Google’s New Tensor G5 Chip for Pixel Phones Has a Major Problem – Here’s Why

Google’s new Tensor G5 chip, built for its latest Pixel 10 phones, is facing criticism from early testers and analysts for frequent throttling and heat buildup, suggesting deeper problems in how the chip was designed.

Wccftech^[1] reported that the Tensor G5 heats up quickly during intensive workloads such as gaming, emulation, and AI processing. This leads to aggressive throttling, where the processor slows down to manage heat, reducing overall performance. In some cases, the chip struggles even with tasks that rely primarily on CPU power and have little to do with the GPU, such as emulating PlayStation 2 games

Lack of Optimization

At its core, the Tensor G5 uses an eight-core CPU made up of one high-performance Cortex-X4 core at 3.78 GHz, five medium-performance Cortex-A725 cores at 3.05 GHz, and two Cortex-A520 efficiency cores at 2.25 GHz. It also includes a fifth-generation TPU for AI workloads, an Imagination IMG DXT-48-1536 GPU, and a Samsung^[2] Exynos 5G modem.

The chip is manufactured using TSMC’s 3nm process, which should, in theory, improve efficiency. However, the performance inconsistencies suggest that Google’s integration of these different components has not been seamless. Each part of the chip, the CPU, GPU, modem, and TPU, comes from different partners, creating a design that is technically functional but also an unoptimized Frankenstein’s monster.

GPU Choice Adds to the Problem

Google’s decision to move from ARM’s Mali GPU to Imagination’s PowerVR-based DXT-48-1536 GPU may have also contributed to the issue. While capable on paper, this GPU operates under strict proprietary control by Imagination, meaning Google^[3] cannot independently fine-tune drivers or optimize performance beyond limited adjustments.

As a result, Google has little flexibility in managing power distribution, heat control, and rendering performance, leaving the chip vulnerable to throttling when pushed to its limits.

Is it a Problem for Google?

But is this performance issue really a problem for Google? Probably not, since Tensor chips were never meant to be performance powerhouses but smart chips that maximize AI performance, which is something Pixel phones already excel at. There is a reason why Google Pixel phones have always been called the “smartest phones in the room.”

Even Google’s latest Pixel 10 launch event was primarily focused on smart features rather than raw hardware performance, which signals that Google cares little for the latter. Pixel fans don’t buy Google’s phones for gaming, but for their camera performance, smart features, and the pure Android experience.