NTD 3 nanometer process, DeepMind optimization support, Tensor G5 chip big

When the Tensor chip was first launched five years ago, Google expanded the possibilities of Pixel phones with advanced machine learning. Google's Vice President of Hardware Peng Yu-ling claimed that the new fifth-generation self-developed chip...


When the Tensor chip was first launched five years ago, Google expanded the possibilities of Pixel phones with advanced machine learning. Google's Vice President of Hardware Peng Yu-ling claimed that the new fifth-generation self-developed chip Tensor G5 is the biggest upgrade in the year, showing the NTD 3-nanometer process and the advantages of cooperation with Google DeepMind.

Tensor G5 is produced using 3-nanometer telco, which allows more transistors to be packaged in wafers. Google has upgraded the Tensor G5 in all aspects, including a TPU (Tensor Processing Unit) performance increase up to 60%, and an average CPU speed increase of 34%, making Pixel 10's daily applications such as web browsing and the latest AI features more agile.

In close cooperation with Google DeepMind, the Tensor G5 became the first chip to execute the latest Gemini Nano model on the device side, unblocking more than 20 generative AI experiences to run on the device side. Gemini Nano has increased speed and efficiency by 2.6 times faster and 2 times more efficient in use cases such as Pixel 10 screenshots and recording audio.

Pixel exclusive features such as Magic Cue, Voice Translate, call notes and Pixel Journal are all driven by Tensor G5. Tensor G5 is paired with Gemini Nano and enhances user familiarity features, such as Scam Detection and editing using Gboard natural voice.

Tensor G5 has a new built-in ISP (Image Signal Processor) that brings excellent shooting performance. Even in low-light environments, ISPs are pre-designed to provide dynamic deblurring and 10-bit videos (applicable to 1080p and 4K 30fps) to capture the highest quality images and improve the Real Tone for natural renderings to everyone in photos and videos.

Tensor G5 also injects upgraded power into camera functions, including Add Me (taken together), Auto Best Take (automatic perfect photo), and Pro Res Zoom (professional analysis focus). The Pixel 10 camera app is the first mobile phone with built-in C2PA content certification. All processes are executed on Tensor G5 and Titan M2 security chips. Secure metadata is established in the image, and the generation and editing process is recorded in the complete recording.

Even with new features, the Tensor G5 helps the Pixel 10 series battery to last more than 30 hours.

Upgrade to save using UFS 4.0

Since the launch of the Pixel 6 series in 2021, Google has been using UFS 3.1 (universal flash memory storage), which was standard for Android phones at that time, but soon after, the flagship phone was upgraded to UFS 4.0, bringing huge progress. First of all, the reading speed is greatly improved. UFS 3.1 supports 2,100MB/s reading speed and 1,200MB/s writing speed, while UFS 4.0 is 4,200MB/s reading speed and 2,800MB/s writing speed respectively. In addition, UFS 4.0's energy efficiency has also been significantly improved.

Now, the new Pixel 10 series is paired with UFS 4.0 and LPDDR5X, making the Tensor G5 have a high-speed memory interface, making the overall mobile phone run faster.

It is worth noting that Google uses "Zoned UFS" in the Pixel 10 Pro XL and Pixel 10 Pro Fold above 512GB. Google has not yet explained the details.

Google did more than just switch to TSMC for Tensor G5 — This is what’s new Google Pixel 10 series finally upgrades to faster UFS 4.0 storage, ‘Zoned UFS’ on Pro

Recommend News