Google has just unveiled its next-generation smartphones, the highly anticipated Pixel 9 and Pixel 9 Pro. At the heart of these devices is the latest iteration of Google’s in-house chip, the Tensor G4. And the first benchmarks have just been released, giving us a — very mixed — glimpse of the new SoC’s performance.
A timid performance gain compared to the Tensor G3
Since Google decided to design its own chips in 2021 with the original Tensor, the goal has always been to deliver the best possible experience on Android, focusing in particular on artificial intelligence and computational photography. With the Tensor G3 last year, Google managed to improve overall performance while remaining substantially behind the competition, both Android and the iPhone. But what about the Tensor G4 in the Pixel 9? The early results are mixed.
It’s crazy how a 4 year old iPhone 12 still outperforms Google’s upcoming Pixel 9 Pro XL. ? pic.twitter.com/59Xl7QE3h2
— sid (@immasiddtweets) August 13, 2024
According to Geekbench benchmarks, the Pixel 9 Pro XL with the Tensor G4 and 16GB of RAM achieves a single-core score of 1950 points and a multi-core score of 4655 points. This is about 10% better than the average score obtained by the previous generation Tensor G3. So the gain is quite modest, especially compared to the leaps made by the competition.
Indeed, the Tensor G4 is still far behind Qualcomm’s Snapdragon 8 Gen 2 that powers most Android flagships this year. And let’s not even talk about the Apple Bionic chips in iPhones that still largely dominate benchmarks. For comparison, the Tensor G4’s single-core score is on par with the A14 Bionic in the iPhone 12s released almost four years ago.
AI and photography still at the heart of the Pixel experience
Of course, benchmarks aren’t everything and don’t necessarily reflect the everyday user experience. And this is precisely where Google intends to stand out with its Pixels. The Tensor G4 brings its share of new features and optimizations, particularly for processing related to artificial intelligence.
Google has also announced that the Pixel 9 will have part of its RAM dedicated to the new Gemini Nano AI model. This should allow for faster processing of AI-related tasks, such as voice recognition, real-time translation and advanced photo editing functions.
Likewise, the Tensor G4 still has its dedicated image processing chip, which should once again work wonders in photography, especially in low light. Pixels have been known for years to be among the best camera phones on the market and this new generation should be no exception, despite fairly poor benchmarks on paper.
In the end, even if the Google Pixel 9’s Tensor G4 doesn’t impress on paper, it will definitely have to be the software experience and AI features that it intends to stand out. Although not all AI processing is done locally, the Pixel 9’s relative power lag doesn’t affect its features at all. Apple has everything it needs to make the experience of its devices much better, and it will have to step up its game on Apple Intelligence and the iPhone A16 to keep up with the competition.