Technology news and discussion on TalkFuse.
Google chose Tensor chips to power Pixel AI features and reduce reliance on Qualcomm
Google replaced Qualcomm’s Snapdragon chips with its own Tensor platform in the Pixel 6 in 2021, aiming to run speech recognition, translation and computational photography directly on the phone. The move lets Google shape hardware around its AI models, including Gemini, while reducing its reliance on cloud processing and costly third-party chips.
Tensor’s first four generations were heavily based on Samsung’s Exynos designs; Google shifted to its own chip design with the Tensor G5, made by TSMC, and the G6 continues that approach. Google says the G6 offers 50 per cent more TPU computing power than the G5, while independent testing found the G5 substantially behind Qualcomm’s leading chip in CPU and graphics performance. Tensor also helps Google avoid the reported $240 to $280 cost of a current top Snapdragon phone chip, but Google still uses Qualcomm silicon in products such as the Pixel Watch 5.
In short
- Google chose Tensor to tailor Pixel phones to its AI features.
- The G6 is claimed to deliver 50 per cent more TPU computing power.
- Snapdragon remains faster in CPU and graphics tests.
Steelman: both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
Google’s Tensor strategy gives it tighter control over how hardware supports Pixel features such as on-device speech recognition, translation and photography, while potentially improving privacy and reducing reliance on cloud processing. Designing around its own AI models may also let Google differentiate Pixel phones and avoid some of the cost and supplier dependence associated with top-end third-party chips.
The case against
Relying on a proprietary chip can make Pixel performance depend on Google’s ability to match specialist chipmakers in areas such as CPU and graphics, where testing found the G5 lagged Qualcomm’s leading chip. The benefits of custom AI hardware may not outweigh weaker general performance or the cost of developing and maintaining the platform, and continued use of Qualcomm silicon in other products shows that supplier dependence is not eliminated.
Explained simply
Google’s Tensor is a processor designed for Pixel phones. Google introduced it in 2021, replacing Qualcomm’s Snapdragon chips in its phones, with the aim of running tasks such as speech recognition, translation and photo processing on the device.
The first four Tensor generations drew heavily on Samsung’s chip designs. From the Tensor G5, Google moved to its own design, manufactured by Taiwan-based TSMC; the G6 follows that approach.
Google can tailor Tensor to its software and AI models, and rely less on processing in the cloud or buying top-end phone chips from Qualcomm. Tests have found the G5 behind Qualcomm’s leading chip in some general performance measures, and Qualcomm silicon still appears in other Google products, including Pixel watches.
Read the full article at the source →
Originally published by Engadget as “Why did Google Pixels stop using Snapdragon chips?”.