Pixel Visual Core
Google’s programmable image and vision processor for Pixel phones
The Pixel Visual Core is a system in package image processor designed by Google for its Pixel smartphones. It uses a custom multi-core architecture to handle computational photography and AI tasks with high energy efficiency.
- Designed by Google in partnership with Intel over four years
- Manufactured by TSMC using 28HPM HKMG process
- Features 8 IPU cores with 512 ALUs each
- Supports Halide language for image processing
- First appeared in Google Pixel 2 in October 2017
What it is
The Pixel Visual Core (PVC) is a specialised chip designed by Google to handle image processing and artificial intelligence tasks on mobile devices. It first appeared in the Google Pixel 2 and 2 XL smartphones in October 2017. Unlike standard image signal processors that have fixed functions, the PVC is fully programmable. This allows it to adapt to new computational photography techniques without needing hardware changes. The chip is built as a system in package, meaning multiple components are combined into a single unit. It serves as a domain-specific architecture tailored for vision and AI workloads.
How it came about
Google previously relied on Qualcomm Snapdragon components for image processing in its Nexus and early Pixel devices. As computational photography became more important, the company sought a more efficient solution. They developed the PVC to reduce power consumption compared to using general-purpose CPUs or GPUs. The design process took four years and involved a partnership with Intel. The project was codenamed Monette Hill. The goal was to create a flexible processor that could handle complex image algorithms while saving battery life. It debuted in the Pixel 2 series and continued in the Pixel 3 before being replaced by the Pixel Neural Core in the Pixel 4.
Inside the chip
The PVC architecture is scalable and can range from two to sixteen cores. Its main building block is the Image Processing Unit (IPU). Each IPU contains a Stencil Processor (STP) which features a two-dimensional array of processing elements. These elements perform calculations on small groups of pixels. The design includes a Line Buffer Pool to store data temporarily, reducing the need for energy-intensive memory access. The chip uses a ring network to connect cores, which helps save power. It supports two instruction sets: a virtual one inspired by RISC-V and a physical Very Long Instruction Word (VLIW) format. Programs are written in Halide, a language that separates algorithm logic from execution scheduling.
How it was made
The SR3HX version of the Pixel Visual Core was manufactured by TSMC. It uses the 28HPM HKMG process node. The chip is packaged as a system in package (SiP). This specific design includes an ARM Cortex-A53 CPU, eight IPU cores, and 512 MB of LPDDR4 memory. Each IPU core contains 512 arithmetic logic units arranged in a 16 by 16 grid. The manufacturing process allows the chip to operate at 426 MHz. The integration of these components into a single package helps manage the complexity of the memory system and data movement.
Where it went
The Pixel Visual Core was used in the Google Pixel 2, Pixel 2 XL, Pixel 3, and Pixel 3 XL smartphones. Its primary role was to accelerate HDR+ photography and other computational imaging features. By offloading these tasks from the main processor, it improved performance and reduced energy use. Google stated that HDR+ could run five times faster on the PVC than on the Snapdragon 835 processor. The chip also supports TensorFlow for machine learning tasks. It helped the Pixel 2 and Pixel 3 achieve high scores in mobile camera benchmarks, with the Pixel 3 tying for the top single-lens score.
Versions and family
The first variant, found in the Pixel 2 series, is labeled SR3HX X726C502. The version in the Pixel 3 series is labeled SR3HX X739F030. Both variants share the same basic architecture but may have minor revisions. The PVC was designed to be either a standalone chip or an intellectual property block for larger systems on a chip. It was succeeded by the Pixel Neural Core starting with the Google Pixel 4. This successor continued the trend of using dedicated hardware for AI and image processing tasks.
Why it matters
The Pixel Visual Core demonstrated the value of domain-specific architectures for mobile photography. It allowed Google to implement advanced image processing algorithms that were too costly for general-purpose processors. The chip claimed to be seven to sixteen times more energy-efficient than the competing Snapdragon 835 for certain tasks. It could perform three trillion operations per second. This efficiency enabled features like HDR+ to run quickly without draining the battery. The use of the Halide programming language also showed how software tools could be tailored to specific hardware designs.
Questions people ask about Pixel Visual Core
What is the Pixel Visual Core used for?
It is an image processor designed for mobile devices. It handles computational photography tasks like HDR+ and runs machine learning models using TensorFlow. It is more energy-efficient than using the main CPU or GPU for these specific jobs.
Which phones used the Pixel Visual Core?
The chip was included in the Google Pixel 2, Pixel 2 XL, Pixel 3, and Pixel 3 XL. It was replaced by the Pixel Neural Core in the Pixel 4 and later models. It was not used in earlier Nexus devices.
How is the Pixel Visual Core programmed?
Developers write programs for the PVC using the Halide language. This language allows them to separate the algorithm from the execution schedule. The code is compiled into a virtual instruction set and then into a physical VLIW format for the hardware.
Good reading and sources
- Pixel Visual Core on WikipediaEncyclopedia
- More photos on Wikimedia CommonsPhotos
- Wikidata recordOpen data
The facts on this page come from the Wikipedia article "Pixel Visual Core", written by its contributors and used under the CC BY-SA 4.0 licence, and from its Wikidata record (CC0). EveryChip rewrote the text in its own words, with AI assistance, and compares every page with the article to make sure no passage is copied. It may be reused under the same CC BY-SA 4.0 licence with a link back.
It may contain mistakes, and figures change over time. Nothing here is investment or buying advice. If something is wrong, tell us. Last refreshed 2026-11-14.
Like Pixel Visual Core
All systems on a chip
🇺🇸 United StatesApple A11 Bionic
Six-core phone SoC with a Neural Engine, used in iPhone 8 and iPhone XSoC · Arm · 2017