RealNC wrote: ā29 Apr 2023, 10:15
Frame time is not affected by LVT. Display scanout time is affected. You can't measure that with software. You need like a 1000FPS camera and record the screen.
Translation / Disambiguation of RealNC
He's correct, but needs disambiguation;
Everyone, RTSS doesn't always measure frametime as rendertime per se (pure GPU render time) because RTSS actually measures (GPU render time + frame-cap-blocking time) when it measures what it calls "frametime".
This is because RTSS is not always capable of determining whether the GPU is rendering or whether the drivers are now waiting (via a sync technology) before releasing the frame. So what RTSS calls "frametime" is actually the sum of (GPU rendering time) + (amount of idling time until frame release).
In other words, a sync technology such as VSYNC ON will wait until the synchronization interval (between refresh cycles), before releasing the frame (terminologies may be "frame swap" or "swapchain" or "blocking Present() call" or "return from a blocking frame-presentation API call")
Which becomes flat with VSYNC ON or RTSS Scanline Sync.
This is where people are getting "confused" by when people say "frametime" because it can means two very different metrics, due to this confusion;
(A) Pure frame rendering time; or
(B) Time interval between frame releases (including all driver/display-based throttles, caps, delays, sync technologies, etc).
If you use uncapped VSYNC OFF, then RTSS frametime is both (A) and (B) simultaneously. But if you use VSYNC ON or other caps (especially driver-based caps), then frametime is only (B) instead.
In other words, confusion is because (A) and (B) can be same or divergent. In other words, frametime is not always rendertime
Kuscheln_Hammer92 wrote: ā29 Apr 2023, 09:51
I have been pouring so much time and many hours into this. I hope it isn't because I'm using DP rather than hdmi. I own an xg2431 paired with an rx 6600. I've created crt motion clarity with the instructions for a lvt with purexp+ adjustments. However, everywhere I read I should expect to see frametime regularly under 16.7ms. I am trying to pair this with RTSS Scanline Sync which I think I finally have nailed. It required me turning on the framelimiter to match hz in the setup. I could not get the tear line to stop moving across the screen without that selected. Is that the minimum frametime I should expect to see despite the LVT? I apologize this is at 60hz custom resolution with a 4500 LVT. I tried basing off of 239.9.. native and then make lvt 2600ish via the vertical calculator but that led to a black screen and obscenely large back porch. Lowering to 4500 allowed the screen to display for me. I would appreciate any and all assistance. Trust me, I have spent at least 18 hours on the topic of both backlight strobing and low latency syncs.
For LVT to reduce latency;
(1) You have to compare lag with the same non-VSYNC-OFF technology (e.g. RTSS Scanline Sync before/after LVT); and
(2) You have to adjust the scanline sync position as a negative index (above top edge).
RTSS Scanline Sync (+VSYNC OFF) isn't less lag than pure VSYNC OFF, but it's one of the lowest-lag non-VRR "framerate=Hz" sync technology available.
RTSS Scanline Sync answers the need of "I love VSYNC ON. But I hate VSYNC ON lag. What's my alternative?". And here, that's where LVT can reduce latency
even further. It's not for people who say "I want something lower lag than VSYNC OFF".
When following RTSS Scanline Sync instructions, reposition the tearline above the top edge of the screen, for minimum strobe latency;
![Image](https://www.blurbusters.com/wp-content/uploads/2019/07/scanout-quick-frame-transport-518x690.png)