GPU Overclocking Is Half Of The Story
When it comes to setting the clock rates on non-reference graphics cards, third-party vendors take different routes. Some overclocked cards employ higher GPU clocks, but they don’t always alter memory data rates. Typically, the GPU speed can be modified afterward through software tools (incidentally, most reference cards can also be tweaked fairly easily). These tools typically allow users to also adjust memory clock speeds, which is a critical warranty item. Experience shows that memory is more sensitive to overclocking. In addition, overclocking often alters 2D clocks as well, causing your card to use more power and requiring more cooling when it sitting there idle.
If you manually overclock a graphics card, you cannot know without some testing if the target settings remain effective, or if throttling in 2D mode is still possible. We've seen plenty of instances where an overclocked card ends up running at its maximum clocks, even on the Windows desktop. That defeats the purpose of throttling entirely, generating a lot of unnecessary heat your cooling subsystem is forced to deal with.
Factory-overclocked cards, however, correctly implement modified speed settings, meaning that they run the tweaked clock speeds in 3D mode and still throttle clock speeds and power consumption in 2D mode, making them more efficient, at least, without any aftermarket tweaking needed.
In this article, we’ll compare a reference ATI Radeon HD 5870 card and the factory-overclocked MSI R5870 Lightning TwinFrozr II, analyzing real-world differences between the two. We'll also run the reference card at overclocked settings as well, because we wanted to see if the pricier aftermarket solution can deliver more value than a reference board tuned to run faster.