Game boy timing comparisons

Blechy

Edit history:

Blechy: 2014-07-17 12:09:47 am

Blechy: 2014-07-16 11:45:53 pm

Blechy: 2014-07-16 11:38:31 pm

2014-07-16 11:34:03 pm

So for awhile now, I've seen a lot of discussion about the VBA emulator having inaccurate timing. The TASVideos page here measures its general accuracy as an emulator, but doesn't draw specific conclusions regarding timing, so I decided to take this task on myself. In my test I compared the following play methods:
• SGB2
• GBPlayer
• VBA 1.7.2 v24 SVN 480
• VBA-M 1.8.0-SVN
• bgb 1.4.1
• Gambatte Qt (r550)

Here is the resulting video:

Method: I recorded each individual video by making a local recording via OBS @ 60fps. I'm aware that this is not completely ideal given that native GB framerate is ~59.73, but the results should be accurate regardless. After the fact, I eyeballed it between play methods without worrying about recording, and the results matched what is shown here.

Conclusion: VBA (and to a lesser extent VBA-M) are simply not accurate for GB emulation in respects to time, and bgb and Gambatte are much better. Again, I've heard claims to this effect for some time now, but have never seen video proof supporting it like this.

Separate SGB1 notes: I did my own tests with SGB1 and my results matched those of Radix's exactly: a ratio of 1.024 (his method measured it even more precisely).

Separate GBA notes: I compared GBPlayer to the VBA and VBA-M versions I list above, and the results were quite different. VBA-M faster than GBP faster than VBA. This is rather unfortunate, as VBA-M supports some games that VBA does not.

Thank you very much to Latte for compiling the videos I created, to Sluip for coordinating with Latte, to cyghfer for letting me use his gear to record, to Krystal for helping me to know what I'm doing, to neskamikaze for helping to create an unshown SGB1 vid, to both nes and Garrison for creating videos that wound up not being shown here, and to anyone else involved.

Thread title:

So for awhile now, I've seen a lot of discussion about the VBA emulator having inaccurate timing. The TASVideos page [url=http://tasvideos.org/EmulatorResources/GBAccuracyTests.html]here[/url] measures its general accuracy as an emulator, but doesn't draw specific conclusions regarding timing, so I decided to take this task on myself. In my test I compared the following play methods:
• SGB2
• GBPlayer
• VBA 1.7.2 v24 SVN 480
• VBA-M 1.8.0-SVN
• bgb 1.4.1
• Gambatte Qt (r550)

Here is the resulting video:
https://www.youtube.com/watch?v=doNLo-mn8Pw&feature=youtu.be

[b]Method:[/b] I recorded each individual video by making a local recording via OBS @ 60fps. I'm aware that this is not completely ideal given that native GB framerate is ~59.73, but the results should be accurate regardless. After the fact, I eyeballed it between play methods without worrying about recording, and the results matched what is shown here.

[b]Conclusion:[/b] VBA (and to a lesser extent VBA-M) are simply not accurate for GB emulation in respects to time, and bgb and Gambatte are much better. Again, I've heard claims to this effect for some time now, but have never seen video proof supporting it like this.

[b]Separate SGB1 notes:[/b] I did my own tests with SGB1 and my results matched [url=https://kb.speeddemosarchive.com/Super_Game_Boy_timing]those of Radix's[/url] exactly: a ratio of 1.024 (his method measured it even more precisely).

[b]Separate GBA notes:[/b] I compared GBPlayer to the VBA and VBA-M versions I list above, and the results were quite different. VBA-M faster than GBP faster than VBA. This is rather unfortunate, as VBA-M supports some games that VBA does not.

Thank you very much to Latte for compiling the videos I created, to Sluip for coordinating with Latte, to cyghfer for letting me use his gear to record, to Krystal for helping me to know what I'm doing, to neskamikaze for helping to create an unshown SGB1 vid, to both nes and Garrison for creating videos that wound up not being shown here, and to anyone else involved.

cyghfer

2014-07-16 11:45:00 pm

sick video

StingerPA

2014-07-17 12:16:59 am

Good stuff. What GB core does Bizhawk use?

