CybersecurityChips & Compute6 min reading time

GPUThor breaks the error correction Nvidia recommended against Rowhammer

The Next Web
Read full post
Researchers at the University of Toronto developed GPUThor, a Rowhammer attack that bypasses Nvidia's error correction on Ampere GPUs, enabling privilege escalation. The attack affects RTX A4000, A4500, A5000, and A6000 cards and cannot be patched without new hardware.

More in Cybersecurity

Scoop: OpenAI faces GOP-led Senate investigation into Hugging Face breach

Covered by 2 sources
Cybersecurity6 min read

Anthropic Discloses Fourth AI Hacking Incident Involving Claude Opus 4.6

Covered by 2 sources
Cybersecurity5 min read

Every AI Incident Has Two Timelines. We Default To One

Forbes