We're currently having a bit of a discussion about 60FPS @60Hz vs 60FPS @120Hz in another forum on a sample & hold display.
John from Digital Foundry says he can definitely see a negative effect when using 120Hz vs 60Hz. (i.e. he says 60Hz is actually better in this case)
However, I find it difficult to find an objective reason beyond subjective perception for this.
If we're e.g. looking at what Mark (@Chief Blur Buster) has posted here I would assume that the difference should be negligible.
In both cases a frame should be effectively visible for ~16.7ms with the difference that @120Hz an image is displayed twice with two ~8.3ms scanouts.
The other difference is of course that the transition from one frame to the next is faster with 120Hz compared to a 60Hz signal.
Does anyone have an explanation for the effect John describes?
Sample & Hold Motion Blur 60FPS @60Hz vs. @120Hz
Re: Sample & Hold Motion Blur 60FPS @60Hz vs. @120Hz
If you're talking about the image itself rather than motion, different refresh rates results in different overdrive settings being applied. The image will change slightly because of that. In the monitors I used, I can't tell the difference. But it's possible that on some monitors it could be visible if the overdrive implementation of the display is really bad.
The bigger difference would be in motion. Meaning frame pacing. If you don't sync 60FPS to exactly every other frame in a 120Hz, non-VRR mode, there's going to be some jitter. That one is very visible, at least by most people.
If you do use VRR, then it depends on how good the display is. However, it should be noted that VRR is driven by the game. Even with a very good display, the frame pacing is going to be ever so slightly uneven compared to a non-VRR, fixed refresh 60Hz mode, so there could be a perceptible difference. I can't see it, but other people might be able to.
You have to consider that on fixed-refresh, non-VRR 60Hz, vsync is going to force very exact frame pacing. Like, ludicrously exact. The exact vsync moment is based on the video signal generated by the video card, and that signal is extremely exact. There's no software involved anywhere. As a result, every frame is going to be scanned out by the display at a very even pace. With VRR, this isn't true anymore. The display will scan out the frames at a pace that depends on software (the game itself and the video driver.) The display scanout time graph is not going to be a perfectly flat line. The video driver usually does its best to make it as flat as possible, but it's not going to be perfect. On my display (native gsync, not just gsync "compatible"), it looks the same as fixed 60FPS@60Hz. There is a difference for sure, but it's not big enough for me to perceive. But this might not be true for other displays.
The bigger difference would be in motion. Meaning frame pacing. If you don't sync 60FPS to exactly every other frame in a 120Hz, non-VRR mode, there's going to be some jitter. That one is very visible, at least by most people.
If you do use VRR, then it depends on how good the display is. However, it should be noted that VRR is driven by the game. Even with a very good display, the frame pacing is going to be ever so slightly uneven compared to a non-VRR, fixed refresh 60Hz mode, so there could be a perceptible difference. I can't see it, but other people might be able to.
You have to consider that on fixed-refresh, non-VRR 60Hz, vsync is going to force very exact frame pacing. Like, ludicrously exact. The exact vsync moment is based on the video signal generated by the video card, and that signal is extremely exact. There's no software involved anywhere. As a result, every frame is going to be scanned out by the display at a very even pace. With VRR, this isn't true anymore. The display will scan out the frames at a pace that depends on software (the game itself and the video driver.) The display scanout time graph is not going to be a perfectly flat line. The video driver usually does its best to make it as flat as possible, but it's not going to be perfect. On my display (native gsync, not just gsync "compatible"), it looks the same as fixed 60FPS@60Hz. There is a difference for sure, but it's not big enough for me to perceive. But this might not be true for other displays.
Steam • GitHub • Stack Overflow
The views and opinions expressed in my posts are my own and do not necessarily reflect the official policy or position of Blur Busters.
The views and opinions expressed in my posts are my own and do not necessarily reflect the official policy or position of Blur Busters.
Re: Sample & Hold Motion Blur 60FPS @60Hz vs. @120Hz
Also, there probably was a context. We'd need to know the exact quote.
- he could be talking about sample-and-hold display modes. I actually saw him doing that on one of the recent Digital Foundry videos. But he said the opposite - that 120Hz is better, obviously, due to less time spent in "paused" mode.
