Google’s Pixel 6 is shaping up to be the company’s most ambitious smartphone in years, largely due to the new, custom-designed Tensor processor, which aims to catapult Google to the forefront of the smartphone market with the power of Google’s years of machine learning experience. Google needs it.
Despite the high popularity of Pixels within tech circles, their phones are not selling well in the US. They don’t sell as many units as Apple and Samsung or smaller players like Oppo, Xiaomi, or Samsung.
Google’s biggest bet is Tensor, which focuses on an AI-boosting TPU that promises improved photos and videos, search and captioning, as well as text-to-speech and other features.
It’s a tall order for any chip, much less one focused mainly on machine learning as its standout feature — but while the Tensor SoC might not launch Google to iPhone-like heights just yet, it could be a crucial first step toward turning its Android also-ran into a top contender.
While the rest of the SoC remains a mystery, it appears that Google will use third-party designs for tasks like the CPU, GPU, and modem.
This means that the Pixel 6 will likely feel similar to other Android smartphones powered by a Qualcomm/Samsung processor, rather than a revolutionary upgrade comparable to Apple’s A-series iPhones.
Google has not provided much information on Tensor’s architecture beyond the TPU, such as the CPU, GPU, and modem. However, rumors suggest that Tensor’s actual architecture will not be disclosed.
This is likely because Google hasn’t made any public announcements about major customizations or advancements it has made here.
Qualcomm and Samsung already do something similar — the Snapdragon 888 uses partially customized versions of Arm’s Cortex-X1, A78, and A55 designs, while Samsung’s Exynos 2100 uses Arm designs for both its CPU and GPU.
XDA notes more specifically that the Tensor will likely be some combination of Arm’s Cortex-A78, Cortex-A76, and Cortex-A55 CPU cores and Arm’s standard Mali GPU.
The difference in Tensor’s performance compared to a Snapdragon 808 or Exynos 2100 may not be significant for overall CPU and GPU performance. This is a great category to be in, especially since Google wants to make a flagship device.
However, Rick Osterloh, Google’s chief technology officer, stated to The Verge that “the standard stuff people see will be very competitive, and the AI stuff won’t be totally different.” The unique things that make Tensor stand out aren’t its speed or efficiency, but how long it can run games and how long it lasts.
Tensor won’t be the magic solution Android fans are looking for. It’s a Google-made custom chip that is specifically designed for Android and Pixel’s hardware. This will allow it to deliver the same performance and power Apple has offered for years with a similar strategy.
TENSOR IS THE FIRST-GENERATION PRODUCT
Tensor is a first-generation product. While it is easy to point out Apple’s custom chips to demand that Google do the same, it is important to remember that Apple’s original A-series iPhone chips were created similarly. Before Apple moved to custom designs, the A4 and A5 chips used standard Arm designs for their CPU cores.
Google and Samsung have rumored partners in manufacturing chips. This makes sense considering it is one of only two companies (besides TSMC) that currently manufacture 5nm chips.
Samsung is also rumored to be working more closely with Google on the actual design of the chips, too, leveraging its Exynos hardware and software in a capacity beyond simply manufacturing.
While we will have to wait for Google to reveal more about Tensor, it is possible that Google could get more help with custom-chip improvements than if it had reinvented the wheel.
Google isn’t new to the idea of a machine learning chip built specifically for Pixel phones.
Google has used dedicated AI chips from its factories before in its phones, such as the Pixel Neural Core, Pixel Visual Core, and Pixel 4 models. The Tensor is less a new advancement than it is a refinement to its past work.
Tensor’s TPU seems to be the next step, offering more powerful AI functionality than the two preceding chips. It is also integrated more closely with the phone. Google noted that it could route image data directly through the TPU.
These benefits might be worth developing a custom SoC. However, we will have to wait and watch how Google implements them and their differences over an Android flagship.
This sentiment is generally attributed to the way that Google emphasizes the TPU in the Pixel 6. It doesn’t feel easy to sell customers on this feature. However, Tensor looks to be a Snapdragon 8888 with some fancy neural network tricks.
Tensor, just like Apple’s in-house chip, maybe the beginning of a longer journey towards a more customized Google chip. This is a much more exciting concept than the subtly improved machine learning demos.
Why is Google Tensor Important?
Google will use these Tensor chips in its other devices like the Pixelbooks or Nest smart speakers. Better interoperability is possible with devices that have unified hardware.
Tensor chips give Google greater control over various aspects of consumer technology.
Google already develops the Android operating system and its associated Google services (like Gmail and Maps). It can adapt the software to these Tensor chips more effectively than other companies. This brings Google closer to Apple’s vertically integrated ecosystem, which has been a key driver of the company’s success for many years.
What is Google Tensor?
This is Google’s first system-on-a-chip (SoC) processor. This is the first SoC of Google. It’s been made by Google before with its Pixel Pixelbook Go and Pixel smartphones. Google has, however, made machine learning (ML) processing chips, called Tensor Processing Units (TPU), for its data centers, since 2017.
What will Google Tensor do for you?
Google said Tensor would improve photo and video processing. It will also power up such services as voice-to-speech or translation software. Google claimed that Tensor would perform new tasks, such as translating between languages and taking still photos when subjects — typically my grandsons — move.
The tensor will use Google’s cloud resources. However, the phone will process more information locally and not have to send it to the cloud.