Mobileye, STMicroelectronics set date for 5th-gen autonomous car sensor fusion chip
The two companies are co-developing the next (5th) generation of Mobileye’s SoC, with a view to equipping Fully Autonomous Driving (FAD) vehicles starting in 2020. To meet power consumption and performance targets, the EyeQ5 will be designed in advanced 10 nm or below FinFET technology node and will feature eight multithreaded CPU cores coupled with eighteen cores of Mobileye’s next-generation vision processors.
Taken together, these enhancements will increase performance 8x times over the current 4th generation EyeQ4. The EyeQ5 will process more than 12 Tera operations per second, while keeping power consumption below 5W, to maintain passive cooling. Engineering samples of EyeQ5 are expected to be available by first half of 2018.
Building on its experience in automotive-grade designs, ST will support state-of-the-art physical implementation, specific memory and high-speed interfaces, and system-in-package design to ensure the EyeQ5 meets the full qualification process aligned with the highest automotive standards. ST will also contribute to the overall safety- and security-related architecture of the product.
“EyeQ5 is designed to serve as the central processor for future fully-autonomous driving for both the sheer computing density, which can handle around 20 high-resolution sensors and for increased functional safety,” said Prof. Amnon Shashua, cofounder, CTO and Chairman of Mobileye. “The EyeQ5 continues the legacy Mobileye began in 2004 with EyeQ1, in which we leveraged our deep understanding of computer vision processing to develop highly optimized architectures to support extremely intensive computations at power levels below 5W to allow passive cooling in an automotive environment.”
EyeQ5’s proprietary accelerator cores are optimized for a wide variety of computer-vision, signal-processing, and machine-learning tasks, including deep neural networks. EyeQ5 features heterogeneous, fully programmable accelerators, with each of the four accelerator types in the chip optimized for its own family of algorithms. This diversity of accelerator architectures enables applications to save both computational time and energy by using the most suitable core for every task.
This optimized assignment ensures the EyeQ5 provides “super-computer” capabilities within a low-power envelope to enable price-efficient passive cooling. Mobileye’s investment in several programmable domain-specific accelerator families is enabled by its focus on the ADAS and autonomous-driving markets.
Mobileye has built the EyeQ5’s security defences based on the integrated Hardware Security Module. This enables system integrators to support over-the-air software updates, secure in-vehicle communication, etc. The root of trust is created based on a secure boot from an encrypted storage device.
EyeQ5 will be delivered to carmakers and Tier1s along with a full suite of hardware accelerated algorithms and applications that are required for autonomous driving. Along with this, Mobileye will support an automotive-grade standard operating system and provide a complete software development kit (SDK) to allow its customers to differentiate their solutions by deploying their algorithms on EyeQ5.
The SDK may also be used for prototyping and deployment of Neural Networks, and for access to Mobileye pre-trained network layers. Uses of EyeQ5 as an Open Software Platform are facilitated by such architectural elements as hardware virtualization and full cache coherency between CPUs and accelerators.
Autonomous driving requires fusion processing of dozens of sensors, including high-resolution cameras, radars, and LiDARs. The sensor-fusion process has to simultaneously grab and process all the sensors’ data. For this purpose, the EyeQ5’s dedicated IOs support at least 40Gbps data bandwidth.
EyeQ5 implements two PCIe Gen4 ports for inter-processor communication, which could enable system expansion with multiple EyeQ5 devices or for connectivity with an application processor. High computational and data bandwidth requirements are supported with four 32-bit LPDDR4 channels, operating at 4267 MT/s (Mega transfers/sec).
Engineering samples of EyeQ5 are expected to be available by first half of 2018. First development hardware with the full suite of applications and SDK are expected by the second half of 2018.
Mobileye: www.mobileye.com/
STMicroelectronics: www.st.com
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
