Combines vision data with radars and laser scanners for robust object detection. 4. Applications of the EyeQ4
The EyeQ4 is engineered to provide "super-computer" capabilities within a strictly constrained automotive power envelope.
: Capability for vehicle detection from any angle, next-generation lane detection, and traffic light detection. Mapping & Localization : Supports Road Experience Management (REM™) for real-time crowd-sourced mapping. Safety Applications
Capable of processing up to 8 cameras simultaneously at 36 frames per second (fps). Architectural Overview eyeq4 datasheet
The EyeQ4 uses a , distributing tasks across different specialized cores to maximize efficiency. 1. CPU Cores Quantity: 4 quad-threaded MIPS InterAptiv cores.
The chip powers features like Traffic Jam Assist, Lane Keep Assist, Automatic Emergency Braking (AEB), and even some highway pilot systems found in vehicles from BMW, Nissan, Volkswagen, and Geely.
datasheet lies in its heterogeneous architecture, which uses different types of proprietary accelerators for specific vision tasks: Combines vision data with radars and laser scanners
While the EyeQ5 targets fully driverless robo-taxis and L4 systems, the for high-volume consumer vehicles due to its highly optimized balance of cost, low power draw, and rich L2+ feature set. 8. Conclusion
+-------------------------------------------------------------+ | Mobileye EyeQ4 Software Stack | +-------------------------------------------------------------+ | [Application Layer] REM (Mapping) / RSS (Safety Policy) | +-------------------------------------------------------------+ | [Perception Layer] Vehicle, Pedestrian, Lane, Sign Detect | +-------------------------------------------------------------+ | [Framework Layer] Deep Learning / Sparse Models Core | +-------------------------------------------------------------+ | [OS / Hardware] RTOS / ASIL Functional Safety Firmware | +-------------------------------------------------------------+ Road Experience Management (REM)
Accommodates variable road profile reconstructions, path planning over construction zones, and localized lane boundaries. : Capability for vehicle detection from any angle,
. This specialized manufacturing process is what allows it to deliver "super-computer" performance within a tiny power envelope. Computational Performance: 2.5 Teraflops (trillions of operations per second). Power Consumption: Approximately , which is less than many standard mobile phone processors. Architecture: A heterogeneous mix of cores designed for specific tasks: Four multi-threaded MIPS cores. VMP (Vector Microcode Processors):
Unlike general-purpose CPUs or graphics-heavy GPUs, the EyeQ4 utilizes a highly specialized, heterogeneous architecture. It balances programmable compute cores with dedicated hardware accelerators to maximize efficiency. Heterogeneous Compute Blocks
Approximately 3 Watts , achieved through a high-efficiency 28nm FD-SOI (Fully Depleted Silicon On Insulator) manufacturing process.