- he could be talking about specific TV or monitor. For example, the newest OLED TV line from LG for some stupid reason, does not support strobed (Black Frame Insertion) mode at 120Hz, but it does at 60Hz. Some monitors or TVs may simply be too slow to switch the pixels on time, resulting in better clarity at 60Hz.
- all the RealNC wrote above. Framerate dropping below the required line (60fps or 120fps respectively), double strobe issue etc.
Out of the whole DF team, John seems to be the most knowledgable in terms of motion clarity issues, but he's far from being an expert even on the non-technical level. It's unlikely, but the one last option, is that he mixed some things up. He knows enough to use BFI mode on his OLED and to play on CRTs when playing retro games. But "errare humanum est" (I probably mispelled this, I know)
- he could be talking about sample-and-hold display modes. I actually saw him doing that on one of the recent Digital Foundry videos. But he said the opposite - that 120Hz is better, obviously, due to less time spent in "paused" mode.
- he could be talking about specific TV or monitor. For example, the newest OLED TV line from LG for some stupid reason, does not support strobed (Black Frame Insertion) mode at 120Hz, but it does at 60Hz. Some monitors or TVs may simply be too slow to switch the pixels on time, resulting in better clarity at 60Hz.
- all the RealNC wrote above. Framerate dropping below the required line (60fps or 120fps respectively), double strobe issue etc.
Out of the whole DF team, John seems to be the most knowledgable in terms of motion clarity issues, but he's far from being an expert even on the non-technical level. It's unlikely, but the one last option, is that he mixed some things up. He knows enough to use BFI mode on his OLED and to play on CRTs when playing retro games. But "errare humanum est" (I probably mispelled this, I know)
Re: Sample & Hold Motion Blur 60FPS @60Hz vs. @120Hz
Shouldn't this exact case be a lot better with 120Hz and V-Sync (but without VRR)?
Let's say there's one frame ready in our frame buffer. But the next frame takes ~20ms to render instead of the necessary 16.7ms.
So the scanout phases would look like this (end time of the scanout in the square brackets) ...
60Hz
Scanout 1 [16.7ms]: frame 0 visible
Scanout 2 [33.3ms]: frame 0 visible (V-Sync kicks in and prevents tearing)
Scanout 3 [50.0ms]: frame 1 visible
120Hz
Scanout 1 [8.3ms]: frame 0 visible
Scanout 2 [16.7ms]: frame 0 visible
Scanout 3 [25.0ms]: frame 0 visible (V-Sync kicks in and prevents tearing)
Scanout 4 [33.3ms]: frame 1 visible
I.e. frame 0 with 60Hz is displayed for 33.3ms while with 120Hz it is displayed for 25ms. Which should lead to a less visible judder?!
Re: Sample & Hold Motion Blur 60FPS @60Hz vs. @120Hz
Sure, for reference I'm talking about this post and this one.
No doubt John has quite some knowledge in this area and it is not about proving him wrong either. We all make mistakes!RonsonPL wrote: ↑03 May 2022, 09:59Out of the whole DF team, John seems to be the most knowledgable in terms of motion clarity issues, but he's far from being an expert even on the non-technical level. It's unlikely, but the one last option, is that he mixed some things up. He knows enough to use BFI mode on his OLED and to play on CRTs when playing retro games. But "errare humanum est" (I probably mispelled this, I know)
Specifically, it's just about 60FPS content in a 120Hz container, without BFI on an OLED TV (i.e. sample & hold).
I simply do not agree with him in this case and have explained it accordingly.
Now I just wanted to cross-check if I missed something that wouldn't be immediately obvious. I also believe him when he says that he sees something subjectively, the only question is what exactly he sees in this case.
In my opinion, however, this is not due to any inherent problem of 60FPS in a 120Hz container.
Re: Sample & Hold Motion Blur 60FPS @60Hz vs. @120Hz
Depends on many things. If we're talking:
- sample-and-hold
- no VRR
- same overdrive behavior
- and flat frame times
than it would look identical but with less latency at 120Hz.
