Recent benchmark leaks of Google's upcoming Tensor G5 processor have revealed mixed results, suggesting that while Google's next-generation chip might not compete with top performers in raw processing power, it could offer unique advantages in specific areas like gaming performance.
Architecture and Manufacturing Process
The Tensor G5, set to power the Pixel 10 lineup in late 2025, will feature a novel '1 + 5 + 2' CPU cluster configuration. Google has opted to manufacture the chip using TSMC's advanced 3nm 'N3E' architecture, incorporating Integrated Fan-Out technology for improved power efficiency. The processor will utilize a Cortex-X4 as its primary core, complemented by five Cortex-A725 cores and two Cortex-A520 cores for efficiency.
Benchmark Performance
Early Geekbench 6 results for the Google Frankel device, running Android 15, have shown concerning figures. The chip scored 1,323 in single-core performance and 4,004 in multi-core tests. These numbers fall notably short of expectations, particularly when compared to the current Pixel 9's scores of 1,800 and 4,573 respectively. However, it's important to note that these results likely represent an early engineering sample, and performance could improve significantly before final release.
Gaming Capabilities
Despite the seemingly disappointing CPU benchmark scores, there's a silver lining for gaming enthusiasts. According to insider reports, the Tensor G5's GPU capabilities might actually surpass those of the Snapdragon 8 Elite, potentially making the Pixel 10 a more compelling option for mobile gaming. This aligns with Google's strategy of optimizing their chips for specific use cases rather than pursuing raw performance metrics.
Strategic Direction
Google's approach with the Tensor series continues to prioritize user experience over benchmark supremacy. A Google executive previously acknowledged that their chips aren't designed to break performance records but rather to enhance the overall user experience through optimized software integration. This strategy, while practical, may pose challenges for resource-intensive applications like on-device Large Language Models and AAA gaming titles.