GT5's Game-Breaking Online Flaw (OP Updated: 11 Feb)

[nasanu]

Ah, so you remember Kenji? We must have been posting on the same sites for a long long time...

Yup Kenji and his green NSX. Plus names like spook, Thoasiii, Dudley etc.

Say I do remember one poster named 'guynamedjohn', your jon... Nah too much of a long shot.

 

[jon-]

I think I posted as "Rabid" back in the day.

It was a long time ago though!

 

[NjLowrider]

Yes i have noticed this in larger rooms. I have a g27 and there is a definite stutter effect when im affected. It feels like boost is on high and glitchy. Corners at normal speeds seem impossible. And some are fine and think im crazy. Its real!

 

[Custer85]

You just realize you said you didn't experience any issue in the first sentence, but then just one sentence later you did?

There is definitely an issue, but it may very well be in large non fixed host lobbies (12+ players) people with low upload bandwidth hurt the performance for everyone.

The first sentence does not mean "...you didn't experience any issue..."

...and i never experienced anything like this.

"this" refering to the problem stated in this thread

In the other sentence iam refering to problems in general, though i have to admit that 2 out of 50 is actually quite average as far as lag is concerned and should not be considered a problem.


@cicua:
The lobbies had different sizes, roughly half of those races featured 10+ cars, the others were below that number. Funny thing was that one of the races that lagged only had 5 cars on track.

 

[V8-Johnboy]

Well credibility should not really need to come into it. Just think logically about what I say if and there is a logical flaw then take it up with me.

But if you want to know my background, not that I think it matters..

As far as career goes I am a web programer and an IT student (doing my degree slowly because of work and general lazyness) at the moment. As part of my degree I've had to do things like construct my own ram out of logic gates, write machine code assembly code to handle memory access and competing requests etc.. I might forget most of it after I pass my tests but my understanding of computing is far above average. Been programming for... No idea, maybe 13 years now. Nothing as serious as game physics though (I really hate math so I avoid building things like that), mostly database web driven stuff and mostly just for fun, I've never taken my career seriously, I'd rather program for fun than money.

As for the GT side of things I very highly doubt there is anyone here with the experience I do. I was racing GT 'online' back in 96 or 97. I was one of the people who made the rules now known as GTP rules (which are really GTOLR/GTF rules), two wheels on track etc. back before GTP even existed (GTP only exists as it does today because the GTF fell over under the weight of this own ego).

I base my opinions here on my knowledge of programming, logic, my extensive knowledge of online racing and I've also been in contact with people who make games for a living just to verify a few things (which I am NOT linking to, too many idiots on here). Take or leave what I say, but judge it on its merits.

This will do for me the bug has just destroyed our V8 season 2 finale Some odd stuff indeed no speed for me for 4 laps then it felt like i was awarded the rocket from Mario kart if not familiar felt like id gained 200bhp Can anyone offer any hope or are we doomed.

 

[griffinz]

This will do for me the bug has just destroyed our V8 season 2 finale Some odd stuff indeed no speed for me for 4 laps then it felt like i was awarded the rocket from Mario kart if not familiar felt like id gained 200bhp Can anyone offer any help or are we doomed.

My suggestion would be to try fixed host mode, and have the host with the highest upload and best quality internet be the host. At a minimum this would require at least a 2Megabit upload to the internet (check at speedtest.net to a server across the country to get 'real' speed).

I am going to do more network logging tonight, more real data will follow.

 

[Sail IC]

Don't worry, we're on it.
I was in here earlier already and still keeping my eye on it.

BTW, all mods and admins are here voluntary.

Really appreciate the work you guys put in.

I think many like I think this is the most important issue we had so far with GT5, by far. I rather have version 1.0 and this fixed than verision 2.2 (current) with this issue.

 

[V8-Johnboy]

Really appreciate the work you guys put in.

I think many like I think this is the most important issue we had so far with GT5, by far. I rather have version 1.0 and this fixed than verision 2.2 (current) with this issue.

Correct anything but this garbage.

 

[nasanu]

My suggestion would be to try fixed host mode, and have the host with the highest upload and best quality internet be the host. At a minimum this would require at least a 2Megabit upload to the internet (check at speedtest.net to a server across the country to get 'real' speed).

I am going to do more network logging tonight, more real data will follow.

