Google made headlines this week when it announced its first-ever custom SoC, Tensor, for its Pixel 6 smartphone due for release this fall. With this move to in-house silicon, Google will be following the path of companies like Apple that are creating their own custom chips.
The upcoming Pixel 6 will contain Google’s first custom SoC. Image used courtesy of Google
Details are scarce at the moment on the hardware specifications for this new mobile computer, but Google promises those details are coming closer to the smartphone’s release.
What We Do Know About Tensor
Tensor is intended to bring AI features to smartphones—making it a “mobile AI computer”—and ready the Android OS for ML functions, according to Rick Osterloh, Google’s senior VP for devices and services.
The chip is said to enhance some existing features of the Pixel 4—for instance its lens suggestion and suite of audio recording capabilities. Lens suggestions can use the phone’s camera to translate foreign languages in real-time. Additionally, Google’s photography computational models capture professional-grade images on a cell phone.
One of the on-device AI features Tensor will provide is foreign text translation. Image used courtesy of Google
Google’s press release on Tensor reports that the chip features a new security core along with Titan M2, Google’s security module, giving it the “most layers of hardware security in any phone.”
Google has historically turned to Qualcomm for its Snapdragon 765G processor; why is the tech giant getting into custom SoC development now? And where will this trend of in-house hardware take the industry in the coming years?
How Google Silicon May Affect Supply Chains
While chip shortages have been well documented in the automotive industry over the past year, manufacturing delays have begun to affect the smartphone industry as well.
Google’s decision to develop an in-house system-on-chip has the potential to put an increased strain on an already precarious supply chain system, which has been subject to hoarding by telecom manufacturers like Huawei, Apple, and others.
Chip demand over the past year. Image used courtesy of Bloomberg
The trend toward in-house optimized hardware could signal further fragmentation of the software and firmware ecosystems across not only the mobile industry but also for computing in general. Meanwhile, Qualcomm is hard at work developing next-generation RF hardware for all connected technology.
Bringing Tensor Processing to Mobile
Now, Google is seeking to further unlock the power of AI and ML, which means optimizing hardware for those applications. Cloud Tensor Processing Units (TPUs) are custom ASICs designed for machine learning applications. TPUs are said to be used for ML models dominated by matrix computations and large data batch sizes.
Designed for low-power devices, Google’s Edge TPUs provide neural network processing for low-power devices. Image used courtesy of Coral.ai
Was this custom cloud-based ASIC the first step in Google’s path towards the new Tensor SoC for the Pixel 6? Will the move away from Snapdragon further complicate the firmware/software intersection for mobile development? Hopefully, starting in the fall of 2021, Google will release some answers.
For lighting, electrical, signage, and technology solutions that allow you to do more call Sverige Energy today at +4(670) 4122522.