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

[BrandonW77]

What was damage set to? Visual damage on/off?

This is what I'm curious about. Visual damage "on" would require more processing power (I'd think) which could make this problem worse. Perhaps some tests are needed with damage on and off to see if there's a difference.

 

[Speedy6543]

Visible damage was on.
For our next race wednesday, we just decided to set it off.

 

[CSLACR]

Visible damage was on.
For our next race wednesday, we just decided to set it off.

And I'll get to be part of those tests/races!

I am slowly developing a theory, need some more pieces yet though.
As we know, the tracks are loaded in online lobbies, but from where, the internet or our PS3?
The obvious answer is the PS3, BUT - Then why do people (usually with slower connections) get stuck on the loading screen sometimes?

 

[jon-]

And I'll get to be part of those tests/races!

I am slowly developing a theory, need some more pieces yet though.
As we know, the tracks are loaded in online lobbies, but from where, the internet or our PS3?
The obvious answer is the PS3, BUT - Then why do people (usually with slower connections) get stuck on the loading screen sometimes?

Because they have trouble syncing all the client data?

 

[CSLACR]

Is that a question or an answer?
I'm theorizing, I really don't know.

I've always found it odd that people can so commonly get stuck loading a track, if they're loading it from their own PS3.

 

[cicua]

And I'll get to be part of those tests/races!

I am slowly developing a theory, need some more pieces yet though.
As we know, the tracks are loaded in online lobbies, but from where, the internet or our PS3?
The obvious answer is the PS3, BUT - Then why do people (usually with slower connections) get stuck on the loading screen sometimes?

You mean at race start? (not just when joining the room)

Perhaps because the PS3's exchange some timing sync to make sure everyone gets green light at the same time. Any kind of network trouble could cause this to go wrong and hang.

At least it can not be the full race track data since that is already loaded before you can go to track.

And for all I know, every copy of GT5 has all tracks and cars on disk, Spa and stealth/chrome/DLC cars included, would be way too much data to load on-demand. But it should be easy for someone to stat the network traffic to see how much comes through before race start.

Custom tracks would be an exception.

 

[priesty_lfc]

Read quite a lot of this, but not seen this mentioned yet,

Do you think HDD size has an impact on this?

GT has a massive install, if people are close to having a full HDD could this impact the performance of when GT draws information from it?

Could relate to people who have the phat as it's a smaller HDD having more issues than those with the slim......

 

[vasho]

Read quite a lot of this, but not seen this mentioned yet,

Do you think HDD size has an impact on this?

GT has a massive install, if people are close to having a full HDD could this impact the performance of when GT draws information from it?

Could relate to people who have the phat as it's a smaller HDD having more issues than those with the slim......

But you can change HDD on the PS3. So that should not be an issue.

I must say i have experienced this bug as well.
It was on a custom toscana track. I was driving in a premium GT86 and my friend in a standard AE86.
During the free run i had no issue of keeping up with him at all or to be faster.
But as the race started i could almost not even keep up at the first straight. And what was funny as well when i overtook him, he than overtook me again but it looked like a sling shot effect. He just flew by me like he would have a ton of grip.

P.s.: Sorry for my bad english.

 

[CSLACR]

You mean at race start? (not just when joining the room)

Perhaps because the PS3's exchange some timing sync to make sure everyone gets green light at the same time. Any kind of network trouble could cause this to go wrong and hang.

At least it can not be the full race track data since that is already loaded before you can go to track.

And for all I know, every copy of GT5 has all tracks and cars on disk, Spa and stealth/chrome/DLC cars included, would be way too much data to load on-demand. But it should be easy for someone to stat the network traffic to see how much comes through before race start.

Custom tracks would be an exception.

I mean in the lobby.
Custom tracks are actually just a sequence of numbers, we all have them on the disc. (I don't mean programming numbers, I mean like section 1, width 2, etc, etc.) (of course they're programming sequence numbers too)

I just find it odd that people get hung up and freeze the lobby "loading" and I wonder if we don't actually all play on an uploaded "online track".
There's also the fact that some people can't load bigger data tracks, like Nurb 24Hr with time and weather.
If we're loading the track from our PS3, track size/detail shouldn't matter.