Interestingly for non-variable overdrives that are tuned for max refresh only, 60fps at 120Hz will have slightly less ghosting because pixels get two voltage boosts per refresh cycle (that's one way to achieve faster response times with less overshoot, compare RTINGS Odyssey G7 charts for OD adaptive-sync at 60-120-240Hz, they are all same). On fast IPS monitors for example 60fps at 240Hz would look almost identical to OLED's instant ghost-free 60Hz/60fps.
I highly recommend u try Smooth Frog and test out combinations you're interested in your self.
- sample-and-hold
- no VRR
- same overdrive behavior
- and flat frame times
than it would look identical but with less latency at 120Hz.
Interestingly for non-variable overdrives that are tuned for max refresh only, 60fps at 120Hz will have slightly less ghosting because pixels get two voltage boosts per refresh cycle (that's one way to achieve faster response times with less overshoot, compare RTINGS Odyssey G7 charts for OD adaptive-sync at 60-120-240Hz, they are all same). On fast IPS monitors for example 60fps at 240Hz would look almost identical to OLED's instant ghost-free 60Hz/60fps.
I highly recommend u try Smooth Frog and test out combinations you're interested in your self.
Compare UFOs | Do you use Blur Reduction? | Smooth Frog | Latency Split Test
Alienware AW2521H, Gigabyte M32Q, Asus VG279QM, Alienware AW2518HF, AOC C24G1, AOC G2790PX, Setup
Alienware AW2521H, Gigabyte M32Q, Asus VG279QM, Alienware AW2518HF, AOC C24G1, AOC G2790PX, Setup
Re: Sample & Hold Motion Blur 60FPS @60Hz vs. @120Hz
That's what I said. "120Hz, non-VRR mode".
Higher refresh looks better, but I'm not sure what the point is here. If you don't have late frames at 60Hz, there's not going to be any judder. If you use 120Hz, then you need to make sure to sync against every other display vsync. This isn't something that will just happen on its own. You need special support for this to happen. Llike a "1/2 vsync" driver option, or a game that natively supports setting the vsync swap interval, which no game I know of has. (Maybe it's possible in some game by tweaking the game engine ini settings.) RetroArch is the only software I know of that offers such option as a normal, user configuration.Let's say there's one frame ready in our frame buffer. But the next frame takes ~20ms to render instead of the necessary 16.7ms.
So the scanout phases would look like this (end time of the scanout in the square brackets) ...
60Hz
Scanout 1 [16.7ms]: frame 0 visible
Scanout 2 [33.3ms]: frame 0 visible (V-Sync kicks in and prevents tearing)
Scanout 3 [50.0ms]: frame 1 visible
120Hz
Scanout 1 [8.3ms]: frame 0 visible
Scanout 2 [16.7ms]: frame 0 visible
Scanout 3 [25.0ms]: frame 0 visible (V-Sync kicks in and prevents tearing)
Scanout 4 [33.3ms]: frame 1 visible
I.e. frame 0 with 60Hz is displayed for 33.3ms while with 120Hz it is displayed for 25ms. Which should lead to a less visible judder?!
If you just set a 60FPS limit in 120Hz, it's not going to be synced to the display. There's going to be hiccups. If you don't do something special to sync against every other frame, 60FPS@120Hz is never going to look good without VRR.
Steam • GitHub • Stack Overflow
The views and opinions expressed in my posts are my own and do not necessarily reflect the official policy or position of Blur Busters.
The views and opinions expressed in my posts are my own and do not necessarily reflect the official policy or position of Blur Busters.
Re: Sample & Hold Motion Blur 60FPS @60Hz vs. @120Hz
Hmm. Are you sure he meant sample and hold? The content at 60Hz displayed in sample and hold 60Hz and in 120Hz, should not differ.
