I generally use benchmarks to help me to tweak my systems to achieve their best performance. Since I'd rather produce content more than to just merely enjoy that created by others, even my gaming experiences have other motives at play. I try to learn what the gaming author did best and what I should try to avoid doing in game production using Unity 3d. I favor the Cinebench series of benchmarks (10, 11.5 and 15) because they have real world application to me. My favorite 3d software is Cinema 4d. It has been true that if I tweaked my system to get the best Cinebench score, I've also gone a long way in tweaking it to give me the best performance in Cinema 4d. In the past, that tweaking has been, almost 100%, at the CPU level. Over time, I've observed that as far as game play is concerned, that OpenGL performance in Cinebench will be the greatest by using the top of the line ATI/AMD cards. However, with the advent of GPU rendering and the paucity of current OpenCL rendering engines that has meant that I have had to place many of my eggs in the GTX CUDA's nest as far as content creation is concerned. Since this thread is under the hierarchy of video card overclocking and with the ascendency of the GPU over the CPU for parallel compute tasks, I hope that I'm not straying to far out of line to point out to those who might actually use Cinema 4d that using GPU rendering engines such as Octanerender (GPU only - mostly unbiased renderer), Thea render (CPU and now GPU - biased renderer) and RedShift (GPU and CPU - biased renderer that, of these three, seeks to overcome the limitation of vram size on the video card) can all improve actual Cinema 4d rendering performance more than by reliance placed solely on CPU rendering and tweaking. Tweaking a GPU to best achieve what the Cinema 4d application is designed to ultimately produce, i.e., an animation, from my experience is enhanced more by finding the sweetspot for memory overclocking. This is particularly so if you have SuperClocked (SC) GPUs. A little GPU core overclocking might also help, but its been my experience that performance gains peter out more quickly with core speed increases, and even more quickly if you have SC GPUs.
If you're into 3d rendering but haven't tweaked either your CPU(s) or GPU(s), why not? If you're into 3d rendering and have tweaked your CPU(s) and/or GPU(s), why did you tweak them, what did you use to tweaked them, how did you tweak them, and by what degree did you tweak them? What was the outcome(s)?