Whatever the case, I'm finding game data/detail has a great effect on online play, whether it's a detailed track, detailed cars, detailed damage, etc.

It doesn't make sense in any way I can think of for details like these to matter, since they don't in offline play, that tells me it's not a personal PS3 data/processing power issue, but clearly only related to online play.

The general consensus is that all this is run by everyone's PS3, but if that's the case, and each person's PS3 loads the track, cars, damage, etc all through their PS3 and we only truly send packets of steering/throttle/braking inputs through the internet, then this wouldn't be possible, no?

 

[Speedy6543]

Do you think HDD size has an impact on this?

GT has a massive install, if people are close to having a full HDD could this impact the performance of when GT draws information from it?

I installed a 250Gb disk in my fatty but there is only about 40Gb on it, I think.

 

[CSLACR]

I installed a 250Gb disk in my fatty but there is only about 40Gb on it, I think.

To my knowledge of hard drives, free space shouldn't really be a major issue in operating speed typically.

 

[BWX]

500GB 7200rpm installed with over 300GB free on slim 120GB. 15 to 25 mbits/sec down, 1mbit/sec up cable. Happened to me before Spec II though, but not all the time, just enough to kill the fun..
Haven't raced online for a while. I'll have to try some online just to experiment with this ''issue'' again, now that other people are aware of it.

 

[priesty_lfc]

But you can change HDD on the PS3. So that should not be an issue.

I know you can, doesn't mean they have tho!

I have upgraded to 320Gb but I have yet to experience this issue (although I have been in a lobby with someone who has)

I installed a 250Gb disk in my fatty but there is only about 40Gb on it, I think.

Fair enough,

Not seen it mentioned so I thought I would as it could have been a cause.

Still hope it can be pin pointed tho as it does exist.

 

[cicua]

I just find it odd that people get hung up and freeze the lobby "loading" and I wonder if we don't actually all play on an uploaded "online track".

I think the lobby hangs happen when a player can not establish network connection with all the other players. So the problem is not getting a lot of track data through, but rather no link at all.

Whatever the case, I'm finding game data/detail has a great effect on online play, whether it's a detailed track, detailed cars, detailed damage, etc.

It doesn't make sense in any way I can think of for details like these to matter, since they don't in offline play, that tells me it's not a personal PS3 data/processing power issue, but clearly only related to online play.

The general consensus is that all this is run by everyone's PS3, but if that's the case, and each person's PS3 loads the track, cars, damage, etc all through their PS3 and we only truly send packets of steering/throttle/braking inputs through the internet, then this wouldn't be possible, no?

Well maybe yes... as the thinking seems to go, the problem starts when the PS3 gets overloaded. So big tracks, premium cars, damage, maybe mics, and above all number of drivers, all increase CPU use and therefore risk.

It sure is strange why offline works. Maybe it does not? I have never checked if my offline replays against wall clock. But others have discussed online/offline usage above, I think it is hard to compare, there are so many differences seen from the game's pow.

 

[Speedy6543]

Still hope it can be pin pointed tho as it does exist.

 

[CSLACR]

I think the lobby hangs happen when a player can not establish network connection with all the other players. So the problem is not getting a lot of track data through, but rather no link at all.



Well maybe yes... as the thinking seems to go, the problem starts when the PS3 gets overloaded. So big tracks, premium cars, damage, maybe mics, and above all number of drivers, all increase CPU use and therefore risk.

It sure is strange why offline works. Maybe it does not? I have never checked if my offline replays against wall clock. But others have discussed online/offline usage above, I think it is hard to compare, there are so many differences seen from the game's pow.

Well it's not the replay itself that's the issue, and I wouldn't expect to find it, because I've never heard people asking why they sometimes run 10 seconds slower a lap when racing offline races.

