An overclocked computer that doesn't fail under
normal computing sessions shouldn't be considered "stable" when
intensive and sustained computing activities could bring it down.
Instead of waiting for the computer to crash in the middle of a game or
while encoding video, burn-in software can be used to test for stability
relatively quickly because of the extreme nature of its computing task.
Most applications--even synthetic benchmarks-- don't
necessarily tax the computer; they merely keep it busy with stalled
pipelines, missed branch predicts, hard drive access and slow memory
requests. The CPU is able to take advantage of these down-times to shut
off sections of the core to save power and reduce heat.
The never-ending "burn cycle" in Core Damage is
designed to prevent the CPU from entering its reduced power state as
much as possible by issuing a sequence of SSE, integer, floating point,
address generating and branch-predicting instructions optimized
specifically for the underlying micro-architecture to achieve a high
degree of core utilization. Stimulating the core in this way produces
more heat as execution units "wake up" to handle the computing task.
An overclocker relies on burn-in software to simulate
the worst-case scenario as to ensure the computer will remain stable
during ordinary and even intensive computing tasks.