Make sure to note the quality setting you have the room running on. That seems to determine how much bandwidth to use or in game terms how many times per second to update the position of other cars. Personally I've never been in a 5 star room that worked with more than a handful of people, and I wont enter 1 star rooms because they are horrible.

 

[Highlandor]

This will do for me the bug has just destroyed our V8 season 2 finale Some odd stuff indeed no speed for me for 4 laps then it felt like i was awarded the rocket from Mario kart if not familiar felt like id gained 200bhp Can anyone offer any hope or are we doomed.

Exactly the same situation as me - I've been running my series for a year now, we still use the SAME regulations (530pp sports soft tyres - road cars only) in one of our races.

We know the lap times and capabilities of our cars because we've been running them a year now.

Exactly the same problem, some guys are going 8-10 secs a lap slower than what they should, somewhere around mid race distance you can feel the car change - hey presto, you're running your "normal" lap times.

No changes to the car whatsoever, nor lobby / race regulations - most of the guys we race with are the same, same umbers in lobby - but this "thing" has hit us out of no-where.

We did a couple of tests tonight:

1) ran cars that had been tuned / modified, but used default settings

2) turned off fuel / tyre wear

On both occaisions we wre "back to normal".

I don't think these are the answers though, even though they did "work" and by making these changes the cars did perform how they should for everyone - I think this has got far deeper roots (no evidence - just a hunch)..

Reason being - Battlefield 3, this is plagued with problems that NEVER occured on the previous game (BC2).

There's no smoke without fire, both such big name titles experiencing so many issues - no way, there's something so not right in the Sony / PS3 / PSN household right now.

If there's any sign of this recurring with the PS4, that's it - off to PC sims for me....and I've been a huge PS / GT fanboy for years. Gutted.

 

[V8-Johnboy]

^ Feel your pain sir.

 

[CSLACR]

As for the GT side of things I very highly doubt there is anyone here with the experience I do. I was racing GT 'online' back in 96 or 97.

How did you manage that?

I've been playing since 98, when the first GT title was released.(except a month of 97 in Japan)
I didn't help "make OLR", I just had the common sense to not drive through grass. I'm still unsure what any of this has to do with the current topic at all either.

 

[dudeondacouch]

How on earth did this thread turn into a throwback to the original GT forums?

Makes me miss the old days of weekly comps. painterr, English Pete... where are you guys?

 

[ADEuk]

How did you manage that?

I used to play GT4 online using XlinkKai. It makes a PS2 think the internet is LAN for multiplayer gaming

 

[Speedy6543]

ADE, do you see these problems in your fully booked shuffle races too?

 

[ADEuk]

No, shuffle races seem to be OK for the most part , but when we switch to PP racing and people use their own cars the lag appears much more noticeable

 

[nasanu]

How did you manage that?

I've been playing since 98, when the first GT title was released.(except a month of 97 in Japan)
I didn't help "make OLR", I just had the common sense to not drive through grass. I'm still unsure what any of this has to do with the current topic at all either.

Online time trial comps where users submit their times to message boards all on the honer system. The fist one for GT was 'V-turns challenge of the week'. Might even have been the first online racing series for consoles. But back on topic..

 

[QB1o]

No, shuffle races seem to be OK for the most part , but when we switch to PP racing and people use their own cars the lag appears much more noticeable

maybe something to do with the power limiter then. hope they fix this soon!

 

[Ealirendur]

How many were in those rooms?

I have not had the problem either... until yesterday in above mentioned RX-7 TC race. 14 identical cars on Grand Valley. Got 1.54 for what felt like a 1.47, the game alternated quickly between slow motion and normal motion, it was like the whole thing was wobbling back and forth. To boot I also got the speed bug; passed the line at 285 km/h where normal was 270. Just horrible.

This was in a room with the usual settings, I wonder if changing any of them could lessen the bug, such as:
Grip reduction = Low
Slipstream = Strong
Tyre/fuel = Off

Anyone tried this?

Not seen it tried yet, but I mentioned some time back that Slipstream = Strong might actually measurably worsen the problem (increased detection zone for it further muddying the interaction between physics and netcode being the reasoning).

Finding out if it (like other controllable settings, especially room quality) has any reliable, measurable effect would be useful either way, obviously. I will try to pinch a managed switch for packet sniffing sooner rather than later (the fan noise will be popular), but gathering testers and results could take a while.

