Newshot

AMD Executive: AI Will Not Replace Supercomputers; High-Precision Computing Remains Irreplaceable

AI summary

AMD's corporate vice president of HPC and sovereign AI, Scott Tease, said most HPC research institutions have recognized that AI "will never replace" traditional high-performance computing, and that high-precision scientific computing relying on double-precision FP64 will persist even as AI applications continue to expand. Tease noted that AI excels at low-precision, highly fault-tolerant data for model training, inference and rapid screening, but frontier research such as climate simulation, nuclear energy, materials, fluid dynamics, chemistry and national security still requires FP64-level numerical precision to ensure reliable and reproducible complex simulations. Tease argued that future scientific computing should not be "AI replacing HPC" but a heterogeneous computing system linking CPUs, GPUs, AI accelerators, traditional supercomputers and quantum computers, and AMD is working with IBM on a pilot at Oak Ridge National Laboratory connecting AMD's "classic" computing stack and the Umbra HPC system with IBM's 129-qubit Quantum Nighthawk quantum computer to Oak Ridge Leadership Computing Facility, hoping to explore quantum error correction, AI-assisted scientific research and t

Translation(model translation, not the source's own text)

Article title: AMD Executive: AI Will Not Replace Supercomputers; High-Precision Computing Remains Irreplaceable

AMD Executive: AI Will Not Replace Supercomputers; High-Precision Computing Remains Irreplaceable

IT之家, November 17: Tom's Hardware, a technology media outlet, published a blog post yesterday (November 16), reporting that Scott Tease, AMD's Corporate Vice President responsible for High Performance Computing (HPC) and Sovereign AI, said most HPC research institutions have recognized that AI “will never replace” traditional high-performance computing. Even as AI applications continue to expand, high-precision scientific computing that relies on double-precision FP64 will continue to exist for a long time.

AMD Instinct MI455X Accelerator
AMD Instinct MI455X Accelerator

Tease pointed out that AI is good at handling low-precision, highly fault-tolerant data and is suitable for model training, inference, and rapid screening; however, frontier research such as climate simulation, nuclear energy, materials, fluid dynamics, chemistry, and national security still requires FP64-level numerical precision to ensure the reliability and repeatability of complex simulation results.

Citing the blog post, IT之家 reported that Tease argues future scientific research compute should not be about “AI replacing HPC,” but should build a heterogeneous computing system that connects CPUs, GPUs, AI accelerators, traditional supercomputers, and quantum computers in coordination, allowing the system to automatically determine which tasks are handed to AI, which are handed to GPU acceleration, and which must enter a quantum environment.

IBM Quantum Nighthawk r2 processor.
IBM Quantum Nighthawk r2 processor.

On the exploration front, AMD is advancing a pilot with IBM at Oak Ridge National Laboratory, connecting AMD's “classical” computing stack and the Umbra HPC system with IBM's 129-qubit Quantum Nighthawk quantum computer at the Oak Ridge Leadership Computing Facility, hoping to explore integration paths for quantum error correction, AI-assisted scientific research, and traditional simulation, and to assess the architecture, software, and workflows needed for future hybrid quantum systems.

The translation is machine-generated and individual terms may be off. Nothing is rewritten or cut; the original is authoritative.

Published by IT之家
This site only summarises and aggregates; copyright stays with the original author. Headlines and summaries follow the source — where they differ, the original wins.
Read the original ↗