Google’s Tensor chip is just the first step to truly great Pixel phones

Google’s Pixel 6 will be the company’s most ambitious smartphone in years, thanks in large part to its 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. And Google needs it: Despite Pixels’ popularity in tech circles, its phones just aren’t popular sellers in the US, barely move the needle as compared to juggernauts like Samsung and Apple or even smaller players like Oppo or Xiaomi.

Tensor is Google’s big bet, centered on the AI-enhancing TPU that promises to improve photos and videos, search, captioning, text-to-speech, and more. It’s a tall order for any chip, let alone one primarily focused on machine learning as a standout feature — but while the Tensor SoC may not take Google to iPhone-like heights just yet, it could be a crucial first step to also run Android – came across a top candidate.

Image: Google

The rest of the SoC is a mystery at this point, but it looks like Google will be using third-party designs for things like the CPU, GPU, and modem — meaning the Pixel 6 will likely be on par with any other Android smartphone out there. is powered by a Qualcomm or Samsung processor for most tasks, rather than some sort of revolutionary upgrade similar to Apple’s A-series powered iPhones.

Google hasn’t given much information on what Tensor’s actual architecture will look like for things like the CPU, GPU, modem, or other key components of the SoC outside of the TPU. But based on rumors and the fact that Google isn’t taking the chance to crow about major tweaks or advancements made here, it’s likely that most of Tensor’s hardware will be outsourced designs. Qualcomm and Samsung are already doing something similar: the Snapdragon 888 uses partially modified versions of Arm’s Cortex-X1, A78 and A55 designs, while Samsung’s Exynos 2100 uses Arm designs for both the CPU and GPU.

XDA notes more specifically that the Tensor will likely be a combination of Arm’s Cortex-A78, Cortex-A76 and Cortex-A55 CPU cores and Arm’s standard Mali GPU. Which means the difference between Tensor and say a Snapdragon 888 or Exynos 2100 might not be that big for things like overall CPU or GPU performance. That’s a good category to be in, especially if Google is trying to really make a flagship device.

The Google Pixel 6

Image: Google

But as Rick Osterloh of Google told The edge“The standard stuff that people look at will be very competitive and the AI ​​stuff will be completely differentiated.” However, the things that make Tensor special and unique isn’t how fast it can run games or how efficient its CPU and battery life is.

That means Tensor probably won’t be the magical solution Android fans had hoped for either: a custom, Google-made chip designed specifically for Android and the Pixel hardware to deliver the kind of performance and power that Apple has been able to. bidding on the iPhone for years with a similar strategy.

But the good news is that Tensor is only a first-generation product; it’s easy to point to Apple’s custom chips and demand that Google do the same, but it’s important to remember that Apple’s first A-series iPhone chips started in a similar way. The A4 and A5 chips had standard Arm designs for the CPU cores (with some optimizations and improvements on top), before Apple moved to fully custom designs in later generations.

It is rumored that Google is collaborating with Samsung to produce the chips. ).

There are also rumors about Samsung to work more closely with Google on the actual design of the chips, leveraging Exynos hardware and software in a capacity beyond manufacturing. We’ll have to wait for Google to reveal more about Tensor to know the extent of that partnership, but it could mean Google gets some extra help with more substantial custom chip improvements than if it had to reinvent the wheel from scratch.

The idea of ​​a purpose-built machine learning chip in a Pixel phone isn’t a new idea for Google either. The company has previously put dedicated internal AI chips into its phones like the Pixel Neural Core and Pixel Visual Core in the Pixel 2, Pixel 3 and Pixel 4 models, making the Tensor less of a brand new advancement and more of a refinement. of his earlier work.

The Pixel 6 has two more fun color choices

Image: Google

Tensor’s TPU seems to be the next step, presumably with more powerful AI functionality than either of those two previous chips. It’s also more closely integrated with the phone, with Google noting that it can do things like pass image data directly through the TPU.

Those benefits might be worth developing a custom SoC — again, we’ll have to wait and see how Google actually implements those AI features and how significant they are over a stock Android flagship. (A sentiment that generally applies to how heavily Google emphasizes the TPU as the primary differentiator in the Pixel 6, something that seems like a tricky thing to actually resell to customers.)

As it stands, the main difference between Tensor and a Snapdragon 888 looks like they’re just some fancy neural network tricks. But as with Apple’s early internal chips, Tensor may well be the first step in a longer journey toward a more custom-built Google chip. And that’s a much more exciting concept than subtly better machine learning demos.


Related: