| smbhax [sys=PC; cat=Utility; reg=EUR] |
| | | https://www.guru3d.com/files-details/rtss-rivatuner-statistics-server-download.html
Aka "RTSS." I found it recommended here https://wiki.gbl.gg/w/Help:Flycast_GGPO#Emulator_Video_settings for limiting the "Flycast" Dreamcast emulator to NTSC output speed, 59.94 fps, as an optimization.
It can load with windows and sits in the system tray. |
|
|
| | | RTSS was preventing some other games--Mesen and USFIV for instance--from capturing correctly in OBS until I set RTSS's "Global" profile to "Application detection level" = "None."
Forcing Flycast to 59.94 fps definitely helps KOF 2000's shadow flicker, for instance--the character shadows are much less glitchy when recorded at 59.94:
|
|
|
| | | | RTSS doesn't interfere with OBS recording when I use it to limit Flycast (Dreamcast games); it DOES interfere with my OBS recording of the Unity game Turbo OutRun Reimagined (see entry 1732), however--having to record that via Display Capture instead. |
|
|
| | |
Very roughly measuring input delay in Super Street Fighter II X for Matching Service (Dreamcast) in the emulator Flycast in Windows, under various settings, using 60 fps cell phone footage of me crudely playing the game using my Hori Real Arcade Pro 2017 edition PS3/PS4 arcade stick on my mid-range gaming laptop, a "Sager NP6858CQ 15.6-Inch Thin Bezel FHD 144Hz Gaming Laptop, Intel i7-10750H, GTX 1660Ti 6GB, 16GB RAM, 1TB NVMe SSD" bought in summer 2020, outputting to a 60 Hz 1080p Asus 5ms monitor.
I measure the delay by pausing the video, then playing back one frame at a time ("." period key for YouTube) and counting the frames between the light punch button being fully pressed, and Ryu starting to play his punching animation.
0:00 - Vulkan video mode
0:33 - OpenGL video mode
0:59 - DX9 video mode
1:21 - DX11 video mode
1:55 - (rebooted just in case after Nvidia driver update) - Vulkan video from here on
2:17 - Linear Interpolation off from here
2:36 - Internal Resolution lowered to Native (from x9) from here
3:11 - 59.94 framerate limiter (via RivaTuner Statistics Server) turned off
Vulkan did seem to be slightly more responsive than the other video modes.
Linear interpolation seemed to be adding close to a frame of delay. I think maybe I had this setting OFF when I was testing Flycast before, because I can't seem to get down to just 1 frame of delay--which I had measured before--with it on. So that's going to stay off; I want Nearest Neighbor scaling and I have that set separately anyway; this setting seemed to produce slightly smoother texture filtering on 3D surfaces.
Raising the Internal Resolution setting makes the pixelization more regular--this is particularly noticeable in the checkerboard pattern in the background of SSF2X health meters, for instance, although that isn't visible here; without raising Internal Resolution there can be some jaggy jankiness--and I had it at x9, max, which admittedly may have been a bit extreme...but Internal Resolution x9 did seem to be adding a fractional amount of delay, now that I look at it closely, so now I'm paranoid about the setting and for peace of mind am going to have keep it at Native (ie 640x480 in non-widescreen). So we'll have some funky jaggies and some a shimmer when scrolling, etc.
Using RivaTuner Statistics Server to limit flycast.exe to 59.94 fps seems to produce slightly more reliable input delay, ie it seems to stay close to 1 frame more and doesn't spike up to 2 frames as much. So I guess I'll leave that in place; it does prevent Flycast's "Fast Forward" function from working in Flycast's OpenGL and Vulkan graphics modes, which is too bad, but ah well, must have minimum delay, mustn't we? MUSTN'T WE????!? = oo |
|
|
| | | | Nvidia Control Panel can in theory set max fps per application--but doesn't allow fractional values. |
|
|
| | |
