Huzzah!
“Minimal Latency” Mode working as designed after the embarrassing problems you guys helped me find
For context, without framerate limiting, input latency averages around 110 ms. This is flipping insane (pun intended).
Nonetheless, that latency debug menu is still coming to Special K’s Framerate Limiter widget.
I’ll be getting some new toys to aid in the development of it. There will be a lot fewer screenshots of CapFrameX when I finish that work, lol.
Here are tests done @ 100 FPS, and I fell out of my chair and broke it
CapFrameX Report:
Special(K) Report.7z (11.9 KB)
Not even worth distributing screenshots for that, the results need to be seen in person to be believed. And I need validation of these results, because if that’s repeatable in other games then Reflex is irrelevant and Special K should be a vendor agnostic path to frame pacing bliss.
Final(?) Update:
NVIDIA’s driver-based limiter has been dethroned (caveats apply)
Since conventional wisdom is that the driver-based framerate limiting has the lowest input latency, let’s challenge that wisdom.
Special K’s almost matching NVIDIA? … about that.
The NVIDIA limiter’s capable of slightly lower latency than SK, but it only achieves the results in the first screenshot because Special K has done low-level tuning of the game’s swapchain for lower latency. Take SK out of the equation and NVIDIA’s actually -not- looking very competitive.
To round things out, here’s RTSS with the same exact low-level SwapChain tune that dramatically assisted the NVIDIA driver.
Special K + NVIDIA actually produces something impressive, showing that combining framerate limiters isn’t necessarily a bad thing.
And to really screw with everyone’s mind. This is all borderless window, because why not? Aside from RTSS most software responds favorably when given an optimized windowed swapchain and can be made to perform indistinguishibly from fullscreen exclusive.