Uploaded March 2025 | Updated September 2026, 1 week ago
AI computing devours energy, not just because of the volume of data involved but also due to chip design. Most chips are built on the von Neumann architecture, which separates compute and memory in an arrangement that’s perfect for conventional computing.But when these traditional processors are asked to compute for AI, they run into what’s called the von Neumann bottleneck – or the lag that happens when data moves slower than computation. At @IBM Research, engineers are exploring ways to complement and overcome the bottleneck with new processors – like AIU Family – designed to break this bottleneck and supercharge AI efficiency. Researchers Manuel Le Gallo-Bourdeau, Geoffrey Burr and Hsinyu (Sidney) Tsai help us navigate the concept: research.ibm.com/blog/why-von-neumann-architecture-is-impeding-the-power-of-ai-computing----
#ibmmainframe #Research #AI #Compute #Processors
AI computing devours energy, not just because of the volume of data involved but also due to chip design. Most chips are built on the von Neumann architecture, which separates compute and memory in an arrangement that’s perfect for conventional computing.But when these traditional processors are asked to compute for AI, they run into what’s called the von Neumann bottleneck – or the lag that happens when data moves slower than computation. At @IBM Research, engineers are exploring ways to complement and overcome the bottleneck with new processors – like AIU Family – designed to break this bottleneck and supercharge AI efficiency. Researchers Manuel Le Gallo-Bourdeau, Geoffrey Burr and Hsinyu (Sidney) Tsai help us navigate the concept: research.ibm.com/blog/why-von-neumann-architecture-is-impeding-the-power-of-ai-computing----
#ibmmainframe #Research #AI #Compute #Processors
