Originally Posted by LifeDisturbens
I have no idea what it is, i only got told from this forum that my rig is bottlenecking.
I don't know why, and i got the instructions to overclock the CPU until it's gone.
Well that is a problem. To understand if you have a bottleneck you really need to know what a bottleneck is. In this case there is a transfer of data between the CPU and GPU. The instructions that are sent to the GPU about the scene from the CPU are processed then displayed. The GPU has a rate at which it can form a scene and the CPU has a rate at which it can send the instructions. If the rate of instructions from the CPU is higher than what the GPU can process then you are GPU limited. In this case you should see a GPU running with 90% to 100% almost constantly (it is also dependent on the coding and engine for scene rendering so fluctuations will be seen in usage no matter the the components.) On the opposite hand if the CPU is sending instructions at its highest possible rate and the GPU is not at its highest processing rate, then you would have a CPU bottleneck. Usually what you would see is a GPU with usage percentages lower than 75%. It doesn't require that the CPU is being 90% to 100% though.
The dirty secret is that you will never likely see a true CPU bottleneck. Why? Because coding for games isn't written in a way that would push the CPU to its max rate constantly. Hence why it is entirely possible to see a CPU with 70% usage with a GPU at 70% usage and you have max your possible frame rate. (Call it a coding bottleneck if you want) It is relatively easy to see a GPU bottleneck but much more difficult to see a CPU bottleneck. You overclock your CPU to increase its ability to transfer instructions to the GPU. The problem becomes even fuzzier in reference to multi-GPU setups because the data has to be split and sent to multiple GPUs as blocks to be processed by a single GPU and then combined once again. Same goes for multi-core processors.
Now that is for games and like applications. A better test of a bottleneck is performing calculations on a CPU and GPU that are Dependant upon one another. Take for instance what I do. I work with incredibly large datasets of LiDAR and aerial imagery. These datasets are comprised of billions of points and cells that need to have a mathematical calculation performed on them. GPU acceleration can speed this up but it first needs to be processed by the CPU into usable chunks of data for the GPU. The coding for this process is highly efficient (I would know because I wrote the software). If I saw the CPU running at 100% and the GPU running at 70% then my CPU is most likely holding the processing back. That doesn't happen though as I see the CPU hanging around 70% and my GPUs at 90% to 100%.
People usually forget the bit about the coding and chalk it up to hardware being the problem causing a bottleneck. They also forget that instructions are handled differently between platforms which result in even more difficulty in pointing to a direct cause of a bottleneck if there is even one. This would be evident by the fact that you can be getting 50 FPS with 60% usage on stock clocks and get 80 FPS with the same 60% usage with an overclock.
Sorry for the long complicated post. I appear to be a bit long winded tonight but hopefully this helps you understand what is going on.