HPE today announced that Subaru selected HPE Cray XD670 servers to orchestrate its image recognition technology and AI. This rollout supports Subaru’s vision of increasing customer value and improving preventive safety—one of the key pillars of the car manufacturer’s comprehensive philosophy. HPE Cray XD670 servers featuring industry-leading and energy efficient direct liquid cooling is expected to accelerate the optimization of AI algorithms, model training, and inferencing for the next-generation EyeSight driver assist system currently under development.
Subaru is recognized by customers and industry organizations1 globally for its outstanding safety performance. Powered by AI, the next-generation EyeSight is being developed as part of Subaru’s commitment to achieving zero fatal traffic accidents by 20302.
The next-generation EyeSight system aims to enhance vehicle safety in diverse real-world scenarios by simultaneously measuring the precise distance to objects and classifying them using AI, using images captured by stereo cameras. Subaru has been providing its stereo camera–based image recognition technology for over 35 years and is now fusing it with the latest AI technologies to deliver greater value to customers to enhance preventive safety. The on-premises deployment of HPE Cray XD670 servers supports this effort by improving the computing power of Subaru’s AI development environment.
We are honored that Subaru has chosen HPE Cray XD670 to support Subaru’s AI initiatives and further bolster the safety capabilities of the EyeSight technology.
Senior Vice President and Managing Director, HPE Japan
“Subaru has an unwavering dedication to developing high-performing vehicles while pursuing the highest level of safety with a goal to eliminate fatal traffic accidents. The remarkably low rear-end collision rate of EyeSight-equipped vehicles stands as a powerful testament to that dedication,” said Hirokazu Mochizuki, senior vice president and managing director for HPE Japan. “The HPE Cray XD670 is a high-performing GPU-accelerated server and ranks at the top of computer vision and LLM workloads performance benchmarks. We are honored that Subaru has chosen HPE Cray XD670 to support Subaru’s AI initiatives and further bolster the safety capabilities of the EyeSight technology.”
Subaru’s vast repository of research data and accumulated technical expertise is a major advantage in the development of advanced driver-assistance systems (ADAS), and SUBARU Lab was established in December 2020 as a strategic hub to strengthen ADAS related in-house software development and AI-based image recognition. Researchers and engineers at SUBARU Lab are rigorously pursuing solutions to real-world driving risks, collecting video data by driving on roads around the world. This image data is used to optimize AI algorithms and train inference models that run on HPE Cray XD670 servers powered by NVIDIA H200 GPUs.
With the energy-efficient HPE Cray XD670 that delivers twice the performance compared to the existing system, we now have an on-premises environment that allows us to maximize our use of GPU resources, which we expect to speed the development of our AI models.
Deputy Chief of SUBARU Lab, and General Manager of ADAS Development Department, Subaru Corporation
“At Subaru, we pursue comprehensive vehicle safety through five key areas of our All-Around Safety philosophy: Primary Safety, Active Safety, Preventive Safety, Passive Safety, and Connected Safety. Subaru’s unique value lies in delivering peace of mind to our customers—a commitment that goes beyond just safety, and one that drives us to advance the development of cutting-edge technologies with a strong sense of urgency,” said Takashi Kanai, Deputy Chief of SUBARU Lab, and General Manager of ADAS Development Department. “With the energy-efficient HPE Cray XD670 that delivers twice the performance compared to the existing system, we now have an on-premises environment that allows us to maximize our use of GPU resources, which we expect to speed the development of our AI models.”
The HPE Cray XD670 system features direct liquid cooling for optimal energy efficiency and performance and is powered by eight NVIDIA H200 GPUs to deliver exceptional capabilities for AI model training and inferencing. HPE Cray XD670 achieved top rankings in the MLPerf Inference v5.0 benchmarks, an industry effort driven by MLCommons to help the industry standardize performance evaluation of AI technologies.
Subaru also selected HPE Tech Care Service for faster access to HPE experts, and intuitive self-help tools via the HPE Support Center. This support experience enables reduced downtime, streamlined operations, and unlocks greater value from the investment in HPE.
SOURCE: HPE
