Google Tensor G5: Benchmarks and everything you need to know

For the past five years, Google Pixel phones have been powered by an in-house system-on-a-chip (SoC) designed by Google. The company's chips have come a long way since then, culminating in the Tensor G5 platform that powers the Google Pixel 10 series. Tensor G5 features more custom components than ever, improving performance and optimization, all while using TSMC's industry-leading 3nm process node.

Tensor G5 doubles down on what Google silicon does best, refining performance for everyday tasks, improving stability and battery life, and adding power for on-device AI processing. In a way, it also doubles down on what Google silicon isn't — Tensor G5 is not a benchmark-killer and will always lag behind Qualcomm and Apple designs. With that being said, it's upgraded in all the right places, making the Google Pixel 10 series a joy to use.

Here's everything you need to know about the Tensor G5 mobile platform, including benchmark tests against all the best Android phones and processors.

Tensor G5: Features, specifications, and upgrades

Tensor G5 marks the highly anticipated switch to TSMC as Google's chip manufacturing partner. The first four generations of Tensor chips were made by Samsung Foundry, which has struggled to keep up with TSMC's advanced fabrication processes. Specifically, the Tensor G5 chip is manufactured using TSMC's 3nm process node — the same one that serves as the foundation for Qualcomm's Snapdragon 8 Elite and Apple's A18 Pro.

Google explains that this new manufacturing technology helps it "pack more transistors into the chip so it’s more powerful and efficient." The company claims the Tensor G5's CPU is 34% faster on average than the Tensor G4. It's even better for AI tasks, as the Tensor G5 tensor processing unit (TPU) is up to 60% more powerful. There's no official word on how much better graphics performance is compared to Tensor G4, but you'll see in our tests that there are clear improvements.

Some of the new Tensor G5 components include a new image-signal processor (ISP), which is expected to result in better photos and videos. Additionally, there's a new display controller, which may be a contributing factor to Google adding PWM sensitivity toggles for the first time on the Pixel 10 Pro. The upgrades extend beyond raw performance, although speeds are definitely improved as well.

Lastly, Google touts Tensor G5 as being capable of running the newest Gemini Nano model on-device. Beyond that, this new Gemini Nano model runs 2.6x faster and 2x more efficiently, according to Google.

Tensor G5: Benchmark tests and comparisons

Past generations of Tensor processors resolved critical issues like overheating and stability. Now, Pixel users are looking to the Tensor G5 to bring performance that's finally on par with Apple and Qualcomm chipsets. Does the Pixel 10, powered by Tensor G5, live up to the hype? The answer is mixed, because while the Tensor G5 platform makes the Pixel 10 better for daily use than ever, it still lags behind in synthetic benchmark tests.

In CPU-heavy benchmarks like Geekbench 6, we see both the Pixel 10 and Pixel 10 Pro models produce slightly higher results than the Pixel 9 series. The Pixel 10 Pro puts up better numbers than the base Pixel 10, even though both are powered by Tensor G5. This is likely because the Pixel 10 lacks the vapor chamber used to cool the Pixel 10 Pro and Pixel 10 Pro XL, resulting in better performance and stability on the more expensive models.

While the Tensor G5 chip beats both the Tensor G4 and Samsung Exynos 2500 processors in single-core tests, the Exynos 2500 actually pulls ahead in multi-core performance. The Tensor G5 platform falls short of even coming close to either the Qualcomm Snapdragon 8 Elite or the Snapdragon 8 Elite for Galaxy processors.

Even the Tensor G5's TPU doesn't make a splash in benchmark test, with memory being the determining factor in AI performance. With just 12GB of RAM, the base-model Pixel 10 loses in the Geekbench AI benchmark to both the Snapdragon 8 Elite for Galaxy and Exynos 2500 chips. The Pixel 10 Pro fares better with 16GB RAM, beating Samsung's chip but still falling short of Qualcomm.

Notably, the Tensor G5 appears to be very underpowered for gaming, according to benchmark results. Neither the Pixel 10 nor the Pixel 10 Pro could break the 20FPS barrier in 3DMark's Wild Life Extreme test, whereas the winner (Snapdragon 8 Elite) posted a result of 38.27FPS. This continues a trend of Pixel phones, and Tensor processors, being very weak where graphical performance is concerned.

Tensor G5: Daily use and performance

Of course, benchmarks only tell part of the Tensor G5 story. The more important part is arguably how the silicon fares in daily use of the Pixel 10, Pixel 10 Pro, and Pixel 10 Pro XL. There's good news and bad news in that department. While our testing of the Pixel 10 Pro and Pixel 10 Pro XL revealed gaming improvements — it ran Minecraft and Genshin Impact at 60FPS, whereas the Pixel 9 Pro would be lucky to hit 40FPS — those are still lackluster figures for games that aren't that demanding.

If mobile gaming is part of your "daily use" of a smartphone, you might want to steer clear of a Tensor G5-powered device. It just isn't going to compete with a Qualcomm or Apple-powered smartphone in terms of graphics. However, if you're not a big smartphone gamer, the Tensor G5 chip seems a lot more appealing.

It adds quality-of-life features like Magic Cue and Camera Coach, improves CPU and TPU performance, and upgrades camera quality with a new ISP. None of those changes is going to make a material difference on a spec sheet or a benchmark result, but they do make an impact in everyday use of a Google Pixel 10.