I wonder if RAM actively fails more often than we realize because I would presume desktop users more frequently switch apps and processes spawn and die on a more regular basis than server processes do.
I wonder if this could explain some system instability I've seen.
The long and short of it is that my desktop is unstable if CPU power management features (sleep states) are enabled. In particular, I would experience seemingly random (Windows) kernel panics, generally non-reproducible but associated with either audio or graphics card use. I suspect some sort of memory corruption because of a handful of non-crashing errors that led to obviously corrupt data inside a running game.
However, memtest86 always gave the memory a clean bill of health, and indeed the system is quite stable after disabling the CPU sleep stages in BIOS.
DMA-driven corruption might explain both the problem and its isolation to audio/graphics I/O.
3 comments
[ 2.3 ms ] story [ 15.5 ms ] threadThe long and short of it is that my desktop is unstable if CPU power management features (sleep states) are enabled. In particular, I would experience seemingly random (Windows) kernel panics, generally non-reproducible but associated with either audio or graphics card use. I suspect some sort of memory corruption because of a handful of non-crashing errors that led to obviously corrupt data inside a running game.
However, memtest86 always gave the memory a clean bill of health, and indeed the system is quite stable after disabling the CPU sleep stages in BIOS.
DMA-driven corruption might explain both the problem and its isolation to audio/graphics I/O.