Oliver wrote:The article talks about a few stability programs.
I have a question about what exactly is the technical difference of what a program such a prime95 vs. memtest86 are doing? How is it that a system could pass memtest86 yet fail prime95? What is the technical fundamental difference of how they are stressing the memory system?
memtest86 exercises the memory controller, CPU cache and DRAM by writing patterns to the memory and reading them back to see if there are inconsistencies. It runs its own operating system (Linux based?) and uses very little I/O. Any problem uncovered by memtest86 is likely to be due to memory / CPU / memory controller, and not due to drivers.
prime95 runs under Windows. The GUI is active, and you have tons of other drivers active while it is running. It is not an exhaustive test of the memory, but it performs lots of calculations and logic operations. So not only is a different area of the CPU being tested, you have interaction with tons of things that Windows does in the background.
I'd first test with memtest86, and only if that ran without errors, go on to prime95, then on to tests for the graphics subsystem (e.g. 3dMark05).