Actually the idea of bottlenecks is a bit too simplistic to describe the complexity of what is going on. On a single frame some parts of it are CPU dominated and some are GPU dominated. Given a doubling in performance of either of those components there will still be parts of the frame that improve. Its too simple to say that the CPU was "the" bottleneck in that benchmark before or that it was the GPU.
So while the benchmark improved, it didn't improve in line with the amount of performance available with the faster CPU, far from it. So what has actually happened is the remaining CPU bottlenecks within a frame where improved thereby improving performance slightly, but there could still be CPU bottlenecks in the benchmark that are yet to be removed and of course the GPU is providing limitations as well.
The CPU and the GPU are always a bottleneck in every game at moments. Its really a question of which upgrade provides the best scaling based on the relative size of the bottlenecks available. In most games and graphics benchmarks adding more GPU performance increases performance at a much better rate than CPU does. But adding CPU is always adding something, its just regularly not even 0.1 fps for a doubling of CPU performance.
I like a simple definition of "bottleneck". If you get a much faster GPU but the FPS remains roughly the same, then that's a CPU bottleneck. If you get a much faster CPU but the FPS remains roughly the same, then that's a GPU bottleneck. The converse is true also, if you swap out the CPU with a much slower CPU but the FPS remains the same, then something else besides the CPU is holding back the FPS.
It's not helpful for this definition of bottleneck to break down what happens in a given frame at a given moment, because it's the actual noticeable hindrance caused by the bottleneck that this definition focuses on.
If only one thing is holding back my FPS, and the FPS would otherwise be much faster, then I can replace only that one thing and see a benefit.
Facebook
Twitter