Cross-Industry Hardware Specification to Accelerate AI Software Development
September 16, 2022 | IntelEstimated reading time: 1 minute
Arm, Intel and Nvidia have jointly authored a paper describing an 8-bit floating point (FP8) specification and its two variants E5M2 and E4M3 to provide a common interchangeable format that works for both artificial intelligence (AI) training and inference. This cross-industry specification alignment will allow AI models to operate and perform consistently across hardware platforms, accelerating AI software development.
Computational requirements for AI have been growing at an exponential rate. New innovation is required across hardware and software to deliver computational throughput needed to advance AI.
One of the promising areas of research to address this growing compute gap is to reduce the numeric precision requirements for deep learning to improve memory and computational efficiencies. Reduced-precision methods exploit the inherent noise-resilient properties of deep neural networks to improve compute efficiency.
Intel plans to support this format specification across its AI product roadmap for CPUs, GPUs and other AI accelerators, including Habana® Gaudi® deep learning accelerators.
FP8 minimizes deviations from existing IEEE 754 floating point formats with a good balance between hardware and software to leverage existing implementations, accelerate adoption and improve developer productivity.
The guiding principle of this format proposal from Arm, Intel and Nvidia is to leverage conventions, concepts and algorithms built on IEEE standardization. This enables the greatest latitude for future AI innovation while still adhering to current industry conventions.
Suggested Items
Siemens Delivers New Solido IP Validation Suite
05/07/2024 | SiemensSiemens Digital Industries Software introduced Solido™ IP Validation Suite software, a comprehensive, automated signoff solution for quality assurance across all design intellectual property (IP) types, including standard cells, memories and IP blocks.
U.S. Air Force Secretary Kendall Flies in AI-Piloted X-62A VISTA
05/06/2024 | Lockheed MartinLockheed Martin Skunk Works joined the U.S. Air Force Test Pilot School and other government and industry partners in hosting U.S. Secretary of the Air Force Frank Kendall to fly in the X-62A Variable In-flight Simulation Test Aircraft (VISTA), a one-of-a-kind aircraft modified to test artificial intelligence (AI) and autonomy capabilities.
Real Time with… IPC APEX EXPO 2024: Software Solutions for Circuit Board Challenges
05/03/2024 | Real Time with...IPC APEX EXPONolan Johnson speaks with Will Webb from Aster Technologies about their software solutions for design teams, manufacturing, test engineers, and process engineers. Aster's software addresses the increasing complexities of circuit boards and the need for alternative testing methods.
Real Time with… IPC APEX EXPO 2024: My Role as a Technology Solutions Director
05/02/2024 | Real Time with...IPC APEX EXPOPeter Tranitz, senior director of technology solutions at IPC, shares insights into his role as the design initiative lead. He details his advocacy work, industry support, and the responsibilities of the design initiative committee. The conversation also covers the revamping of standards, the IPC Design Competition, and the implementation of design rules in software tools.
Boeing Validates Software for Future Manned Unmanned Refueling Missions
05/02/2024 | BoeingBoeing has advanced its manned-unmanned teaming (MUM-T) technology using a digital F/A-18 Super Hornet and MQ-25 Stingray. The testing shows the software is maturing for future U.S. Navy use and a potential to deploy the teaming capability on both F/A-18 Block II and III Super Hornets.