Super Star Future-Intelligent Driving Computing Chip "Jingzhe R1"

Intelligent driving computing chip "Jingzhe R1"

Product Description:

"Jingzhe R1" is an intelligent driving computing chip launched by Super Star Future based on its self-developed NPU "Pinghu". It adopts TSMC's 12nm advanced process to ensure product safety and reliability. "Jingzhe R1" has powerful processing capabilities, which can meet the low-power real-time computing needs of multiple scenarios, provide Al computing power up to 20TOPS@INT8 and general computing power of 32KDMIPS, and support SIMD expansion acceleration. The typical power consumption of the entire chip is less than 10W, and it can support passive heat dissipation. "Jingzhe R1" has comprehensive sensor access capabilities and can support multi-chip and multi-die high-bandwidth expansion. "Jingzhe R1" is positioned in L2+ level intelligent driving applications, which can reduce the overall cost by 30%-50% while realizing the same level of intelligent driving solutions. It is an accurate choice for lightweight intelligent driving products.

Unique advantages:

1. Automotive-grade product definition and implementation to ensure product safety and reliability:

- Using TSMC 12nm advanced process technology;

- Meets AEC-Q100 Grade-2 requirements.

2. Powerful processing capability configuration to meet the low-power real-time computing needs of multiple scenarios:

- 8-core RISC CPU, general computing power up to 32 KDMIPS, support SIMD expansion acceleration;

- 4-core 20 TOPS INT8 computing power neural network processing NPU;

- The typical power consumption of the whole chip is less than 10W, which can support passive cooling design.

3. Self-developed high-efficiency neural network processing core NPU, which combines extreme energy efficiency and flexibility:

- Industry-leading energy efficiency ratio of 5 TOPS/W under the same process conditions;

- Supports common CNN/Transformer model calculations and custom operator extension development;

- Convolution operation efficiency > 85%, average chip computing utilization > 70%, fully tapping the chip computing power.

4. Comprehensive sensor access capabilities, supporting multi-chip/multi-die high-bandwidth expansion:

- Integrated 3*4 Lane MIPI, 2*USB3.0/2.0, 2*RGMII 1000Mbps communication interface;

- Integrated PCI-E Gen3 x4, 2*Chiplet D2D expansion interface.

Application scenarios:

"Jingzhe R1" can be applied to intelligent driving pre-installed mass production scenarios and edge intelligent scenarios such as vehicle-road collaboration, including active safety, L2+ integrated driving and parking, regional intelligent driving, human-machine co-driving, vehicle-road collaboration, etc.

Based on "Jingzhe R1", a variety of intelligent driving reference designs can be created according to different application scenario requirements and sensor configurations, including the high-end integrated driving and parking reference solution NOVA-ADCU Ultra, the lightweight integrated driving and parking reference solution NOVA-ADCU Pro, the binocular intelligent driving reference solution NOVA-ADCU Air (Bino), and the roadside perception reference solution NOVA-ADCU RS.

Future prospects:

Research data shows that from January to September 2022, the L2 and L2+ ADAS assembly rate of passenger cars in my country has reached 33.5%. At the same time, Shanghai has issued a policy: by 2025, the proportion of new cars with L2 and L3 functions must exceed 70%. After years of exploration, the focus of the intelligent driving market has begun to shift from open-scenario L4 to pre-installed mass-produced L2+ systems. The biggest market opportunity in the next few years will still come from mass-produced solutions represented by integrated driving and parking, and integrated cabin and parking. At the same time, the value share of automotive chips in the total cost of vehicles will continue to increase to more than 50%. It is estimated that by 2025, the scale of automotive electronics will exceed 800 billion yuan, and intelligent driving chips will exceed 100 billion yuan.

"Jingzhe R1" can be applied to intelligent driving application scenarios with the largest cargo volume and the greatest commercial potential, including 8M pixel front-view integrated machine, lightweight parking integrated machine, V2X roadside perception, etc., to play a significant economic benefit and industrial driving role. It is expected that in 2023, the company is expected to win the chip orders for mass-produced models from at least 2 automakers (including a leading benchmark automaker), achieve 30 million in revenue, and become a leading domestic intelligent driving chip manufacturer. It is expected to receive nearly 100 million yuan in orders in 2024.