Retropie 4.3 INput Lag - Help Wanted

mikemolang

Thank you.

The Samsung played the same as the BenQ monitor, both which outperformed the Sanyo. Because of this (and the fact that the BenQ gets rave reviews as being a fast gaming monitor, I am really doubting it is a monitor problem. In fact, my bet is that anyone else running the same config as me is going to get the same results, whether they think theirs is fast, slow, or indifferent. As dankcushions said, "maybe 6 frames is 'normal'", Indeed, that is all I really want to prove at this point.

From the beginning I have wondered out loud if I was just being too critical, and said that I was really hoping that someone could provide factual data. So, I have a new idea!

TO ALL:

If you would be so kind, post a video as I did here (

) and here (

) for your setup, displaying the games shown. I will analyze your video to determine your input lag. That is all that is important thing to me right now; quantitatively determining the amount of input lag that people are getting on these games, regardless of their hardware or configuration.

If everyone's is the same as mine, I can be done.
If anyone's is better than mine, I can know what is different.
If mine is better than everyone else's... Well, you get the idea.

Best,
MML

dankcushions

@mikemolang i might do a video, but inputlag is a well researched topic for retropie, with all the "factual data" you could ever want. you should read https://retropie.org.uk/forum/topic/2019/an-input-lag-investigation

mikemolang

@dankcushions
Yes, I love his page. Read and re-read may times. And I still have the same basic question as before...

For anyone's configuration that uses RPi3B, Retropie 4.3, lr-mame2003 emulator, and 0.78 ROMs, when playing these specific games, what is your input lag measured in frames, (or please provide a video so I can compute the input lag in frames).

Then, quantitatively, if everyone's input lag under these conditions (but with whatever controllers, monitors, etc. that they happen to use) is essentially the same as mine, I am done. No need to bother with different monitors, controllers, emulators, settings, or anything else. If they are the same, there is nothing more to learn. I got my answer.

If anyone's is better than mine, I can know how much better, I can learn what is different about THEIR SYSTEM (as opposed to learning all the things that can be made different in general), and then I can proceed to the next step (if warranted) by addressing the specific things that are different.

Until I have an apples-to-apples difference in terms of frame lag, suggestions of things to change and discussions of how different things act in different ways won't answer my question.

As my boss is fond of saying, "I asked what time it is, but you keep telling me how a watch works. I need to know what time it is." (BTW, that comment is not directed at you in particular. Please, take no offense.)

Regards,
MML

dankcushions

@mikemolang what i'm saying is you already have an abundance of scientific data on that page. rather than recreate it in a new scenario, why not recreate the scenarios in the link and see how your system compares to them? only if your system has the same results as brunnis does it make sense to do mame2003-specific testing.

mikemolang

@dankcushions

Fair question!

He has a different RPI, and older Retropie version, and (unless I am mistaken) some of the emulators he is using are not standard in retropie.

I am trying to baseline the modern hardware and retropie revision with default config.

It's ok if you choose not to upload a video.

Thanks,
MML

dankcushions

@mikemolang said in Retropie 4.3 INput Lag - Help Wanted:

He has a different RPI,

he's the same RPI (RPI3B is the only RPI3, other than the compute module).

and older Retropie version

retropie version shouldn't matter. emulator version does, but look at the original libretro forum thread and it has been updated a few months ago with current data.

and (unless I am mistaken) some of the emulators he is using are not standard in retropie.

they are all available on retropie + pi3, except bsnes.

you're misunderstanding what i'm saying. we have a bunch of data on libretro and input lag. you're using a libretro emulator. this data is your baseline. you should retest with the same scenarios as brunnis, and then we rule out everything that isn't mame2003 and/or the specific games you're using.

comparing videos from two different cameras/setups isn't scientific. the only useful comparisons you could make are across emulators on your own setup. then you can find out if mame2003 and/or these games have specific issues.

dankcushions

on this link: https://forums.libretro.com/t/an-input-lag-investigation/4407

test cases 1 and 3 are basically 'vanilla retropie' - these would be the ones i would compare against.

jrburke99

I have dealt with this for some time as well. Mostly noticeable in Super Mario Bros. (NES). I was abel to tighten up my controls noticeably by setting Frame_Delay to 5. This gave me much tighter controls, and I didn't get any bad side-effects. I was running RetroPie 4.2. After upgrading to 4.3, I found that the lag has come back. I now have to set frame_delay to 8, and even then, it is a tad too laggy to play a tight game. I may have to go back to my 4.2 build.

I also have an NES Classic that is hacked to play other systems. There is no noticeable controller lag on that system. And it is nothing more than a single-board computer running a flavor of Linux and some emulators (Retroarch as a matter of fact). Controls are very tight. Feels like playing on actual hardware. And it doesn't matter what TV I put it on, or whether I have Game mode turned on.

AndersHP

Is this an effect of the update from 4.2 to 4.3 or did you just start with your first install?

jrburke99

@andershp said in Retropie 4.3 INput Lag - Help Wanted:

Is this an effect of the update from 4.2 to 4.3 or did you just start with your first install?

If you are asking me, this is a result of the upgrade, AND with a fresh install. My setup was running great. I had little-to-no input lag. Everything was great. I imaged my SD card as a backup.

Put on a fresh 4.3 install, configured it exactly the same as my 4.2 build. I have written instructions that I follow so that my builds get all the correct settings. But with 4.3 and the same settings, I found input lag again in Super Mario Bros. (running on lr-fceumm). I double checked in the RetroArch GUI menu while running lr-fceumm, and my settings were taking correctly. I use frame_delay=5 to get around the input lag. I had to set this up to 8, and even then, it was tad laggier than with my 4.2 build.

I went back and re-imaged my card back to my 4.2 backup, and everything was tight again. From there, I ran Update All Installed Packages to get it up to 4.3. Rebooted, and again, I saw worse input lag.

So I see the same result whether it is an upgrade, or a fresh install.

I also have noticed Playstation slowness in lr-pcsx-rearmed. With my 4.2 build, it all worked great with the enhanced resolution and speed hack. Upgrading to 4.3, PSX games are a bit slower.

AndersHP

Just did the 4.3 update, and tested Sonic (Megadrive), Kirbys Adventure (NES) and Gyruss (MAME2003) with no input lag.

But I also had some input lag when I first started my RetroPie build. I read the above mentioned investigation, and here's what I think helped:

In my retroarch.cfg file, in the configs/all, I run with these settings (not sure which of these is default, so here's all of them related to video):

video_disable_composition = "true"
video_max_swapchain_images = 2
video_frame_delay = 5
video_threaded = "false"
video_smooth = "false"

If anyone has suggestions to these, I'll be happy to try them out.

Omnija

I didn't read the whole thread but noticed the power supply is connected to an extension. Normally extensions don't always carry power well, so i was wondering if you tried directly into the wall?

mikemolang

Ok, I give up. The request was simple and would give developers early feedback on real world performance with each release, which seems to be of some value, based on the feedback above on rev 4.3. It would also allow newbies to quickly check their setup against a standard benchmark and take away a lot of subjectivity, saving them and the support team countless hours of chasing problems that do not exist.

Good luck,
MML