Turning tuning off is interesting (or shuffle, or recommended garage, as those posters noted); I can't easily see a basis myself, but I couldn't rule it out completely just based on that. Anybody experienced it with Karts?

Stuff like draft is calculated on your local console and has nothing at all to do with the network.

Citation needed, as that calculation could easily be dependent on another system communicating at least position/accel vectors (it could even want quaternions or something on that order). Sorry, but it can't be ruled out unless:

• It's tested properly
• PD comments
• Someone disassembles and makes sense of the code

(Yes, those last two are obviously fairly unlikely).

Why rule things out completely without testing, in such a complex system? Just test them in order of their perceived likelihood *to you*. Fair?

Griffinz, esh, keep up the good work!

 

[nasanu]

Citation needed, as that calculation could easily be dependent on another system communicating at least position/accel vectors (it could even want quaternions or something on that order). Sorry, but it can't be ruled out unless:

If you require a citation for everything then you may as well give up now because without input from someone from PD we can not confirm anything. What we can do look at the game and infer.

Ever been hit by someone on your screen but they say they never hit you on theirs? Ever seen a car lagging really badly, hit a wall and rejoin the track losing no speed like it never went off? We all have. This is caused by the game using prediction.

The game receives info about driver 2, he is going 200kmph in a straight line. The code will keep him going straight at 200 until the next chunk of data arrives. But what if he braked in that time? The code will place him back where he should be causing the car to jump on the screen. If he happened to be right behind you and you braked also, the code would run him right into you, then place him behind you again as if nothing had happened. Except one vital point, he hit you. This means that the local model is what you interact with, not the data that is coming in because he never actually hit you on his screen and the data he sent does not say he did.

 

[Ealirendur]

If you require a citation for everything then you may as well give up now because without input from someone from PD we can not confirm anything. What we can do look at the game and infer.

Ever been hit by someone on your screen but they say they never hit you on theirs? Ever seen a car lagging really badly, hit a wall and rejoin the track losing no speed like it never went off? We all have. This is caused by the game using prediction.

The game receives info about driver 2, he is going 200kmph in a straight line. The code will keep him going straight at 200 until the next chunk of data arrives. But what if he braked in that time? The code will place him back where he should be causing the car to jump on the screen. If he happened to be right behind you and you braked also, the code would run him right into you, then place him behind you again as if nothing had happened. Except one vital point, he hit you. This means that the local model is what you interact with, not the data that is coming in because he never actually hit you on his screen and the data he sent does not say he did.

I get that; it's what I meant by interpolation elsewhere. I've got a replay that shows exactly this sort of thing for a lagging driver from a race at the weekend, and quite excitingly too. His replay probably shows something else.

I'm not arguing that GT5 will not calculate the draft effects itself locally, based on its current model of the other car. That doesn't mean it doesn't try to get the "correct" data before guessing, or partially block on the I/O, or something. Juggling all those pipelines must get complicated when you're waiting on something so slow and unreliable.

Point is, it might be nothing, it might be down the list, but that doesn't mean we can't try it - it was mentioned because of the free run/race difference, which is perfectly reasonable to wonder about.

 

[CSLACR]

I used to play GT4 online using XlinkKai. It makes a PS2 think the internet is LAN for multiplayer gaming

Online time trial comps where users submit their times to message boards all on the honer system. The fist one for GT was 'V-turns challenge of the week'. Might even have been the first online racing series for consoles. But back on topic..

The point is GT didn't exist at the time you claimed to have been playing it.
Reading the whole post helps, instead of just the first sentence.💡

 

[nasanu]

I'm not arguing that GT5 will not calculate the draft effects itself locally, based on its current model of the other car. That doesn't mean it doesn't try to get the "correct" data before guessing, or partially block on the I/O, or something. Juggling all those pipelines must get complicated when you're waiting on something so slow and unreliable.

But if it can get 'correct' data then the model would be updated with it... At some point you need to consider that the people making the game have some brains. What you are suggesting is that they are not only producing duplicate data (programing 101), but they are purposely letting the local model get out of date (producing lag) when they could easily update those variables when they get the new ones to calculate the drag. Is there really a person in existence smart enough to be a games programer but also so stupid as to do that?