Maybe this way: If it were an uploaded, shared online track that we all raced on online, and someone was experiencing issues downloading the track as they drive it, I could see all types of issues cropping up, frame rate issues, low grip issues, etc.
If it's actually a system overload and just slow PS3's having the issue, I would expect to see more evidence for that already.

I don't know, just theorizing some more.

Amar seems confident it affects high speed internet users before slower connections, which would mean it had nothing to do with the PS3 processing at all, but only be an internet issue, and I'm inclined to think that's the case.
Not the PS3 processing power, but something related to to much data being transferred online.
If it's to much data being transferred for some connections/systems to handle, what is this bulk of data causing the problem then?

That's where I expect we'll find the answer.
I think the first thing that should be tested is whether or not things like standards/premiums cause the issue, big data tracks, and/or damage, or any combination of them.
Once we know for certain if one or all of these have a play in the matter, we can narrow it down from there easier.

 

[cicua]

I think the first thing that should be tested is whether or not things like standards/premiums cause the issue, big data tracks, and/or damage, or any combination of them.
Once we know for certain if one or all of these have a play in the matter, we can narrow it down from there easier.

This. We need to be able to trigger the problem before we can figure out how to avoid it.

My 2 cents:

Nurb type V+weather/time in WRC cars (most detailed)
Damage, mics, background music on
Full grid, most importantly
Anything else can we add to tax the system and network? Use horns, wipers and headlight blinking at all times?

We run a 1 lap race, all cars stay close together so they are all in view, and we can create some damage along the way. Race pace is not necessary, just reasonably swift.

Then we all go home and stopwatch check our replays. It should give an exact answer if something was wrong for anyone (hopefully it will be). Then we can scale down and figure out when it stops.

Some say they get the problem often (I never did), I hope they will join.

Anyone up for it, wednesday around 22.00 (10PM) GMT?

 

[Ealirendur]

We should discuss how to gather some data. While denying the issue exists is stupid, our crazy Aussie is right that we need some data to isolate what factors are having an effect.
Ultimately, when this is hopefully brought to PD's attention, data gathered could be useful in arguing our point.
Say, a questionnaire on surveymonkey. Ask about perceived frequency of occurrence, some controls (like DS3 users, NAT type, bandwidth etc) and some sample controls (GTP WRS division, member of online leagues/clubs etc). Exact questions should be discussed and then, with help of some moderators, ask for the survey to be featured in the news. Statistically we'll have a major issue of people quitting GT5 because of the issue we're investigating, but in no way this makes the whole exercise worthless.

Thanks, Dan, I completely agree. I know you meant the other crazy Australian ;-)

I still can't quite believe no one has logged races through wireshark/netmon et al, but I shall try to attend to that in coming weeks. The magazine latency test was interesting.

What I was alluding to earlier was to do with trying to sort out what is going on from data that is actually available for us to test with - the replay wall clock issue is one, and comparing enough replays at 12-16 hosts certainly won't hurt.

Estimated figures for bandwidth per person for the fixed/mesh and room quality types would be useful, and should be possible. With enough data for differential analysis, you would be able to work out how GT5's code changes packet distribution, if indeed it does, in a mesh.

It is worth noting (again, as others have said) that a replay of an online race is only an approximation for any one other than the one who recorded it - jumpiness/twitchiness and freaky driving are usually indicators of lag to a given driver from the recording player.

How much another driver and network problems affect you can be agreed to depend at least at minimum on collision detection zones, and draft (both complicated by gaps in telemetry).

Rubber on track, armco movement, weather, and damage are other possibles, but the only way it can be turned into a falsifiable test is running the same participants with one factor you can actually change (for visible damage, say) with relatively stable conditions (someone's sibling/child firing up limewire, iView/bbcplayer/netflix et al partway through clearly isn't stable conditions). Record all replays, hopefully at least one party log it through wireshark, lather, rinse repeat. All change from premiums to standards, go again. Flip to strong draft, try again. Drop 1 or 2 participants, do that at least twice, and so forth. Preferably have speedtest/measurementlabs NDT results for everyone between each race, without leaving the room. Then move to a different track. Repeat until you have enough data points. Ideally, have more than one party (possibly use an observer in live timing) video each race to compare to replays (not livestreamed, Speedy! )
Then change to fixed host, and start again.

Sound time consuming?

Large data sets from qualitative survey data does sound easier. Apologies for the wall of text, didn't have time to make it shorter.

EDIT: Also, do what Cicua said.

 

[ajns73]

To my knowledge of hard drives, free space shouldn't really be a major issue in operating speed typically.

PC hard drives with modern operating systems are one thing. They do benefit from more free space for different reasons that I won't bore you with here. The question is does the PS3 OS function like Windows 7 or Mac OS? If the answer is yes, then HD performance will suffer as the disk appraoches being full.

 

[CSLACR]

PC hard drives with modern operating systems are one thing. They do benefit from more free space for different reasons that I won't bore you with here. The question is does the PS3 OS function like Windows 7 or Mac OS? If the answer is yes, then HD performance will suffer as the disk appraoches being full.

Not saying it's impossible, but I believe some people with the issue have already been said to have lots of free room on their HDD's.

This. We need to be able to trigger the problem before we can figure out how to avoid it.

My 2 cents:

Nurb type V+weather/time in WRC cars (most detailed)
Damage, mics, background music on
Full grid, most importantly
Anything else can we add to tax the system and network? Use horns, wipers and headlight blinking at all times?

We run a 1 lap race, all cars stay close together so they are all in view, and we can create some damage along the way. Race pace is not necessary, just reasonably swift.

Then we all go home and stopwatch check our replays. It should give an exact answer if something was wrong for anyone (hopefully it will be). Then we can scale down and figure out when it stops.

Some say they get the problem often (I never did), I hope they will join.

Anyone up for it, wednesday around 22.00 (10PM) GMT?

I could sign up and potentially join, but we need people that get the problem to join first and foremost.
Also my time is limited to about 45 minutes at that specific time.

Maybe let's start a sign up, the more people getting the issue the better, for Wednesday at 10:00PM GMT (5PM EST)?

 

[Speedy6543]

Working further on this post.

Good news!
As I do a livestream every week of the race, which is viewable here, you guys can all see the difference between my quali grip and race grip.
Also, take note that the GT5 in race timer, in this livestream video, is a 100% accurate. Which wasn't the case when I reviewed the replay of this race........
You can all take out your stopwatch and FFWD to 21'00.0min and see my second lap which I already timed in the replay and being 14 sec off then.

 

[ajns73]

Not saying it's impossible, but I believe some people with the issue have already been said to have lots of free room on their HDD's.

Roger that. Makes my post a waste of space and time.

Here's something that isn't hopefully. I read through the rest of the thread and has some ideas. First, it was interesting what you said about hanging at the loading screen. You'd kind of have to know what PD is "loading" at that time to have a definitive answer, but it seems more logical that the track loads locally and information about where the other players are on the track is reported and represented on that track. But, like I said, you'd need an answer from PD to know for sure.

I think this is going to end up being a network issue. Others with more network knowledge than I have raised important issues. One other thing that might be interesting to track is whether people with issues are using DSL versus "Cable" versus dedicated network lines and whether they are hardwired to their gateway or if they are trying to use a wireless connection.

 

[Ealirendur]

PC hard drives with modern operating systems are one thing. They do benefit from more free space for different reasons that I won't bore you with here. The question is does the PS3 OS function like Windows 7 or Mac OS? If the answer is yes, then HD performance will suffer as the disk appraoches being full.

It would be hard to evaluate the performance of the PS3 FS (encrypted SPUFS, apparently?) against NTFS or HFS, given that it's encrypted and proprietary. I found links under this interesting, particularly the old PDF about forensic analysis. There may well be better information out there since certain more recent, uh, revelations, but evaluating them during or in relation to online play would be fairly problematic to say the least.

