Robert Triggs / Android Authority
With Google already lifting the lid ever so slightly on the upcoming Pixel 7 series, expectations and rumors about the phone’s next-gen silicon are beginning to circulate. The upcoming chip, tentatively dubbed the Google Tensor 2, follows Google’s first foray into the semi-custom silicon space and will power Google’s next-generation handsets.
There are plenty of changes, tweaks, and outright improvements Google could make to its next SoC. Nothing is ever perfect on a first try, after all. So let’s wade through the current rumors and reasonable expectations about what will be coming our way later in 2022.
Catch up: Everything you need to know about the first-gen Google Tensor
Google Tensor 2 expectations and rumors
Robert Triggs / Android Authority
Early Google Tensor 2 rumors point to a few givens for the next-gen chip. It will be co-developed with Samsung, just like its predecessor, and will almost certainly be built on a newer, smaller 4nm process. It will likely be the same 4nm LPE process Samsung uses for its own Exynos 2200 and Qualcomm’s Snapdragon 8 Gen 1 flagship silicon, or perhaps Samsung’s slightly newer and more efficient 4nm LPP process. What’s less certain and open to a few possible variations is the makeup of the chip’s CPU, GPU, and other components.
On the CPU side, no mixing and matching between Armv9 and Armv8 CPU cores creates a choice between sticking with older components or moving to higher-performance but potentially higher-cost CPU parts. A move to off-the-shelf Armv9 components is a likely next step, as Google doesn’t design custom CPU cores and we can’t see an X2/A78 or some such combination. Upgrading the current big Arm Cortex-X1, medium Cortex-A76, and little Cortex-A55 cores to the Cortex-X2, Cortex-A710, and A510 seen in 2022’s flagship Android smartphones would provide a decent bump in performance. Particularly as this would mean moving to a more modern middle core compared to 2018’s rather dated A76.
Read more: Arm Cortex-X2, A710, and A510 deep dive
The other unknown is whether Google will stand by its 2+2+4 cluster configuration or move to a more conventional 1+3+4 CPU setup. Either way, the decision will tell us a lot about whether Google feels that this arrangement offered the right balance of performance and power consumption in its first-generation chipset.
Key ingredients include Armv9 CPU cores and second-gen TPU, built on Samsung’s 4nm process.
Google Tensor 2 will come packing an upgraded version of the company’s machine learning smarts. The current-gen Tensor Processing Unit (TPU) couples the chip’s AI processing capabilities right alongside multimedia pipelines like the audio and image processing units. Close integration between these parts allows Google to leverage its industry-leading AI algorithms for natural language processing, image recognition, and photo processing inside the Pixel 6 series, which helps the phones punch well above their price tag. See Magic Eraser, Pixel’s signature HDR image processing, and on-device speech captioning for examples.
What is an SoC? Everything you need to know about smartphone chipsets
While we don’t know exactly what TPU improvements Google will make with Tensor 2, we’re expecting another tightly coupled, bespoke piece of AI silicon to create unique experiences for the Pixel 7 series.
What we’d like to see
Robert Triggs / Android Authority
Sticking with the TPU/media suite for a moment, there is room for improvement. We’d like to see Google Tensor 2 revamp the ISP’s throughput to support improved video capabilities. For example, 4K 60fps video only works from the primary camera but it would be great to have the same feature support for the ultrawide, telephoto, and even the selfie camera too. 8K recording might be useful for more serious videographers also. It’s not clear if this is a limitation of the current chipset or the camera modules, but we know that Samsung’s Exynos chipsets have these capabilities, and they would be much-coveted features for the Pixel 7 series. Likewise, there are no simultaneous multi-lens capabilities in the Pixel 6, a feature we’ve seen in other smartphones. A wider, more robust imaging pipeline would be the answer.
While not a hardware wish, opening the TPU up to third-party developers would also be a feature in Google’s cap. The current TPU is very capable, but its potential remains limited to Google’s in-house applications. Nice to have, of course, but there are so many AI/ML applications that could also benefit Google’s hardware acceleration.
See also: How on-device machine learning has changed the way we use our phones
Graphics is an area that’s seen big gains in 2022 and we would like to see Google Tensor 2 level the playing field. Moving to an Arm Mali-G710 from the G78 is a likely progression, which could yield a 20% or greater performance uplift. However, that might not be enough to close the gap on the Snapdragon 8 Plus Gen 1 that will appear in handsets around the same time as the Pixel 7, let alone the yet to be announced Snapdragon 8 Gen 2 that will power 2023 flagships.
Samsung’s partnership with AMD has produced a decent, if not ground-breaking, Xclipse 920 GPU in the Exynos 2200. If Samsung can license the fruits of its RDNA2 collaboration, a suped-up Xclipse GPU for the Google Tensor 2 could lend the chipset some serious gaming chops. While gaming isn’t Google’s priority, competitive frame rates would certainly help make the Pixel 7 a jack of all trades.
Beefed up image, graphics, and modem capabilities top our Tensor 2 wishlist.
Finally, early codenames point to the Tensor 2 featuring a Samsung Exynos modem, which is to be expected. However, the codename suggests an as-yet-unannounced Exynos 5300 model. This could be an error at this stage, of course, but could also point to a semi-custom and/or integrated component explicitly produced for Google. The original Google Tensor used an external Samsung modem. Moving to an integrated component would be a boon for power consumption and device battery life, particularly when chugging data down those fast 5G pipelines.
Read more: Snapdragon 8 Gen 1 vs Google Tensor
Built for the Google Pixel 7
The original Google Tensor wasn’t a particularly ground-breaking SoC. Still, it was highly innovative in the machine learning department, which in turn enabled Google to stuff its Pixel 6 and Pixel 6 Pro smartphones with impressive AI-based features. Rather than an out-and-out benchmark thrashing chip, Google and Samsung designed silicon uniquely tailored to its mobile vision and needs. We can almost certainly expect this ethos to run through the core of the Google Tensor 2’s design.