Since the 2021 Google Pixel 6, Google’s in-house Tensor chip has remained one of the biggest drawbacks of its phones.
While steady improvements in other areas have boosted Pixel’s popularity, the lineup still lags behind other Android flagships — all because of the underwhelming Tensor chip.
With the Pixel 10’s Tensor G5, Google finally had the chance to turn things around.
It could have finally offered a flagship that delivered a proper balance of performance and battery life. Instead, it fumbled that opportunity.
Google’s Tensor problems trace back to Samsung
A partnership that never should have been
Designing and building an SoC in-house is expensive.
So, to get a head start, Google used Samsung’s Exynos chip as a base platform. It also used Samsung’s process node to fabricate the chip.
Since the 2021 Pixel 6, Pixel fanboys have blamed Samsung — and, to a certain extent, Google — for the issues plaguing Tensor chips.
While Google gradually shifted to off-the-shelf CPU and GPU designs from Arm, it continued to rely on Samsung’s foundry and modem for Tensor chips.
This led to all Pixel phones suffering from two major issues: poor thermal performance and mobile connectivity issues.
The Exynos 5400 modem inside the Pixel 9 largely fixed the latter problem, though it arrived a couple of years too late.
Samsung and TSMC rank among the world’s largest foundries, capable of producing the latest chips on cutting-edge process nodes.
However, the two are not alike, with TSMC pulling ahead of its Korean competitor with its fabrication technology in the last few years.
The same chip fabricated on TSMC’s and Samsung’s process nodes can show significant differences in power efficiency and performance.
The 2022 Galaxy S22 is the prime example of this. It launched with a 4nm Samsung-fabricated Snapdragon 8 Gen 1 chip, which was widely criticized for its poor thermal performance.
A few months later, Qualcomm switched to TSMC for the Snapdragon 8+ Gen 1. The underlying CPU architecture remained the same; just the switch alone brought a massive improvement in thermals and power consumption.
Since then, Qualcomm has stuck to TSMC’s node for fabricating its new Snapdragon chips.
Likewise, Apple has close ties with TSMC, booking out its newest node for the initial few months for its new A-series chips. If Apple’s ready to pay billions of dollars to TSMC, there has to be a good reason for it.
After years of growing criticism from the Pixel community, Google finally turned to TSMC’s foundry to fabricate the Tensor G5 this year.
Given TSMC’s solid track record, it raised expectations ahead of the Pixel 10’s launch that the move would finally fix one of the Pixel lineup’s biggest flaws — poor battery life.
Tensor G5 proves TSMC alone can’t fix Google’s problems
Google fell way short with its execution
With the Pixel 10 lineup now out for over a month, it’s clear that Google’s move to TSMC hasn’t translated into a major boost in battery life. Other chips from Qualcomm and Apple still deliver far better efficiency.
Yes, the Pixel 10 runs cooler than previous Pixels, and while that’s an equally important improvement, it alone is not enough.
It wouldn’t be as disappointing if Google had used the available thermal headroom to boost the Tensor G5’s CPU performance instead.
But that’s not the case either, with the Pixel 10’s Tensor G5 failing to beat even flagship Snapdragon chipsets from two years ago. The company stuck to older ARM CPU cores, instead of adopting the latest option available.
Worse, the new PowerVR GPU has turned out to be a major disappointment, delivering average to below-par gaming performance.
The unusually low benchmark scores suggest outdated drivers on the Pixel 10 could be to blame. But until Google acknowledges and addresses the issue, nothing is certain.
The Tensor’s AI performance is impressive, but that alone won’t be sufficient in the long run. Plus, when you spend $1,000 on a smartphone, you expect to play all popular Android games on it smoothly.
Google didn’t capitalize on the Tensor G5’s TSMC advantage
A new foundry, but the same old Pixel problems
Despite using one of TSMC’s latest process nodes, the Tensor G5 falls short on all key metrics.
Its performance lags far behind the latest flagship chips from Apple and Qualcomm, and the GPU is underwhelming. And the new Pixels don’t even offer standout battery life to make up for the average performance.
The Tensor G5 marks Google’s fifth-generation Tensor chip, yet it continues to suffer from many of the same problems as the original.
In contrast, Xiaomi’s new in-house flagship SoC — the Xring O1 — launched earlier this year using the same 3nm TSMC process as the Tensor, but manages to rival Qualcomm’s Snapdragon 8 Elite in performance.
If anything, this highlights Google’s lack of ambition and effort with its Tensor lineup.
When a company like Xiaomi can nearly match the best chips on the market with only its second in-house SoC, there’s little excuse for Google still struggling to catch up after five generations.
As a longtime Pixel fan, I hoped the Pixel 10’s Tensor G5 would finally fix the lineup’s long-standing issues with performance and battery life. I wasn’t expecting flagship-level power, but I did think improved endurance would make up for it.
Sadly, that’s not the case either, with the Pixel 10 Pro XL delivering above-average battery life.
This would have worked a couple of years ago, but not now, when the likes of the OnePlus 13 and other Android phones can easily survive 1.5 days of heavy use.
At this point, it’s hard to expect Google to make any major leaps with its Tensor chips. The company seems content on focusing on AI, and it’s likely using it as a reason to sidestep meaningful performance gains elsewhere.