 

[ajns73]

It would be hard to evaluate the perfmormance of the PS3 FS (encrypted SPUFS, apparently?) against NTFS or HPFS, given that it's encrypted and proprietary. I found links under this interesting, particularly the old PDF about forensic analysis. There may well be better information out there since certain more recent, uh, revelations, but evaluating them during or in relation to online play would be fairly problematic to say the least.

Good point about the encryption. I had completely forgotten about that. It seems that this angle has been somewhat eliminated since people who are not approaching full disks are having the same issues. However, until it is completely disproved, you have to consider it since qualitative analysis from players indicate they felt that the PS3 was "working hard." Unfortunately, I don't play online. One thing people might pay attention to is r/w to/from the disk when they are having the issue.

EDIT: Just checked out the link. I am assuming these are CSI techs working on real evidence for real cases. The question that I has was: "If these guys can't do it, nor can they compel Sony to do it, what hope would we ever have?"

 

[CSLACR]

Working further on this post.

Good news!
As I do a livestream every week of the race, which is viewable here, you guys can all see the difference between my quali grip and race grip.
Also, take note that the GT5 in race timer, in this livestream video, is a 100% accurate. Which wasn't the case when I reviewed the replay of this race........
You can all take out your stopwatch and FFWD to 21'00.0min and see my second lap which I already timed in the replay and being 14 sec off then.

Unfortunately my PC is to sucky to watch it right.
Videos don't play right, freezing, etc for minutes at a time.

 

[Ealirendur]

EDIT: Just checked out the link. I am assuming these are CSI techs working on real evidence for real cases. The question that I has was: "If these guys can't do it, nor can they compel Sony to do it, what hope would we ever have?"

On the bright side, the information there is quite old, and I didn't use much google-fu yet. But working with actually (at least mostly) controllable variables and data seems much, much more reasonable.

 

[Jubby]

I wonder if this flaw would have anything in common with a timed race ending issue. When I was doing my 24h race at Nurburgring, the time remaining versus the actual time completed was off by over 2 minutes. I didn't worry about it at the time (it made no difference in my race). I searched for others, as I remember seeing a photo of the same idea, and came up with 2 others. One in B-spec, and one not specified(my guess is A-spec).
First
Second
I'd imagine we wouldn't notice it as much outside of long races that are times, or online racing where seconds really count.

This also raises a question of... tests and special events. I wonder if these could be effected, if this flaw reaches beyond online.

 

[CSLACR]

That is very interesting. I'm going to time a 4 hour race tonight. Of course, since I'm usually not affected I expect mine might be accurate.

 

[r2dc2]

It is worth noting (again, as others have said) that a replay of an online race is only an approximation for any one other than the one who recorded it - jumpiness/twitchiness and freaky driving are usually indicators of lag to a given driver from the recording player.

Here's another odd thing I noticed. I have a recorded replay of a clean, unbugged 5 lap Spa race with 12 cars on the grid. The first 2 times I viewed the replay, some cars midfield were lagging, jumping, twitching and freaking out quite obviously on the first lap. By the third time I viewed the replay in its entirety, the laggy twitchy freakiness in the midfield cars were totally gone! Same replay, all the parameters were unchanged, but no more twitchy cars. To my layman's point of view, it seems like the more often you watch a stored replay, the smoother the PS3 is able to make it by interpolating the missing bits, filling in the blanks and saving that onto the already existing replay data. This replay was recorded almost 4 months ago. But now if I were to load it and watch it again, there are no more twitchy cars right from the get go.

It was a reasonably close 5 lap race, so I have no reason to suspect that the 'phantom lag' was present in that race. But why did my replay 'smooth out' with repeated viewings?

I can deal with the whole online/offline being different. I can deal with that fact that because of that, the online and offline laptimes will never match. I don't care.

What really riles me is that in online racing, some of us are going like our cars have 3 wheels while the others have 4. There is no consistency amongst cars/timing online. How can we race like this?

 

[BestCaliPlayer]

It's definitely a connection issue....but it doesn't relate to the PS3 hardware.