ZERO_MILES
Enthusiast
I'm not a professional player, should I bother? I only know how to use regular I/O boards.
- Thread starter Niko
- Start date
What you are saying is true, I meant that maybe is more effective investing on a fast I/O rather than using more than 2gb and the usual CPU that are used on the TTX2 if you want to play mainly at fighting games but it's just a thought 
ZERO_MILES
Enthusiast
Oh yes I understand, I am doing these HW upgrades regardless of USFIV, the parts are very very very cheap.
Well "it depends" but realistically not. JVS cards all have differing latency depending on who made them and what revision it is. Some X2 games also have a 15ms polling loop that won't continue until the full 15ms is up to compensate for slow cards. FastIO is a direct memory access controller so the latency is almost 0.
Unless you already have FastIO DMACs you'll have to buy Type X2s that already have them as they're rarely sold pulled from a system and cost more than a TTX2 itself. If you don't have a Jammafier or FastIO wired cabinets you'll have to rewire your cabinets for FastIO as well which will leave you unable to use other systems than the TTX2.
Not really gonna be worth it in the long run if you don't have the hardware and a cabinet wired for FastIO.
The fast I/O that I've recently bought in the forum along with the pcie has already the jamma connector so you don't necessarily have to use a jammafier or a dedicated cabinet
Good point, I forgot that Vewlix cabinets can come with jamma connectors.
I see this kind of comment a lot when discussions about input lag come up. And I think you should never sell yourself short. You might not spend your days practicing 1 frame links, but these games that we put on these cabs are meant to be responsive. You shouldn't ever have to wonder if you missed something because you were too slow, or because the game didn't get your input in time.
We might as well just emulate once we stop worrying about those details, in my mind at least. I know for me avoiding lag is an essential part of my setups.
issue with that is, well jvs traffic is slow the way taito do it,
So the "emulators" should be fast if not faster then jvs, just due to how slow the winapi for them is
fast io on the other hand the emulators will be slower as it is direct memory access
And if we want to talk about lag, the 2nd source of lag all of these games have is forced vsync and vsysc on computers is lag,
Guess that means your going to want to take a fast io card, run it in a normal pc so you have no emulation of the fast io,
and the games run as native windows games, you then will need to edit every game, or use a dxhook to remove this vsync
once thats done you would have a no emulation, fastest inputs possible. with a bit of screen tearing due to lack of vsync
It's funny really, how arcade used to be the best, and home ports were always substandard in some way.
Now you're honestly better off ditching the arcade stuff and just rolling with a Windows PC and Steam, and Brook adapters for most games. The home ports are better, the setup is more responsive, the games are braid dead easy to get and install.
There's no real upside that I can think of aside from collector enjoyment to running a game like USFIV on an X2 box. This is a site for collector enjoyment, don't get me wrong, have fun with the authentic stuff. Just how things go now that arcade gear is all embedded Windows PCs I guess.
Now you're honestly better off ditching the arcade stuff and just rolling with a Windows PC and Steam, and Brook adapters for most games. The home ports are better, the setup is more responsive, the games are braid dead easy to get and install.
There's no real upside that I can think of aside from collector enjoyment to running a game like USFIV on an X2 box. This is a site for collector enjoyment, don't get me wrong, have fun with the authentic stuff. Just how things go now that arcade gear is all embedded Windows PCs I guess.
Tbh, I really enjoy playing USFIV on a viewlix cab. It's super comfy as compared to my lap or the floor or whatever. I only get to do so when I go to Japan though haha.
Oh, playing on a cab is awesome, I'm all about playing on cabs. I'm just saying, stick a PC in that Vewlix (or a PS4 whatever) and run the home version. Set it up right and it will have less lag than the arcade in Japan.
Not as authentic. But honestly a better experience. That's all I was really remarking on.
That's not exactly true because fastio is still more responsive than usb.... but not that it really matters as X4 has moved back to a USB based IO system and doesn't use fastio.
Do we have numbers for fastio? I know a Brook UFB over USB has been tested under 4ms, so it's probably academic, just curious
streetfighterken
Grand Master
Seems like if you're looking for an authentic standard, wouldn't it be consoles that tournaments run? Even if fast i/o is technically faster?
Roughly form my understanding
Real jvs by the time it is 1.5 frames per read so around 1.8 MS
according to brizzo with some jvs testing,
The jvs emu on the otherhand Clocks in instant we are talking 0.1MS so nano seconds but, also has the call to windows to check for controls, so depending on your device, But my keyboard clocks in about 1.3 MS
Fast io could in thoery read well fast as it wants over Pcie, its got alot of lanes to go thought
so if i had to guess based on feel cuz I dont own one to test, id say about 1MS or so BUTT
the sad reality, is Something prob no one knows besides me and niko n coery is alot of these games run internal fast io to jvs emulators, so effectually the games are still limited to about jvs performance.
then another handful of games do dumb shit like memorymap of file controls, so reading that adds so much overread very few titles even attempt to just read flat out over the pcie bus
I swear most people who go on about how great Fast IO is think it's faster just because it has "fast" in the name.
Here is a high level overview on how Fast IO works.
Part 1
Game pushes a read request to the DMAC driver.
DMAC polls the IO board for inputs
DMAC pushes a D/QWORD to memory.
Poll rate from game is about 1000hz. It was a very long time ago i looked at its poll rate so i could be wrong with this
Part 2
Game reads memory region used in part 1
Part 1 and 2 are completely independent of each other.
so yes the game can read the inputs super fast as it is a simple memory read and binary decode but where the latency comes from is the poll to update of part 1.
I have not timed this so i can't answer but it is in the magnitude of 100x fast than JVS.
Part 1 has the benefit of offloading most of the polling process to a co processor on the DMAC card.
----
When i drafted a dirty implementation of doing a Fast IO to JVS the poll response time for the JVS get buttons command is near instant as it is a pull from memory decode and respond
Where as JVS incorporates the polling as part of the read input. this would in theory make JVS more accurate but fast IO negate this by its speed.
---
to summarize if i was to choose the best build for playing all TTX2 games nxl and normal. i would choose a fast IO and convert it to JVS via software.
This also opens up to making customer action button combos for EG exiting any game
Part 1
Game pushes a read request to the DMAC driver.
DMAC polls the IO board for inputs
DMAC pushes a D/QWORD to memory.
Poll rate from game is about 1000hz. It was a very long time ago i looked at its poll rate so i could be wrong with this
Part 2
Game reads memory region used in part 1
Part 1 and 2 are completely independent of each other.
so yes the game can read the inputs super fast as it is a simple memory read and binary decode but where the latency comes from is the poll to update of part 1.
I have not timed this so i can't answer but it is in the magnitude of 100x fast than JVS.
Part 1 has the benefit of offloading most of the polling process to a co processor on the DMAC card.
----
When i drafted a dirty implementation of doing a Fast IO to JVS the poll response time for the JVS get buttons command is near instant as it is a pull from memory decode and respond
Where as JVS incorporates the polling as part of the read input. this would in theory make JVS more accurate but fast IO negate this by its speed.
---
to summarize if i was to choose the best build for playing all TTX2 games nxl and normal. i would choose a fast IO and convert it to JVS via software.
This also opens up to making customer action button combos for EG exiting any game
Last edited:
