By the time this message gets printed it *is* too late for the process that was impacted. It crashed. If a bad CPU was at fault for the crash, the "badness" could have been hundreds or thousands of instructions before the crash. But, not all "badness" is created equal. For the CPU to pass validation, get into production, boot and be running good code in the first place, it can't be *totally* broken. That's why the kinds of issues that this helps with are usually fundamentally subtle. It might take a *very* specific sequence of events to trigger an issue. It might only happen at *one* tiny place in *one* program. Most of the time, the CPU is still well enough to get into the kernel, spit this message out, and keep on going. BTW, I work on Linux at Intel.