Google Pixel 10 Pro XL performance review: Tensor G5 won’t woo you in benchmarks but impresses where it matters

• By Sanket Vijayasarathy
Published On: 3 September 2025 | 09:56 IST

I’ve been using the Pixel 10 Pro XL for less than a week now, and there are numerous AI features and other upgrades that I need to test thoroughly before I have my full review ready. One of the upgrades I feel I can talk about right now is the new Tensor G5 chipset. Google claims it is over 30 per cent faster than the previous Tensor G4, which is great if you’re coming from an earlier Pixel, but it’s still well behind the likes of the Snapdragon 8 Elite. However, what I’ve understood in the few days using the newest Pixel is that the Tensor G5 chip shows Google is on the right track in terms of real-world performance and on-device AI experiences.

Disclaimer: This isn’t a full Pixel 10 Pro XL review. I’ll cover that in detail later. Here, I’m focusing on the new Tensor G5 chipset and how it impacts performance, battery life, and AI features.

Numbers tell you half the story

Let’s get one thing out of the way: Tensor chips have never been great in benchmarks, and the new Tensor G5 sticks to tradition. Our smartphone testing team ran the Pixel 10 Pro XL through various benchmark apps, such as Geekbench and AnTuTu, and the numbers were underwhelming compared to those of rival phones like the iPhone 16 Pro Max, Vivo X200 Pro, and Galaxy S25 Ultra. Where the Pixel 10 Pro XL (and Pixel 10) score under 15 lakh on AnTuTu, rival phones with the Snapdragon 8 Elite and MediaTek Dimensity 9400 surpass 20 lakh. Even the iPhone 16 Pro Max’s A18 chipset crosses 17 lakh on AnTuTu. And you’ll find similar differences in numbers on Geekbench as well.

But the Tensor G5 isn’t designed for raw benchmark numbers. It’s designed to handle on-device AI tasks, which are aplenty on the new Pixel phones, and improve power and thermal efficiency. The latter two seem to have been addressed by moving to TSMC’s 3nm process from Samsung Foundry.

The Tensor G5 isn’t designed for raw benchmark numbers. It’s designed to handle on-device AI tasks and improve power and thermal efficiency.

TSMC is known for creating power and thermal-efficient chips that throttle less than Samsung-made Tensor chips that came before. And this is immediately evident in our CPU Throttle test, where the Tensor G5 managed to maintain around 46 per cent of peak performance under heavy workloads, significantly better than the Snapdragon 8 Elite-powered Galaxy S25 Ultra’s 21.40 per cent and last year’s Tensor G4-powered Pixel 9 Pro XL’s 39 per cent.

Tensor G5 shows some real-world improvements

If you’re not a numbers person and can conveniently ignore benchmark scores, the Tensor G5 processor feels like a solid upgrade. It has been snappy during my limited time using the Pixel 10 Pro XL as my primary phone. There’s a noticeable fluidity when opening or switching between apps, with fewer crashes in between. The phone barely gets hot while using AI features or the camera app, which seems to be due to a combination of improved thermal efficiency and the vapour chamber cooling. I have yet to test gaming on the Pixel 10 Pro personally, but our testing team ran games like BGMI and COD Mobile for 30 minutes each, noticing an average temperature rise of around 5.3 degrees Celsius. The Pixel 9 Pro XL performed slightly better with an average increase of 3.8 degrees Celsius.

The Tensor G5 is snappier than before and doesn’t throttle or heat as much, but it doesn’t feel more power-efficient, likely due to AI features like Magic Cue running in the background.

While Tensor G5 is snappier, it doesn’t feel more power-efficient than Tensor G4. For some reason, the PCMark battery test refused to work properly, so I won’t be able to provide benchmark data on how long the phone’s screen can stay on from 100 to 20 per cent. However, real-world usage may provide a better understanding. I was able to achieve anywhere between 4 and 5 hours of screen-on time with always-on display enabled and working between 5G connectivity and Wi-Fi. My usage involved testing the cameras, browsing through Instagram, streaming music on YouTube Music, and texting on WhatsApp.

Yes, the Pixel 9 Pro XL gets a bigger battery than before, and I’m sure the Tensor G5’s improved power efficiency helps the phone last beyond a full day of use. However, I also feel that the extra battery and efficiency are being consumed by AI features constantly running in the background, such as Magic Cue. As a result, you’ll still need to plug the phone in at night (or the next morning) to start the day with confidence.

AI is where Tensor G5 truly matters

Perhaps the most important aspect of the Tensor G5 will be its ability to handle the many on-device AI features that come with the Pixel 10 phones. The chip is equipped with a 60 per cent more powerful TPU (Tensor Processing Unit) used for AI tasks, and it is the first chip to run the newest Gemini Nano model, allowing for on-device generative AI experiences. Features like Magic Cue, which runs continuously in the background to pull up useful information when you need it, AI Voice Translate, and Personal Journal are some proactive, on-device AI tasks powered by the Tensor G5 chip. While I haven’t tested all these AI features in depth yet, it’s clear that Google views Tensor as a chip for AI tasks (now more than ever) rather than one that will compete with flagships from Qualcomm and Apple on raw numbers.

At the end of the day, the Tensor G5 isn’t trying to beat Snapdragon or Apple at their own game. It’s trying to play a different one – one that’s less about chasing benchmark crowns and more about enabling meaningful AI features while keeping things cool and efficient. From what I’ve seen so far, it’s a step in the right direction. If you’re buying a Pixel 10 phone, you’re not choosing it for the highest frame rates in Genshin Impact; you’re choosing it for Google’s software and AI vision, and the Tensor G5 finally feels like the kind of hardware that can keep up with that vision.