Blechy

2014-07-17 12:20:53 am

Unfortunately my rig is not powerful enough to local record while Bizhawk is going without it affecting the video, so I did not include it & don't know its technical details. I mainly wanted to focus on the more accessible emulators.

Warepire

2014-07-17 01:53:44 am

Heavy Metal Powered

BizHawk uses Gambatte as a core.

bangerra

2014-07-17 09:46:43 am

Good to see GBPlayer and SGB2 are indeed identical in speed.

Partystar

Edit history:

Partystar: 2014-07-17 10:08:20 am

2014-07-17 10:07:24 am

Thats a really cool comparison video! and indeed good to know that the Super Game Boy 2 has the same speed as the GB Player. I wonder how accurate the 3DS VC for Game Boy is. I feel the NES VC has lag at least with SMB it feels like it, correct me if I'm wrong. But for game boy that might be a diferent story of course.

Greenalink

2014-07-17 10:37:45 am

DS Dictator

I can record 3DS VC of Super Mario Land directly if you want me to?

LeonPowPow

2014-07-17 01:27:18 pm

Super cool

Radix

2014-07-17 05:41:57 pm

I'm addicted to games

Cool to see my old numbers still hold up!

pld

2015-11-20 02:16:32 am

OneHans

I was thinking about doing a comparison video of the Game Cube GB Player, Super Game Boy 2 and Super Game Boy because I haven't found a video yet which compares the two accurate platforms (GB Player and SGB2) with the inaccurate one (SGB1). I could need a little help with this one though. I own all 3 devices and I can capture them locally (I think I will go with Duck Tales for GB since this game has an in-game timer as well). Now my question how can I arrange all 3 recordings in one video like in the video above? Can I add video files in OBS or how do I do it?

Sir VG

2015-11-20 06:49:17 am

Fucking Weeaboo

The easiest way is to use an AVS file combined with VirtualDub. I've done 2 side by side videos myself, and 3 would just be a slight variation of it.

Code:

a = AviSource("Path to video file with video file name")
b = AviSource("Path to video file with video file name")

StackHorizontal(a,b)

That is what I use for stacking 2 horizontal. You should just have to add a "c" variation to get 3 horizontal.

Code:

a = AviSource("Path to video file with video file name")
b = AviSource("Path to video file with video file name")
c = AviSource("Path to video file with video file name")

StackHorizontal(a,b,c)

To create an AVS, just put that into a text file and save it with an AVS extension instead of a TXT extension. And as long as you have Avisynth installed, you should be good to go. Open the AVS in VirtualDub.

pld

2015-11-20 08:00:24 am

OneHans

wow, that will help me a lot, thanks

charleon

2015-11-20 08:03:05 am

What is a man?

thanks Sir VG, I've personally had this question for other purposes as well for a long time Cheesy

pld

Edit history:

pld: 2015-12-01 07:10:04 am

2015-12-01 01:36:05 am

OneHans

@Blechy I hope it's ok when I continue your thread for my findings as I think it fits in this thread.

I haven't seen a video yet which compares the timings of the different Game Boy players available for different consoles so I felt like fixing that. In my video I have compared the original Super Game Boy (SGB1, top left recording), Super Game Boy 2 (SGB2, top right recording) and Game Cube Game Boy Player (GBP, lower recording). All 3 devices have been recorded with the GV-USB2 capture card at 59.94 frames per second. I have used Amarec 3.10 as capture software and the raw files have been encoded with Yua. All recordings start and end at the same frame of the game to ensure accuracy. As game I have picked Castlevania Legends as it features an in-game timer which doesn’t just go for 5 minutes but a bit longer.
I have disabled the audio of the Super Game Boy 1 because it is running way faster than the other 2 devices and that would cause one big mess with the audio (audio is already a bit off because of different speeds of SGB2 and GBP). Not to mention the disturbing high pitch of the SGB1.

