Tensor G4 vs Tensor G3 – Here Are All The Differences You Need To Know Between Google’s Pixel Smartphone Chipsets Of Two Generations

The Tensor G3 debuted in Google’s Pixel 8 and Pixel 8 Pro last year, and while the company introduced a range of improvements, the chipset turned out to be significantly slower than the competition. As is typical of successive releases, the Tensor G4 is expected to bring various upgrades, but how extensive will they be? Here, we look at all the differences between the two SoCs so you can decide if upgrading to any of the forthcoming Pixel 9 models is worth it.

Tensor G4 vs Tensor G3 - Specification differences show a downgrade in the total number of cores

On paper, it would appear that the Tensor G4 is slightly downgraded compared to the Tensor G3, as a leaked CPU configuration of the upcoming SoC revealed that it features a ‘1 + 3 + 4’ cluster. In contrast, the Tensor G3 has a ‘1 + 4 + 4’ configuration, which means that the Tensor G4 will potentially deliver lower multi-core performance, right? Not exactly, because Google relies on faster cores to complete the job. For instance, instead of the Cortex-X3 and Cortex-A715, which were a part of the Tensor G3, the upcoming silicon should get its performance boosted thanks to the inclusion of the Cortex-X4 and Cortex-A720.

Related Story Tensor G4 Delivers Mixed Performance Across Various Pixel 9 Models, Base Version Slower Than Pixel 8 In Multi-Core, Pixel 9 Pro Barely Faster Than Pixel 8 Pro

As for the low-power cores, the Cortex-A520 will boast better capabilities than the Cortex-A510, assuming the Tensor G4 can keep its thermals in check. Coming to the GPU side of things, future buyers wanting to get their hands on any Pixel 9 model will be disappointed to learn that the Tensor G4 and Tensor G3 share the same ARM Mali-G715 GPU, with the only difference being that the newer SoC’s graphics processor is clocked at a slightly higher 940MHz.

Based on a multitude of tests that we have seen in the past, a slightly overclocked GPU does little to bump real-world performance, so the actual benefit will likely be witnessed when