Home / Technology / Google's Pixel's AI Focus: Ditching Snapdragon for Custom Silicon
Google's Pixel's AI Focus: Ditching Snapdragon for Custom Silicon
15 Nov
Summary
- Google replaced Snapdragon with its own Tensor chip for enhanced AI.
- Tensor G6 offers 50% more TPU compute than the G5.
- Qualcomm Snapdragon chips outperform Tensor in raw power and graphics.
Google shifted its Pixel smartphone strategy in 2021, replacing Qualcomm's Snapdragon processors with its custom-designed Tensor chips to bolster artificial intelligence capabilities. This move aimed to improve on-device AI functions like speech transcription and translation, driven by the Tensor Processing Unit (TPU).
The Tensor platform, initially a collaboration with Samsung, has evolved significantly. The recent Tensor G6 chip, fabricated by TSMC, reportedly offers 50 percent more TPU compute than its predecessor, the G5. It also claims substantially faster and more energy-efficient AI task performance when paired with Gemini Nano.
However, comparative tests indicate that flagship Snapdragon chips maintain a lead in raw CPU and graphics performance. Google appears to accept this benchmark deficit as a trade-off for prioritizing its AI features and controlling its chip roadmap. Notably, Google's Pixel Watch 5 series continues to utilize Qualcomm's Snapdragon chips, as do certain other wearable devices.
This strategic decision allows Google to avoid the significant costs associated with Qualcomm's flagship chips, estimated between $240 to $280 per unit for the latest generation. This diverted budget supports the refinement of Tensor's core AI strengths, positioning Google's services for its unique machine learning models.