The Super Game Boy 1 and 2 have been attached to my Super Famicom and the Game Boy Player has been used with my Game Cube.

Lengths of the recordings in the video:

Super Game Boy 1: 40434 frames (@59.94 = 11:14.34)
Super Game Boy 2: 41413 frames (@59.94 = 11:30.54)
Game Boy Player : 41379 frames (@59.94 = 11:30.20)

It was already known for years that the Super Game Boy 1 is running faster than original Game Boy’s frame rate so no surprise here. SGB1 is running 979 frames (00:16.20) faster than the Super Game Boy 2. What surprised me a little bit though is that the Game Boy Player is also not as accurate as the SGB2 (GBP is running 34 frames faster or 00:00.34 than SGB2). It’s not as dramatic as the SGB1 but still worth mentioning.

Also I wanted to say that I have tested different games as well (Super Mario Land 2 for example) with similar results (SGB1 running like crazy, GBP several frames faster than SGB2) so it’s not an issue of a single game and seems to be hardware related (or maybe software related/ GB Player disc?). This needs additional testings from someone who is more knowledgeable than me.

What I haven’t tested yet is the GBI (Game Boy Interface, http://www.gc-forever.com/forums/viewtopic.php?f=37&t=2782) for the Game Boy Player which is said to fix the „speed issue“ of the GBP. Would be nice if someone with GBI and SGB2 could test and compare them.

@Blechy I hope it's ok when I continue your thread for my findings as I think it fits in this thread.

https://www.youtube.com/watch?v=alV7QTMewSA

I haven't seen a video yet which compares the timings of the different Game Boy players available for different consoles so I felt like fixing that. In my video I have compared the original Super Game Boy (SGB1, top left recording), Super Game Boy 2 (SGB2, top right recording) and Game Cube Game Boy Player (GBP, lower recording). All 3 devices have been recorded with the GV-USB2 capture card at 59.94 frames per second. I have used Amarec 3.10 as capture software and the raw files have been encoded with Yua. All recordings start and end at the same frame of the game to ensure accuracy. As game I have picked Castlevania Legends as it features an in-game timer which doesn’t just go for 5 minutes but a bit longer.
I have disabled the audio of the Super Game Boy 1 because it is running way faster than the other 2 devices and that would cause one big mess with the audio (audio is already a bit off because of different speeds of SGB2 and GBP). Not to mention the disturbing high pitch of the SGB1.

The Super Game Boy 1 and 2 have been attached to my Super Famicom and the Game Boy Player has been used with my Game Cube.

Lengths of the recordings in the video:

Super Game Boy 1: 40434 frames (@59.94 = 11:14.34)
Super Game Boy 2: 41413 frames (@59.94 = 11:30.54)
Game Boy Player  :  41379 frames (@59.94 = 11:30.20)

It was already known for years that the Super Game Boy 1 is running faster than original Game Boy’s frame rate so no surprise here. SGB1 is running 979 frames (00:16.20) faster than the Super Game Boy 2. What surprised me a little bit though is that the Game Boy Player is also not as accurate as the SGB2 (GBP is running 34 frames faster or 00:00.34 than SGB2). It’s not as dramatic as the SGB1 but still worth mentioning.

Also I wanted to say that I have tested different games as well (Super Mario Land 2 for example) with similar results (SGB1 running like crazy, GBP several frames faster than SGB2) so it’s not an issue of a single game and seems to be hardware related (or maybe software related/ GB Player disc?). This needs additional testings from someone who is more knowledgeable than me.

What I haven’t tested yet is the GBI (Game Boy Interface, http://www.gc-forever.com/forums/viewtopic.php?f=37&t=2782) for the Game Boy Player which is said to fix the „speed issue“ of the GBP. Would be nice if someone with GBI and SGB2 could test and compare them.

Sir VG

2015-12-01 05:53:09 am

Fucking Weeaboo

Hmmm, I have the GBP and my GC is modded (Qoob). Hopefully I can remember how to boot DOLs.

