@HPCpodcast-110: John Gustafson on AI, HPC, Making Every Bit Count – In Depth
Can AI and high-performance computing get better answers with fewer bits? HPC pioneer John Gustafson, author of Every Bit Counts, joins Doug Black and Shahin Khan to challenge the default reliance on 64-bit floating point, and explain why numerical precision belongs at the center of the debate over computing performance, memory bandwidth and energy use.
Gustafson makes the case for posit arithmetic, a number format designed to get more useful information from every bit. The conversation explores how matching precision to the task could reduce data movement and energy consumption while preserving the accuracy applications need. His prescription: let the computer handle more of those decisions.
The discussion also takes on the Ozaki Scheme, the limits of LINPACK, the promise of dataflow architectures, and the enduring relevance of Gustafson’s Law. We also touch on his “SC Perennial” status—he has attended every SC conference since 1988—and his work at Vq Research, using 3D printing to rethink how capacitors, inductors and resistors are built and integrated.
A candid conversation with a computing pioneer who keeps challenging assumptions.
For more, revisit episode 37 of the @HPCpodcast, John’s previous appearance, which laid the foundation for today’s discussion.
The OrionX editorial team manages the content on this website.