Workload and traffic model
Build a traffic matrix across the documented CPU, graphics, AI and I/O engines. Model concurrent bursts, locality and worst-case response time at shared memory before choosing arbitration or buffering.
GOOGLE / SILICON DESIGN
Google’s TSMC 3 nm Pixel 10 SoC, with a faster TPU/CPU and new ISP.
Original source · Retrieved 2026-09-09
Chipwise / NanoPhysics · photographer · Included in this image
Image courtesy of Chipwise · Used with permission. Link the original Chipwise source page and retain its markings and credits. Permission reported by the Museum operator on 2026-09-09. Display resizing and format conversion only; no removal of annotations or watermarks. No onward redistribution or comparison-export licence is granted by the Museum.
Google Tensor G5. Backside/deprocessed die view, as identified by source
Source photography; preparation and visible markings are preserved. This photograph supplies no independently verified geometric measurements.
| Identity & fabrication | |
|---|---|
| Process | Manufacturer / primary sourceGoogle Scope: silicon-die |
| CPU & clock domains | |
| CPU performance claim | Manufacturer / primary sourceGoogle Scope: silicon-die Google’s comparison with Tensor G4; manufacturer result, not a Museum measurement. |
| AI & security | |
| TPU performance | Manufacturer / primary sourceGoogle Scope: silicon-die Google’s comparison with Tensor G4; manufacturer result, not a Museum measurement. |
| AI | Manufacturer / primary sourceGoogle Scope: silicon-die |
| Security | Manufacturer / primary sourceGoogle Scope: silicon-die |
| Camera, video & display | |
| ISP | Manufacturer / primary sourceGoogle Scope: silicon-die |
| Connectivity & devices | |
| Target device | Manufacturer / primary sourceGoogle Scope: silicon-die |
Start with documented specifications, then use the study prompts below. The prompts are engineering guidance, not a claim about this product’s private RTL or physical implementation.
Build a traffic matrix across the documented CPU, graphics, AI and I/O engines. Model concurrent bursts, locality and worst-case response time at shared memory before choosing arbitration or buffering.
Choose a representative producer/consumer path and specify acceptance, ordering, backpressure, outstanding capacity and reset recovery. Turn these contracts into assertions and test scenarios before optimizing the datapath.
Obtain macro abstracts, pin maps, timing libraries and activity-based power estimates. Explore macro proximity, routing channels and power-grid reachability together. An annotated die photograph can suggest study questions, but cannot supply a netlist, PDK or signoff constraints.
Google’s TSMC 3 nm Pixel 10 SoC, with a faster TPU/CPU and new ISP.
This record has identification and source evidence. A deeper architectural article has not yet been written. Visible patterns alone do not establish block functions, capacity or an internal protocol.
Chipwise / NanoPhysics
Google Tensor G5. Frontside die view, as identified by source
View original die photoAccess to a source does not grant permission to reproduce its images.
Chipwise / NanoPhysics
Google Tensor G5. Backside/deprocessed die view, as identified by source
View original die photoAccess to a source does not grant permission to reproduce its images.
5 reasons why Google Tensor G5 is a game-changer for Pixel ↗ · Google
Google Tensor G5 — Chipwise image source ↗ · Chipwise / NanoPhysics