Even though the launch of the Pixel 10 series is far away, we’re already starting to get some juicy details about what to expect with Google’s next flagships. According to a report by Android Authority, Google’s upcoming Tensor G5 chip, set to power the Pixel 10, will bring some noticeable changes, though not everything might meet expectations.
For the first time, Google has built the Tensor G5 completely in-house, moving away from Samsung’s involvement in previous designs. This shift is meant to make the new chip more competitive, as earlier Tensor processors have lagged behind rivals in performance and efficiency.
The Tensor G5’s CPU layout has been tweaked, but some decisions might raise eyebrows. Google has stuck with a single Cortex-X4 core, skipping the newer Cortex-X925, despite the potential improvements. Here’s the full CPU breakdown:
- Big Core: 1x Cortex-X4
- Mid Cores: 5x Cortex-A725 (up from 3 on the Tensor G4)
- Little Cores: 2x Cortex-A520 (reduced from 4)
These adjustments could improve multi-core performance slightly, but the reuse of the Cortex-X4 suggests Google is playing it safe. The most surprising change lies in the GPU. Instead of the usual Arm Mali cores, Google is switching to a GPU from Imagination Technologies. The new IMG DXT-48-1536 will run at 1.1 GHz and bring two major upgrades: Raytracing support and GPU virtualization.
Tensor chips have always prioritized AI, and the G5 is no exception. Google has upgraded the TPU (Tensor Processing Unit), boasting a 40% increase in operations per second (TOPS). However, real-world performance gains are limited to just 14% according to internal benchmarks. The new TPU also supports on-device training and includes RISC-V cores for additional flexibility.
The Tensor G5 will be fabricated on TSMC’s 3nm N3E process, the same technology used in Apple’s A18 Pro. However, the G5 is physically larger, measuring 121 mm² compared to Apple’s 105 mm², which could impact thermal performance and efficiency.
While the Tensor G5’s specs might not blow anyone away, most enthusiasts claim the real strength of Google’s chips lies in software optimization. However, I’d like to point out that in some cases, you need all the processing power. This is where the new Snapdragon 8 Elite might blow the Tensor G5 out of the water, like it does with the Tensor G4.