Strange behavior in MAME ( lag/control )

The Core Mame has a LAG frame compared to the official version. I did the test with several games, for example Mortal Kombat (Like most games) has intentional delay in their movements.

  • Blow and kick in 4 frames in Mame, 5 frames in RA.
  • Lock, move and bend in 3 frames in Mame, 4 frames in RA.
  • Skip on 6 frames in Mame, 7 frames in RA.

The Mame Independent, I usually configure in the right joystick the gear lever for cars games (left -1 / down-2 / up-3 / right-4), to play them in manual, it is very comfortable. In Retroarch I can assign the option to the buttons in the control menu but do not work as expected. It seems that it was very close or several will be selected before touching a point, especially right. Try to configure the control in the MAME GUI but I can not get the menu with the control, I have to use keyboard and it does not work correctly.

Another thing I see, the shortcuts and buttons of Retroch and Mame are overlapped, all MAME should be deactivated.

I’m getting 4 frames of input lag for MK’s blow/kick in RA’s MAME (current), and exactly the same result in FBNeo.

Are you using the frame-advance method for counting latency? If so, that gives an illusory frame because of the way input is handed off between the frontend and the core, but if you measure with a high-speed camera, that additional frame doesn’t actually exist.


This happens to me by pressing the blow.

In Native MAME: 3 empty frames and run on 4.

In RetroArch with Mame, also in FBNeo: 4 empty frames and executes in the 5.

No, I’m using the default configuration, the latency options are deactivated.

I am pausing the game and advancing by frames with the “K” key. In MAME it can also be done with “Shift+P”.

Wouldn’t that be a “Yes” to the question? :grin:


I believe @hunterk is right about the illusory frame, most likely you got the same latency on both sides and would be able to confirm using a high-speed camera. What you refer as 5 is probably the same 4 frames of input lag i was mentioning (when measuring input lag in RA, you are supposed to start from 0 after pressing K the first time).

1 Like


Sorry. Then yes. :sweat_smile:

1 Like

At this moment I do not have a camera, when I have it, I do the test. But when you play if you notice the difference, especially in Mortal Kombat 2/3 that is very precise.

@hunterk What I did was place RetroArch next to MAME, remove the automatic pause and run a blow. It runs on both at the same time and you can notice the retardation in RetroArch.

1 Like

It was already tested with a high-speed camera. I can’t remember if it was brunnis or @Tatsuya79 , though.

I understand, I’m sure I’m self-suggested, that time to play perfect mk step long ago. :cry:

I made this video, captured at 60fps, I do not know if it will have any technical value, or the two together affects the retroarch behavior. In normal time and slow motion, in normal time you can advance by frames.

Well, it appears your MAME and RetroArch setups aren’t running at the same speed or something, it’s visible when observing closely the 2 timers, somehow your MAME is a few frames behind and that difference increases over time, i’m sure this is somehow related.


Isn’t MAME on Retroarch usually running a bit faster in order to sync all games at 60hz in non- gsync screens?

Framerate on RA is displayed in the video, it’s not 60, so that’s not the case here. Also, my point was that comparing apples with melons is probably a bad idea.

FWIW i investigated the mk driver in FBNeo, i’m confident the current input lag is the minimum we can achieve through normal emulation, and i also know the libretro port isn’t adding any input lag in FBNeo (because i’ve seen a few games reacting at next frame, meaning literally 0 frames of input lag), so i really doubt standalone MAME could achieve less input lag here, except if they did implement a runahead feature.

1 Like

No, I pressed PLAYER 1 separated. Here I did the test with a full Round. Assigning Player 1 on the same button. In fact, Retroarch loads a faster frame.

No idea because it does not show 60fps, Street Figther if running at 60 and has the same LAG.

I do not have settings apart from the directories. Either way, delete the configuration folders to do this test.

I discovered the delay because he was showing him a friend (of those times) as they were emulated, he did not know. We played a while in Mame and then we went to Retroch, it was when I felt the delay.

I do not want to sound presumptuous, but I was a high-level player of Mortal Kombat (2D) at those times. I have no way to check the delay, but if I tell you that it is delayed, it is because it has it. :tipping_hand_man:t2:

PS: NBNeo also has the delay, then, the error is not MAME.

I’ve noticed this as well, which has resulted in some frustration when it comes to playing Fighting games. It’s not just MAME but FBNeo in which I have had lag as well which really screws up delicate combos and tight input moves. Not just Retroarch but even in Retropie I’ve had experienced some lag. Doesn’t Retroarch have performance markers and other means to test for lag?

Street Fighter isn’t supposed to run at 60Hz, very few arcade games are supposed to run at 60Hz. There is also some desync between screens on your sf2 video.

Indeed, and i wonder if that illusory frame comes from the fact MAME is somehow delaying frame rendering, hence getting additional time to account for inputs. You’ll need a high-speed camera to verify the actual input lag, but as said above someone already did the comparisons and confirmed there is no additional input lag between MAME and its libretro core.

@hunterk If it is an illusion of the interface, then, in other emulators it should happen the same?

I compared another emulators with the cores.

  • Genesis: It is identical.
  • NES and SNES: In Retroarch it is even better, they have a frame less than the original emulator.

So, why is it delayed when I record the video in real time? And why does it also happen in FBNeo?

This week I’m going to get some time and I’m going to get a camera, I’m also going to hit a LED to the control button, a frame can go unnoticed if there is no electrical signal that notifies it.

A good way to feel the delay is, spends 10/20 fights in Mame and immediately happens to Retroarch.


Load MK1 in FBNeo, which has “the same delay as the Core Mame”, active Run-Ahead “4” and the delay disappears, the punch is executed instantaneous.

But, I discovered something curious. FBNEO has mixed delay. The punch is 5 as the Core Mame and the Block has 3 as independent MAME.

So, this would be like that.

Blow and kick.

  • 4 Mame
  • 5 RA-Mame
  • 5 RA-FBNeo

Block. Move and crouch

  • 3 Mame
  • 4 RA-MAme
  • 3 RA-FBNeo


  • 6 Mame
  • 7 RA-Mame.
  • 6 RA-FBNeo