But I think what you are really saying is just test everything. We should leave the discussion of 'why' behind because I don't think we can solve it.

BTW have we even established that is does effect the physics? I know what I experienced did not, have others had this happen without a frame rate drop? Has the obvious answer of delayed input been ruled out yet?

The point is GT didn't exist at the time you claimed to have been playing it.
Reading the whole post helps, instead of just the first sentence.💡

Oh who cares. Its well over a decade ago. The tune of the post should have told you that I was not exactly sure about the dates.

 

[Divecable]

Fallen victim to the joys of an occasional handicap flicker.
Notice it straight off the start where a well tuned car including trans setup is overtaken by 5 or 6 before the first corner.
Issue is new to me .Just as new is a px21(headset)
I have extended the usb output with an extension usb into a hub for the headset,keyboard and the wheel. .Going to try removing devices next time im flickering.
Anybody else with a similar setup or device having the same issues ?

 

[Ealirendur]

But if it can get 'correct' data then the model would be updated with it... At some point you need to consider that the people making the game have some brains...yada yada... Is there really a person in existence smart enough to be a games programer but also so stupid as to do that?

The model has to be updated from the incoming data. I think we can agree that the way it has to do this is not entirely uncomplicated; consider the effect of a car effectively teleporting 10 meters on draft and collision detection. Even if there's an event/interrupt driven thread to feed the network data into the physics model, it could get hairy. Consider the way NTP works when it notices your time is way out vs small changes - it uses small incremental changes (slew) rather just up and reset the clock when it doesn't have to, as it's less disruptive. Engineering/programming can also be about choosing how gracefully (or not) you fail.

Anyways....

But I think what you are really saying is just test everything. We should leave the discussion of 'why' behind because I don't think we can solve it.

BTW have we even established that is does effect the physics? I know what I experienced did not, have others had this happen without a frame rate drop? Has the obvious answer of delayed input been ruled out yet?

I meant test the stuff we actually have control over, in order of perceived relevance. If it doesn't work, test something else. The order we decide to test may be different, but that's fine, as long as we're testing stuff one change at a time, and can usefully compare them.

Leaving the why behind is fine by me.

 

[cicua]

Finding out if it (like other controllable settings, especially room quality) has any reliable, measurable effect would be useful either way, obviously. I will try to pinch a managed switch for packet sniffing sooner rather than later (the fan noise will be popular), but gathering testers and results could take a while.

Count me in. I can dump and analyze traffic too. If you run a test we could capture on each end and compare.

Maybe we can lure people to stay and take part in testing after a real race, like the WRS-online on wednesdays. But perhaps it would be more relevant to start out with a small room and get some baseline patterns of the traffic when things work right, just to know what to expect.

I am not sure we can find anything that can solve the problem, but I can't escape being a nerd.

 

[griffinz]

Here is the data from trying 14 people in a fixed host room.

Hosted on a comcast business class cable modem, typical U/D to many speedtest.net servers across USA 12/65Megabit, latency 5ms-75ms. PS3 gigabit wired.


Track: Cape Ring Inside (tight track worst case)
Room Quality: High Voice: Standard
Aids off, except ABS 1

Grid Start with False Start check
Boost: Off
Penalty: Weak
Visible Damage: Off
Mechanical Damage: Light
Slipstream Strength: Strong
Grip Reduction: Real
PP: 575 (rx-7 tc, elise rm were primary cars).

The game is trying to use a ton of upload bandwidth, based on the usage there in a 14 player non fixed host room it would require every player to have at a minimum 2Mbit upload, and with some headroom 3-4Mbit for 14 players. I wish I could monitor PS3 CPU usage somehow because in this case I honestly think my host PS3 was maxed out and accounted for the lag during the race. I would really like to try with a second PS3 hosting but not racing, but I don't currently have access to another PS3. I think beyond that the only possibly thing to try to work around this would be to put 16 PS3's on a local network, which is pretty unrealistic.

Please Polyphony Digital put your best guys on the netcode for the next patch.

 

[Speedy6543]

^^Nice work there griffinz. 👍

 

[V8-Johnboy]

Probably a daft question but are PD actually aware as yet of the seriousness of our problem??

 

[MTLclllutch]

my upload speed is only around 500kbs for my ps3 so this isnt good news