ARM Cortex-X1 cores, two Cortex-A76 cores and four Cortex-A55 cores that are based on a 5nm design, the company told ArsTechnica.

Based on this information, we can understand why Google Tensor has an edge over other competing chips, like Samsung's Exynos 2100 and Snapdragon 888 or Snapdragon 888 Plus.

The other two chipsets also feature a tri-cluster design, like Tensor, but are equipped with a single ARM Cortex-X1 core, three Cortex-A78 cores, and four Cortex-A55 cores.

Here is a quick comparison of CPU core configuration and clock speeds for different cores in Google Tensor, Snapdragon 888, Snapdragon 888 Plus, and Exynos 2100 chips:

SoC Google TensorQualcomm Snapdragon 888/888 Plus Samsung Exynos 2100

Configuration

processor

2x ARM Cortex-X1 @ 2.8GHz 2x ARM Cortex-A76 @ 2.25GHz 4x ARM Cortex-A55 @ 1.8GHz

Snapdragon 888:

1x ARM Cortex-X1 @ 2.84GHz

Snapdragon 888 Plus:

1x ARM Cortex-X1 @ 2.995GHz

3x ARM Cortex-A78 @ 2.4GHz

4x ARM Cortex-A55 @ 1.8GHz

1x ARM Cortex-X1 @ 2.9GHz 3x ARM Cortex-A78 @ 2.8GHz 4x ARM Cortex-A55 @ 2.2GHz

The Tensor prioritizes efficiency

Phil Carmack, VP and GM of Google Silicon, told ArsTechnica the company's reasoning for choosing two of ARM's Cortex-X1 cores instead of just one: "The processor will be able to distribute the load between the two Cortex-X1 cores, even for moderately important tasks, and it will contribute to more efficient performance. »

Carmack demonstrates a use case by sharing an example camera. From recording to rendering, and from Google Lens detection to ML, multiple tasks are performed at the same time when using the camera.

Therefore, multiple components of the SoC must work in harmony. Along with the camera hardware, the CPU, GPU, ISP (Image Signal Processor) and ML Processing Unit combine forces to contribute to a lag-free camera experience.

If Google were to stick with a single high-performance Cortex-X1 core in the Tensor, as is the case with its Snapdragon and Exynos counterparts, this workload would amount to "average" Cortex-A76 cores running at full capacity, but always with a lag.

Conversely, two Cortex-X1 cores can run the same workload with higher efficiency and lower power consumption than average cores. Better energy efficiency when performing tasks results in less heat generation and better battery backup.

In particular, the Pixel 5 or Pixel 4a 5G, which used the Snapdragon 765G chipset, were plagued with serious heating issues, especially when using the camera. Therefore, a custom processor architecture should, in theory, allow the Pixel 6 and Pixel 6 Pro to allocate resources optimally.

On the one hand, if Google chooses to use two Cortex-X1 cores instead of one, it's a little shocking to see that Tensor uses mid-size cores of at least three generations.

The Snapdragon 888 and Exynos 2100 use intermediate cores based on Cortex-A78, which is relatively more efficient than the Cortex-A76 implemented in Tensor. Google, unfortunately, didn't bother to offer solid reasoning on this.

Also, for low-intensity operations such as keeping the screen always on (AOD) and Now Playing, Google Tensor has a special Context Hub. Again, a unit dedicated to low-power tasks is a step towards greater energy efficiency.

Google's SoC Tensor uses a 20-core GPU and a 5G modem from Samsung.

Apart from the change in chip design, the Google Tensor has already been equipped with a Mali-G78 GPU, which is identical to that of the Exynos 2100. According to Google, it is a 20-core graphics processor, specially designed to deliver superior gaming performance.

It also claims that the GPU has 370% better performance than the Pixel 5. Real-world performance won't be known until we get the devices to run graphics benchmarks and test games on them.

Google Tensor is likely to rely on Samsung's Exynos 5123 modem for its 5G capabilities in most markets rather than opting for a Qualcomm modem.

Signs pointing to a Samsung modem in the Google Pixel 6 and Pixel 6 Pro were first spotted in the Android 12 beta by XDA and later confirmed in a report by Reuters.

The Exynos modem supports 5G Sub-6GHz and mmWave frequencies. But recent findings suggest that only certain carrier-locked variants of the Pixel 6 support both types of 5G signals, while unlocked models only support 5G Sub-6GHz.

Google Tensor bets on security

Google Tensor presents the second generation of its dedicated security chip: Titan M2, which succeeds the first generation Titan security chip, present in premium Pixel phones since the Google Pixel 3.

According to Google, the new security chip is designed to protect sensitive data such as passwords and PINs from online breaches, as well as physical attack techniques including "electromagnetic scanning, of tension and even the injection of faults by laser".

Along with the Titan M2 chip, Pixel 6 handsets will also feature a Tensor Security Core, a processor-based subsystem specifically designed to perform sensitive tasks by isolating them from other apps accessing that data.

Artificial intelligence is the main focus of Tensor.

Despite the advertised performance, Google hasn't built a custom processor to deliver better power efficiency than Qualcomm or other competitors.

The main reason, as Google bluntly shared, is to provide a stable and secure platform to run artificial intelligence (AI) and machine learning (ML) tasks on the handset itself, without depending on cloud infrastructure.

In fact, the chipset takes its name from Google's Tensor Processing Units, AI-accelerating processors used in its data centers.

Looking back, Google could be hinting at a custom SoC by showcasing dedicated AI chips including Pixel Visual Core and Pixel Neural Core.

Apart from the optimized processor, Google Tensor SoC also has a dedicated TPU, commonly known as NPU or neural processing unit, for running AI-based applications on Pixel 6 and Pixel 6 Pro.

Due to its nature and Google's experience with machine learning, Tensor is designed to run such models on the devices themselves.

This advanced architecture allows Tensor to perform complex tasks like Automatic Speech Recognition (ASR), which will actively translate any other language to your phone's default language in apps like Messages, WhatsApp and Recorder or even tools as Live Caption.

Improved speech recognition also allows Tensor to interpret pauses and punctuation marks in speech more accurately and using half the power of previous Pixel phones.

In addition to better speech processing, Tensor brings significant improvements to photography. First, the chipset now facilitates computational videography and photography using Google's HDRNet.

This machine learning algorithm ensures the Pixel 6 and Pixel 6 Pro capture the most vivid and accurate colors in every frame. The tensor also facilitates the use of functions such as the Face Unblur function, which helps to correct blurred faces in moving photos, the Magic Eraser function, which helps to eliminate unwanted objects from images, and better perception of the skin tone for people of color.

Why is Tensor essential for Pixel?

As Google tirelessly repeated throughout the Pixel 6 launch event, Tensor ensures that Google's latest advancements in AI can be delivered directly to its latest and future phones. That would be difficult to achieve with a generic SoC like the Snapdragon 888, especially with limited control over Qualcomm's chip design process.

Another reason Google chose a custom SoC with two ARM Cortex-X1 cores instead of just one is to ensure greater power efficiency and less heat loss. Unlike previous Google phones, like the Pixel 5, the new Pixel 6 is less likely to heat up when performing common tasks, like capturing 4K video.

The Snapdragon 888 and Exynos 2100 were also criticized for poor heat management to compensate for higher initial performance. However, higher amounts of heat over extended periods of time can cause throttling and ultimately slow down and lose performance, which is the primary goal.

Finally, if Google chose a bespoke processor, it was to draw the world's attention to its efforts to regain its dominant position in the mobile telephony sector.

Brands such as Samsung, Apple and Huawei are already making their own custom chips, while OPPO is also said to be working on its own. It is therefore essential for Google to go the extra mile and prove its skills in order to remain relevant in the mobile sector.