ZenicReverie

2015-12-18 04:43:43 pm

Waiting hurts my soul...

Would be nice to see a test on original GB/GBC/GBA/DS hardware as well.

MeGotsThis

2015-12-19 07:26:35 pm

Here is a different perspective on the timing for the Gameboy Player. This one is about the Input Delay.

It is quite known that there is some input lag on the Gameboy Player from the Startup Disc software. But no one seem to have a definite number of how much lag there is. Using the wavebird controller and gamecube adapter, we can measure how much delay it really is compared to emulators at least. The Gameboy Player seems to have 8 frames of input delay compared to emulator.

Also I can compared the lag between different platforms if requested. All I can to get an adapter for gamecube/wavebird to whatever platform and assume that adapter has no to very minimal lag (i.e. Super Mario 64 on N64 vs Wii vs Wii U. I do plan to test with the Gameboy Player Interface to see what kind of input delay there is.

presjpolk

2015-12-21 05:20:56 am

HELLO!

Isn't that the Reddit video that used an LCD and a wavebird and so introduced lag unrelated to the GBP?

kirbymastah

Edit history:

kirbymastah: 2015-12-21 06:43:53 am

2015-12-21 06:33:46 am

<(^_^)>

Quote from presjpolk:

Isn't that the Reddit video that used an LCD and a wavebird and so introduced lag unrelated to the GBP?

Yes. Apparently he didn't quite learn that his "experiment" is completely flawed and have terrible results.

also he used an HDTV

presjpolk

2015-12-21 07:50:31 am

HELLO!

Heh. I mean there *is* lag. But when I tried to measure it, I used a wired controller and my CRT BVM, messing with a frame-perfect trick, and from that I tried to estimate the lag. But it wasn't 8 frames. If I recall it was ~3?

kirbymastah

2015-12-21 08:16:28 am

<(^_^)>

Quote from presjpolk:

Heh. I mean there *is* lag. But when I tried to measure it, I used a wired controller and my CRT BVM, messing with a frame-perfect trick, and from that I tried to estimate the lag. But it wasn't 8 frames. If I recall it was ~3?

People generally agree that the GBP introduces ~2 frames of input lag at worst. 8 frames is basically unplayable for speedruns lmao

Emptyeye

2015-12-21 08:44:21 am

Talk to the Hand

Quote from kirbymastah:

People generally agree that the GBP introduces ~2 frames of input lag at worst. 8 frames is basically unplayable for speedruns lmao

Yeah, I know Pres and I have had some conversations for the GBP and input lag. Now, yes, I have a lot of GBP experience what with speedrunning Metroid II for ~2.5 years, so I'm probably used to whatever lag is there by now to the point that I wouldn't notice it. That said, I like to think I'd freaking notice 8 frames of lag (Which is roughly, what, an eighth of a second delay?), especially before I was used to it.

TheMageling

Edit history:

TheMageling: 2015-12-21 09:54:44 am

2015-12-21 09:54:29 am

Quote from Emptyeye:

Yeah, I know Pres and I have had some conversations for the GBP and input lag. Now, yes, I have a lot of GBP experience what with speedrunning Metroid II for ~2.5 years, so I'm probably used to whatever lag is there by now to the point that I wouldn't notice it. That said, I like to think I'd freaking notice 8 frames of lag (Which is roughly, what, an eighth of a second delay?), especially before I was used to it.

Yeah, 8 frames would be ridiculous. I would definitely have noticed that much difference going from emulator for practice to GBP for actual runs in multiple games. 2ish frames seems legit, just enough to screw me up in the Goron dancing minigame in Oracle of Ages.

presjpolk

2015-12-21 01:20:28 pm

HELLO!

However it seems even smaller if you use the Game Boy Interface ultra low latency player instead of the standard one.

http://www.gc-forever.com/forums/viewtopic.php?t=2782

Which is easily played with this (bought inexpensively from the maker in the UK, worked completely fine on my US GC)

http://uk.codejunkies.com/Products/GCNWii-SD-Media-Launcher__EF000580.aspx