It would've made much more sense if he was talking about the issue explained here:
Actually, no matter if you have 60fps on 120Hz or 60Hz (both in sample and hold mode), you can't see anything in fast motion anyway
(unless the 60fps gets interpolated to 120fps first, but that would no longer be 60fps content)
Personally, I can definitely say the motion clarity on 50fps at 50Hz is way better than at 100fps displayed at 200Hz, due to the issue illustrated above on the Chief's graphic. That's what I call "double strobe" issue to simplify.
edit: I just read the thread topic there, it's about VRR, so maybe John from DF talks about something specific to VRR mode
-
- Posts: 181
- Joined: 06 Aug 2015, 17:16
- Chief Blur Buster
- Site Admin
- Posts: 11653
- Joined: 05 Dec 2013, 15:44
- Location: Toronto / Hamilton, Ontario, Canada
- Contact:
Re: Sample & Hold Motion Blur 60FPS @60Hz vs. @120Hz
Welcome!
You’ve posted in the correct forums — and the above posters are (more or less) correct.
However, there’s additional human psychovisual factors that creates a double-image optical illusion on a perfect sample-and-hold display (no impulsing) — much in the same way that sometimes a vibrating string on a guitar looks like two strings. (stutter = a form of vibration around the human eye gaze axis)
Perfect sample-and-hold display (GtG=0ms) means 60fps looks perfectly identical at 60Hz, 120Hz, 180Hz, etc.
However, GtG curves can be different for 60Hz and 120Hz.
There is yet another cause: The fast GtG curve of OLED can raise the threshold of “slideshow to motion”. That’s why 60fps stutters more on an OLED than on a slow LCD.
Slower GtG lowers the threshold of stutter detectability (even for perfect frame pacing).
It is a complete continuum from stutter-to-blur: www.testufo.com/eyetracking#speed=-1
This is also observed in framerate ramping on VRR; www.testufo.com/vrr where low framerates looks like stutter and high framerates, the stutter amplitude becomes the blur amplitude (stutter and blur are the same thing: persistence …. And the variable is the frequency, aka the frame rate). Like how a slow-vibrating string vibrates visibly, and a fast-vibrating string blends into blur. Motion on a display (during eye-tracking situation) is the same.
A fast GtG will raise the framerate threshold of where stutter blends to blur, and vice versa. (be noted, the specific human will vary in threshold — the stutter blending to blur is related to a specific human’s flicker fusion threshold)
A faster-GtG display will raise the framerate where stutter goes too fast and blends to blur (like a fast vibrating guitar string).
The non-impulsed two-image effect can also be caused by the interpretation of the extremities of the stutter amplitude — the outer extremes of the stutter-vibration around the axis of the moving eye gaze.
This produces a phantom double-image effect even on a perfect 0ms GtG sample-and-hold display. This is also sort of observed at lower frame rates on LCDs too, at approximately ~20fps-ish though some humans say they see it in 30fps-ish, and others, 40fps or 60fps.
Concidentially, OLED displays have the highest threshold where stutter blends to blur, and some humans still see stutter on OLEDs at 60fps 60Hz, because the stutter-to-blur continuum threshold is above the number 60 for that specific human for that specific display (OLED).
You can watch the RTINGS video to understand how stutter and persistence blur are exactly the same thing:
Since this is a display physics question, I have moved this thread to the Area 51 forum.
More testing is needed for measuring the stutter-to-blur continuum thresholds for humans, because it is affected by many variables (the specific human’s flicker fusion threshold, the specific display GtG curve behavior, etc) but a common way to determine this is to watch a VRR framerate ramping animation and decide which framerate threshold — whereupon below is stutter and above is blur.
Sample-and-hold persistence blur is simply stutter vibrating too fast that it blends to blur (like a fast-vibrate guitar string) … that’s why fast-vibrating strings sometimes looks like two strings too.
(For simplicty’s sake: Assumes perfect frame pacing, and no VSYNC-related stutter — just stutter because of too-low framerate)
A related thread on retina refresh rate thresholds are:
Human Visible Sensitivity Thresholds
Head of Blur Busters - BlurBusters.com | TestUFO.com | Follow @BlurBusters on Twitter
Forum Rules wrote: 1. Rule #1: Be Nice. This is published forum rule #1. Even To Newbies & People You Disagree With!
2. Please report rule violations If you see a post that violates forum rules, then report the post.
3. ALWAYS respect indie testers here. See how indies are bootstrapping Blur Busters research!