As Google prepares to celebrate the 10th anniversary of the Pixel line, all eyes are on the “Laguna” project—the internal codename for the Tensor G5 chipset. For the first time since the transition to custom silicon, Google is reportedly moving away from Samsung Foundry and partnering with TSMC to manufacture the G5 on their advanced 3nm (N3E) process node.
Breaking the Thermal Ceiling
The move to TSMC is more than just a logistical shift; it is a tactical strike against the overheating and throttling issues that have plagued previous Pixel generations. Leaked internal benchmarks suggest that while the 15% speed boost in raw CPU performance (single-core) is a respectable year-over-year gain, the real headline is the 40% jump in power efficiency.
By utilizing the 3nm architecture, the Tensor G5 is expected to consume significantly less power during intensive tasks like 4K video recording and long-exposure computational photography. Early prototype testing indicates that the Pixel 10 can maintain peak performance for nearly double the duration of the Pixel 9 before thermal throttling kicks in.
AI at the Edge: The Revamped NPU
Google’s strategy has never been about winning the “spec war” against Apple or Qualcomm. Instead, the Tensor G5 doubles down on On-Device AI. The new Neural Processing Unit (NPU) is rumored to be 60% more powerful, specifically optimized for “Anticipatory AI” features.
According to leaks, this will enable:
-
Real-time Video Magic Eraser: The ability to remove moving objects from 4K/60fps video instantly on the device.
-
Offline Gemini Ultra: A “distilled” version of Google’s most powerful LLM running locally without needing a cloud connection.
-
Proactive Battery Management: An AI co-pilot that predicts your app usage patterns to shave off unnecessary background cycles.
Tensor G5 vs. Tensor G4: The Performance Gap
| Metric | Tensor G4 (Samsung 4nm) | Tensor G5 (TSMC 3nm) |
| CPU Architecture | 1+3+4 Configuration | 1+5+2 Configuration |
| Single-Core (Geekbench 6) | ~1,900 | ~2,285 (+20%) |
| Multi-Core (Geekbench 6) | ~4,600 | ~6,200 (+35%) |
| Thermal Efficiency | Baseline | +40% Improvement |
| GPU | Mali-G715 | Imagination DXT-48-1536 |
The GPU Gamble
While the CPU and AI gains are undisputed, some leaks suggest a radical change in the graphics department. Google is reportedly ditching the ARM Mali GPU in favor of a custom solution from Imagination Technologies. This new GPU architecture is designed to handle ray-tracing and high-fidelity gaming, though initial benchmarks show it currently trails behind the Snapdragon 8 Elite’s Adreno GPU. Google appears to be prioritizing sustained stability over peak frame rates, aiming for a “jitter-free” experience rather than raw benchmarks.
