I have a laptop with an i7-4700HQ plus GTX 850M combo, meaning Optimus with all its benefits and problems.
When I first installed osu, after a bit of fiddling with the settings I found that fullscreen mode running on the dedicated gpu results in what I perceived as no cursor latency. I'd love to use borderless like in virtually every other game I play but that causes a slight but noticeable delay.
Now when I tried out the cutting edge build (b20150825.10) I noticed that the familiar "slight but noticeable delay" is there in both fullscreen and borderless modes.
Long story short, I made a table showing perceived and reported latencies for every setting combination. When you look at the table you'll notice that I found the combination I'll be using when OpenGL hits stable but I thought you guys might use this to figure something out as I've already found a couple posts mentioning similar issues.
If nothing else, adding a note that integrated gpus have a potentially lower latency in such scenarios (as Tom94 seems to be aware) might be helpful.
The tableAll tests are done on a clean install of Windows 10, default settings in both Intel and Nvidia graphics drivers, the laptop was plugged in, no fps limit in osu.
Perceived latency is a number on an arbitrary scale where 0 is the least latency I ever experienced in osu. A number larger than 2 would be what feels like more than 100ms latency like I was experiencing in CuttingEdge-Integrated-Borderless main menu. I wish there was a proper way to measure this.
FPS was measured on the offset wizard map because it's the only map where it was relatively stable.
What puzzles me is the large difference in performance in fullscreen and borderless modes as well as the better performance of the integrated gpu in the cutting edge build.