Emulator comparison in Any% NMG (And possible re-revision of emulator rules)
2 years ago
Finland
ModeratorAhvena
He/Him, They/Them
2 years ago

This issue has been under discussion several times on both CTR speedrunning discord and CTR Time Trials discord so I decided to initiate a thread here so that the topic wouldn't be lost under ongoing other topics.

At the moment in the Any% NMG emu category most of the runs have been done with either Mednafen 1.26.1 and Retroarch 1.9.0 (beetle PSX HW).

Some of you may know that during this year I have been gathering data from both console runs and emulator runs. I compared how big of a weight hub movement and load times had on the resulting RTA time.

Mentioned IGT sheet can be seen here: https://docs.google.com/spreadsheets/d/1KzfMfQpnNhN262IUZ9h-yf9zUT-jK0rldnO6oZcxvF4

Of course there might be a lot of variables I haven't looked into since hub movement is dependable on how optimized runners hub world movement is (plus PAL vs NTSC differences). Also lag plays considerable role here too. Dataset could be much bigger too. But in my opinion this comparison still gives pretty good estimate to which emulator performs better.

Fastest Mednafen hub+load: 17:51.470 (NTSC-U with 3 pauses) Fastest Retroarch hub+load: 17:32.550 (NTSC-U with 3 pauses) Slowest Mednafen hub+load: 18:10.460 (NTSC-U with 0 pauses) Slowest Retroarch hub+load: 18:01.010 (NTSC-U with 0 pauses) Mednafen delta: 00:18.990 Retroarch delta: 00:28.460 Average Mednafen hub+load: 18:00.914 Average Retroarch hub+load: 17:45.123 Average delta: 00:15.790

As you know, on console hardware side CTR speedrunning is a bit region locked so I made same comparison but with PAL vs NTSC-U

Fastest PAL hub+load: 14:43.940 Fastest NTSC-U hub+load: 14:45.910 Slowest PAL hub+load: 15:02.350 Slowest NTSC-U hub+load: 15:09.970 PAL delta: 00:18.410 NTSC-U delta: 00:24.060 Average PAL hub+load: 14:54.941 Average NTSC-U hub+load: 15:01.942 Average delta: 00:07.001

This makes it pretty clear that Retroarch has quite big advantage compared to Mednafen.

So point of this all is to raise discussion if only one emulator should be allowed for CTR.

TheRedhotbr, Natty and 3 others like this
Portugal

imho, I'm fine with just having one emulator only for it, especially because mednaffen is MUCH easier to setup than Retroarch and to run it, also probably the simplest to just stream and not heavy on the machine.

There's also a load/auto-splitter, done by Redhot, for mednaffen that works wonders and was shared on the Crash discord and also the Time Trial discord.

I was pretty pro-retroarch at the start just because it would work smoother on my pc but I've switched to mednaffen and it's ALOT less taxing on my PC, especially when I stream. It's a win-win in a way.

Ahvena likes this
Finland

This should be a no-brainer at this point. Only allowing mednafen is the way to go to ensure equal competition for everyone.

Zibang and Ahvena like this
United Kingdom

Agree with what has been mentioned already and think we should just go with Mednafen. I understand the emulator is fairly lightweight compared to some others so it should be fine for streaming/recording & playing. Also with RedHot's load remover and auto splitter he has developed that should allow for us to track total IGTs and consider removing loads if we see fit.

I don't think we need to allow more than one emulator either, regardless of the rules on the main Crash boards (and make sure mods update the CTR rules accurately if/when a decision has been made). The main argument I have heard is that Mednafen currently is not compatible with Mac, but I don't think this issue is worth throwing away our chance at fair competition.

There's also the issue of current runs on the boards which used RetroArch. I don't think the advantage is huge enough for these to be removed (although WR is on that emu when we already knew it was unfair :-D)

Ahvena, Spikestuff, and FreezeChamp like this
Victoria, Australia

[quote=Natty]I don't think we need to allow more than one emulator either, regardless of the rules on the main Crash boards (and make sure mods update the CTR rules accurately if/when a decision has been made).[/quote]

That's a hard agree I personally have with, some other things are going to be done soonish™ for the broader range, but focusing on the narrow field that is CTR which is a lagish game, and is dependent on a few things, allowing freedom of what you choose as your choice is completely fair for the community.

Especially when you can back it up in that narrow field.

Edited by the author 2 years ago
Ahvena and FreezeChamp like this
Finland
ModeratorAhvena
He/Him, They/Them
2 years ago

My own opinion on this is also pretty clear.

Make newest version of Mednafen only allowed emulator. Now we are also able to use RedHot's load removed which is already integrated with livesplit. Only downside is that Mednafen is not available for Mac but I guess some sacrifices need to be made. Point on current WR stands but I also agree that there's no reason to remove those times because they are not in any way unbeatable.

There's also enough reasons for CTR to have their own set of emulator rules.

FreezeChamp, Faceoff, and Casu like this
France

Made the switch from RA to Mednafen recently and it is indeed a big buff in loading time and input lag. I also agree on the mednafen last stable version rule, and as i dont know much about rule changes in SRcom, what would be the next step for the rule to be applied ?

Edited by the author 2 years ago
Ahvena likes this
United Kingdom

Is it possible to run Mednafen on Apple Silicon? At least on the surface it seems it doesn't work with CrossOver or Parallels on M1?

Ohio, USA

Would Duckstation be acceptable in the future? That is what I have been using runs on.

Victoria, Australia

[quote=FuzionSynth]Would DuckStation be acceptable in the future?[/quote]

You missed the brief period of time where DuckStation was an acceptable emulator for CTR. As outlined and explained here CTR wants to keep it to a single static emulator and that emulator is mednafen.