Dream Chip ADAS SoC shows power promise
Dream Chip Technologies has announced record power efficiency of its ADAS System-on-Chip (SoC) for automotive computer vision applications, fabricated on GLOBALFOUNDRIES's 22FDX semiconductor process at the foundry's Fab 1 facility in Dresden, Germany.
The SoC was created in close cooperation with Arm, ArterisIP, Cadence, GLOBALFOUNDRIES, and INVECAS as part of the European Commission's ENIAC THINGS2DO reference development platform, where about 40 partners in Europe cooperated to propel the FDSOI-Design Ecosystem. The chip offers high performance image acquisition and processing capabilities and supports AI / Neural Network (NN) vision operation with a total of 1 TOPS at 500 MHz on 4 parallel engines. With all functions including quad-core Arm Cortex-A53, Tensilica DSPs, and LPDDR4-Interfaces activated, the SoC shows single digit power dissipation without the need for forced cooling, which is of significant importance for embedding in automotive environments.
The SoC incorporates Dream Chip Technologies' image signal processing pipeline, working in conjunction with Cadence Tensilica Vision P6 DSPs and a quad-core cluster of Arm Cortex-A53 processors. In addition, a lock-step pair of Arm Cortex-R5 processors provides functional safety and the SoC is interconnected with an ArterisIP FlexNoC network-on-chip. Memory bandwidth is provided by a dual-channel LPDDR4-interface from INVECAS. 2 DDR-memory-chips and the SoC are mounted together on a chip-carrier, so that the module is providing 4 GigaBytes in total system memory.
The System Module is the centre piece of a new ADAS platform from Dream Chip Technologies targeted at automotive applications with a need for cost, performance and low power for embedding into the car without the need for forced cooling, such as fan or liquid. It is targeted to take over the central image recognition and manipulation tasks, based on camera capture and due to its tiny power footprint is geared to be integrated with the camera module.
Dream Chip Technologies is part of GLOBALFOUNDRIES's FDXcelerator Partner Program and supports the European automotive industry with design and engineering services. Dr. Jens Benndorf, Managing Director and Co-Founder of Dream Chip Technologies said: "Being able to collaborate with Arm, ArterisIP, Cadence, GLOBALFOUNDRIES and INVECAS and getting very early access to the 22FDX technology and the IP cores was an exciting experience and a tremendous benefit for Dream Chip Technologies, as we were able to drive Computer Vision MPSoC design in Europe to the next level. We have received an excellent technical support from these companies while meeting very aggressive schedule constraints. The power consumption we measured on the silicon have fully met the expectations and the result is highly competitive for the ADAS Computer Vision MPSoC market."
Of particular importance is the new and reduced power footprint of this SoC in 22FDX-Technology from GLOBALFOUNDRIES. AI/NN-operation for image recognition is available today, but most of the solutions need active cooling. Implementation of Dream Chip Technologies' SoC on GLOBALFOUNDRIES's 22FDX platform demonstrated single digit (1.0) Watt and cooling targets for designers managing power dissipation. If needed, the SoC bears the potential to increase the performance even further up to 2 TOPS at 1.0 GHz by applying GLOBALFOUNDRIES's forward body-bias capabilities and other optimization techniques.
"Building the best power efficiency and machine learning performances with a fully integrated SoC chip will pave the way for self-driving cars and accelerate ADAS adoption," said Sanjay Charagulla, Senior Director - Vertical Market Segments, CMOS BU of GLOBALFOUNDRIES.
Professor Blume, Head of Architectures and Systems Group of the Leibniz University of Hannover, Germany, states: "This SoC ranges among the most power efficient chips for AI. We have run implementations; code optimizations and benchmarks and we are amazed of the effectiveness of the Tensilica Vision P6-engines in this SoC architecture for image recognition and other image transformations. One TOPS on a chip without fan is very impressive."
