Interference Risks In Processing-Using-DRAM (U. of Tokyo, ETH Zurich, CISPA, Riken)
Researchers from The University of Tokyo, ETH Zurich, CISPA, and RIKEN published a technical paper titled “PuDGhost: Experimental Analysis of Computation Result Corruption in Processing-using-DRAM Operations on Real DRAM Chips and Implications for Future Systems.” Abstract excerpt: “We reveal PuDGhost, an interference phenomenon where a PuD operation in a given column produces erroneous results due... » read more The post Interference Risks In Processing-Using-DRAM (U. of Tokyo, ETH Zurich, CISPA, Riken) appeared first on Semiconductor Engineering .
As Processing-Using-DRAM (PuD) becomes a critical pathway for energy-efficient computing, the identification of fundamental interference risks is timely for guiding its development and adoption.
A sophisticated reader should care because these findings highlight potential vulnerabilities in advanced memory architectures that could impact the reliability and security of future high-performance computing systems and AI accelerators.
The understanding of PuD's reliability now includes specific interference phenomena, necessitating new design considerations, verification methodologies, and potential architectural adjustments to prevent computational errors.
- · Memory verification tool developers
- · Secure hardware architects
- · Academic research institutions
- · Semiconductor companies focused on robust systems
- · Developers of PuD without robust error mitigation
- · Companies reliant on insecure or unreliable in-memory computing
- · Users of systems employing unverified PuD implementations
The immediate consequence is an increased focus on error correction and fault tolerance mechanisms within PuD and in-memory computing designs.
This could lead to a delay in the widespread commercial deployment of certain PuD technologies until these interference issues are thoroughly addressed.
Long-term, successfully mitigating these risks could establish more secure and reliable foundations for future AI and high-performance computing architectures, potentially accelerating their adoption in sensitive applications.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at Semiconductor Engineering