Input Lag

Gaming-related threads & forum meta.
Post Reply
User avatar
Porcupine
Virtual-On Positive
Posts: 592
Joined: 19 Jul 2012, 19:07

Input Lag

Post by Porcupine »

http://www.displaylag.com/video-game-in ... -database/

Finally, here's proof of something I've been telling fighting game players for decades. (No one ever disagreed with me, just that it was hard to confirm). All computers and video game consoles have inherent input lag. According to the link, it varies from a minimum of 4 frames for the best games, 5 frames for many games, down to 8 frames in some games. The link also contains their testing methods and further explanation. Depending on how you view their testing methods, and multiple other theoretical considerations, you might have a slightly different interpretation of these results, by a +1 frame difference. (I've given the games the benefit of the doubt here, and subtracted 1 frame from their tested results).

Previously, I always justified my claim by theoretical arguments (which changed slightly over the years, as hardware technology changed). Assume that the monitor displays its image all at once, at the very instant that image has been fully received, instead of scanning like in the CRT days. (Plasma is said to display synchronously. LCD monitors supposedly still scan to display, however many scan the whole image much quicker than one frame, in roughly 2 ms, which can be taken as instant). If you press a button the exact instant the monitor displays frame 0, the monitor is currently being sent the signal for frame 1. The console must then be readying frame 2 for output. This consists of overhead tasks that are effectively outside the game engine, such as copying an already-rendered game image to the proper display memory, rescaling the image to the output resolution, then converting the image to an HDMI or analog signal. All these tasks don't need to take up 1 whole frame of time, but the console could be hard-coded to allocate 1 frame to this step. Meanwhile, the game must simultaneously be rendering frame 3. It has exactly 1/60th of a second to complete this task, otherwise the game slows down. Your button press cannot affect this frame, because it has already started to be drawn. Finally, frame 4 is the earliest possible frame that can show the result of your button press.

The overhead step doesn't have to exist, so it's possible to have a console with 3 frames of input lag. In the CRT days, on a CRT with no digital processing, or only line-by-line digital processing, it was possible to have a console with 2.5 frames of input lag, assuming you react to an image at the center of the screen.

http://www.displaylag.com/display-database/

Lately, third parties have started to properly measure the input lag of many monitors and TVs as well. It appears that most computer monitors have a display lag of 10 ms. ASUS isn't the fastest like everyone kept claiming all these years. Their monitors are the same as the every other brand. Bad monitors, and good TVs in game/PC modes have display lags of 16 ms to 33 ms, so 1~2 frames. Most TVs are at least capable of keeping the display lag under 48 ms, or 3 frames. Bad TVs, or most TVs with too many processing features turned on, have display lags of up to 100~150 ms, or 6~9 frames.

The dangerous thing with TV input lag is that it varies wildly with the particular model, with no officially published results. Studying all the data carefully, there is no correlation with manufacturer, year, type of display (LCD, plasma, etc), cost, resolution, refresh rate, response time, or anything else. It's a total crapshoot. The only pattern is that very similar models, from the same manufacturer, from the same year, may have a similar input lag to each other. Even that is not guaranteed, as I recently found out Samsung TVs have a secret sub-model number associated with them. Usually there are several sub-models per TV model, and you cannot guarantee which you will get unless you buy in person and look on the actual item's box. The sub-models are made in different countries, and may not even be the same type of display! (LCD and plasma won't mix, but you may get an LCD panel with fundamentally different pixel shape and performance).

http://www.rtings.com/tv/tests/video-games/input-lag

The Samsung UN40JU6500 bedroom television I just bought has a 1.5 frame lag in game mode, but a 3 frame lag in PC mode (which is unfortunately the one I'm forced to use because of other quirks of this model). I noticed the difference before I read the data regarding it. I play my Xbox 360 online in the living room on a Sony KDL-46W5100 though. I can't find lag data on it but comparing to my Samsung it feels better, so I'm guessing it might be 1 frame lag.
User avatar
twistedsymphony
Virtual-On Positive
Posts: 212
Joined: 25 Feb 2013, 16:46
Gamertag: twistedsymphony
Location: NH, USA
Contact:

Re: Input Lag

Post by twistedsymphony »

IMO measuring input lag in the game code is moot, everyone playing that game is going to experience the same amount of lag so does it really matter if Killer Instinct has lower lag than Ultra Street Fighter IV?

I suppose it's beneficial for cross-platform comparison but ultimately you're always going to have a level playing field.

Input lag due to the controller/joystick is definitely something worth caring about though. here's a good resource for that: http://www.teyah.net/sticklag/results.html
User avatar
Porcupine
Virtual-On Positive
Posts: 592
Joined: 19 Jul 2012, 19:07

Re: Input Lag

Post by Porcupine »

Nice link, too bad they didn't measure many controllers. Come to think of it, I wonder if input lag could vary between models of the same console? One difference I know is the slim PS3s have a lower quality audio DAC than the fats. This only affects people who use the analog audio out. The obvious difference, there is no highpass frequency filter, so in the system menu all keystrokes go "click....PAK!!" from the recording ending in a clipped wave. Aside from that it was a very low quality, jittery DAC in general. They could very well use different upscaling and video DAC chips as well.

Non-universal lag is of course a major advantage, but universal lag is originally what concerned me. I was talking about this subject decades ago, when there was no online play and everyone played on the same hardware. Knowing the input lag still helps in theorizing the best ways to play. For example, a few people in fighting games might say "I can block this 6-frame punch on reaction" or "I can react to the hit spark of this throw and break within its 4-frame window" and you could respond "Unlikely, with the 5-frame input lag of this game." Beyond that, knowing input lag lets you figure out how feasible various things are. Is reacting in 10-frames just slightly harder than reacting in 13-frames? Or is it twice as hard, because there is a 7-frame input lag to begin with?

Such knowledge can also help timing in frame-perfect maneuvers, either for glitches or merely a general desire to execute more tightly. In VO games, do the sticks have the same input lag as the turbo button, and the triggers? I know parts of the answer. In VOOT, the triggers in close combat respond at least one frame slower than in projectile combat. In both cases, the triggers also respond at least one frame slower than the turbo button and the sticks. This is both by feel and from theory. Any game that lets the user do simultaneous button press attacks (in this case, CW trigger) should wait an extra frame to make it easier to press both buttons at the same time.

Just this past year, I also discovered (using a turbo controller set to various speeds) that in all VO games, pressing CW trigger for any action responds up to one frame quicker than pressing LT or RT alone. As soon as both triggers are pressed, the game responds. But if only one trigger is pressed, the game waits for the other, before shooting. Most other games don't do that, they wait either way. Knowing this is helpful in several ways I'll talk about later. The turbo controller method also confirmed that VOOT CC has an extra delay (didn't test VOOM and Force) except for some reason jumping CC does not.
User avatar
twistedsymphony
Virtual-On Positive
Posts: 212
Joined: 25 Feb 2013, 16:46
Gamertag: twistedsymphony
Location: NH, USA
Contact:

Re: Input Lag

Post by twistedsymphony »

some very good points that I hadn't considered. Also interesting about the CW in VO, but it makes sense since they game would need to know if you're trying to fire LW or RW separate or if you're firing CW and just haven't pulled the 2nd trigger yet.
Post Reply