Identifying the Flaws in Google Pixel’s New Tensor G5 Chip

Despite the anticipation surrounding Google’s latest Tensor G5 chip, its release has not met expectations from both tech enthusiasts and consumers. The primary concern under scrutiny is the chip’s tendency to throttle, which detracts from its overall performance. Analysts suggest that this issue may stem from Google’s fragmented approach to its architecture.

Understanding the Architecture of Google’s Tensor G5 Chip

The Tensor G5 chip features a complex architectural design, which includes the following components:

An Eight-Core CPU:
- One high-performance Cortex-X4 core operating at 3.78 GHz.
- Five medium-performance Cortex-A725 cores running at 3.05 GHz.
- Two efficiency-driven Cortex-A520 cores clocked at 2.25 GHz.
A Fifth-Generation TPU: This component is dedicated to managing machine learning and artificial intelligence tasks effectively.
An Imagination IMG DXT-48-1536 GPU: This integrated PowerVR-series GPU operates at 1.10 GHz and offers theoretical performance on par with leading mobile GPUs like the Adreno 732/740 and ARM Mali G715 MP7, albeit without ray-tracing capabilities.
A Samsung Exynos 5G Modem: This modem enhances connectivity options.

Fabricated on TSMC’s advanced 3nm process node, the Tensor G5 chip promises increased transistor density along with improved performance and energy efficiency.

Challenges of Google’s Fragmented Chip Design Approach

Recent discussions have highlighted the Tensor G5 chip’s issues with overheating and throttling, particularly during heavy-duty gaming sessions. This limitation raises significant concerns about gaming performance and overall user experience.

While the transition from the ARM Mali GPU to the Imagination IMG DXT-48-1536 GPU has drawn some criticism, it does not fully account for the performance issues observed. Intriguingly, the Tensor G5 has demonstrated throttling issues even during simpler tasks, such as PlayStation 2 emulation, which relies predominantly on CPU power rather than GPU resources.

In comparison, Qualcomm’s Snapdragon 8 Elite Gen 5 easily outshines the Tensor G5 in recent Geekbench 6 and 3DMark benchmarking tests. This performance superiority can be attributed to Qualcomm’s use of custom Oryon CPU cores, with the primary core clocked at 4.60 GHz and additional performance cores at 3.62 GHz. Moreover, Qualcomm has incorporated numerous optimizations, including enhanced L2 cache management, with both core types featuring a robust 12 MB of L2 cache.

In stark contrast, Google’s use of standard ARM Cortex CPU cores lacks the same level of refinement and optimization tailored for the Pixel 10. Furthermore, even though Google collaborated with Imagination to develop the IMG DXT-48-1536 GPU, it still relinquishes complete control over foundational driver updates and hardware-specific code to Imagination, as highlighted in discussions on driver management.

The analogy of Google’s chip design approach is reminiscent of purchasing a pre-made suit with minor alterations—functional yet lacking the appeal and finesse of a bespoke design. If Google prioritizes cost-effectiveness over comprehensive optimization in its chip design strategy, it may continue to trail behind competitors in pure performance, notwithstanding the presence of transformative elements like the TPU.

Source & Images