[Online] Race Quality and effect per rating.

Discussion in 'Gran Turismo 6' started by PJTierney, Feb 4, 2015.

  1. PJTierney

    PJTierney

    Messages:
    3,288
    Location:
    Ireland
    [​IMG]

    Hi all.

    I have started hosting lobbies recently and have a query regarding the Race Quality setting.

    Here is what the GT6 Online Manual says about it:
    While that is a good explanation of the setting as a whole, it does not detail how a Room changes depending on which level (or "star rating") the Race Quality is set to.

    So, from your collective experience, can anyone give me some feedback on what happens in:
    • A "4 star Room"
    • A "3 star Room"
    • A "2 star Room"
    • A "1 star Room"
    I would imagine that a "5 star Room" has everything turned to maximum, be it frequency with which data is transmitted between drivers, particle/smoke effects, graphical rendering capabilities, frame rate etc.

    I would also imagine that down the levels, each of these features is reduced or removed entirely.

    If anyone has any input or experience on racing in different qualities of Room and can share some thoughts, that would be much appreciated :)
     
  2. VBR

    VBR Premium

    Messages:
    6,866
    Location:
    United Kingdom

    That setting only affects the amount/quality of telemetry that is sent between players, nothing else is affected.


    :tup:
     
    Chrunch Houston and PJTierney like this.
  3. Chrunch Houston

    Chrunch Houston Premium

    Messages:
    2,237
    Location:
    United States
    @VBR is exactly right.

    I have a friend in the Philippines, he was getting disconnected, so he asked if I would lower the race quality. I now run a four star room all of the time, and really don't notice a difference.

    Even with a three star room, I really didn't notice anything, but I only tried that once.
     
    VBR likes this.
  4. bulls23

    bulls23

    Messages:
    19
    So the actual refresh rate / accuracy of other car's positions isn't affected by that setting? When will I notice the difference? Maybe in stuff like the replays of an online race?
     
  5. VBR

    VBR Premium

    Messages:
    6,866
    Location:
    United Kingdom
    No, it is affected by that setting! The telemetry, or "Driving Information" as PD calls it, includes everything needed to recreate your opponents driving performance on your PS3.


    You'll notice the difference if you are racing against someone who has a recommended 2 star connection, & your room is set to a 4 star connection. This might result in lag, or a disconnection for the player with the lower bandwidth.
     
    bulls23 likes this.
  6. PJTierney

    PJTierney

    Messages:
    3,288
    Location:
    Ireland
    Would a lower setting increase the chance of the "Invisible Cars" issue occurring?
     
  7. Sick Cylinder

    Sick Cylinder

    Messages:
    4,817
    I don't know the answer to that one.

    Before a race we check the number of cars visible e.g. Mr Smith types 9 if he sees 9 cars. The other drivers type 9 and if the Host agrees there are 9 we start.

    If cars are invisible to the host they are listed in the room, but don't go on track so ask all drivers to go on track. If people can't see the correct number they should leave and rejoin the room. If several people have this problem the Host should start a new room.

    To get the most stable connection I reduce the room to four stars even if it recommends 5, disable voice chat as this takes up bandwidth, use a public room - avoid private rooms and club rooms.
     
    VBR and PJTierney like this.
  8. PJTierney

    PJTierney

    Messages:
    3,288
    Location:
    Ireland
    Why avoid Private/Club rooms, does the filter do something that affects connection quality?
     
  9. Sick Cylinder

    Sick Cylinder

    Messages:
    4,817
    I used to use Private Rooms on GT5, then with GT6 I was having a lot of problems with disconnections including sometimes multiple Host disconnections - I took advice from Furinkazen and switched to public rooms and found that the disconnection problem which had plagued us was virtually solved.

    I believe that Club rooms are a variation on a private room so I avoid them.

    If you use public rooms you will get unwanted visitors even if you label the room PRIVATE, but you can use the kick function.
     
  10. VBR

    VBR Premium

    Messages:
    6,866
    Location:
    United Kingdom
    No idea. I remember that turning the music off in GT5 solved most issues that people had online, that & using a wired connection. But some still have trouble after doing all that.
     
    Goodwolf likes this.
  11. watermelon punch

    watermelon punch

    Messages:
    6,377
    Location:
    United States
    I thought lowering the star setting would decrease the chance of Invisible Cars.

    I thought that was totally about the fact that in the old GT5 friends lounge, you couldn't set it to fixed host. :confused:
    (Fixed host being the cure for a lot of connection problems.)

    Now in the club lobby, you can set it to fixed host.
     
  12. Sick Cylinder

    Sick Cylinder

    Messages:
    4,817
    When I tried fixed host in GT5 it had the disadvantage that it selected a lower race quality. Also (I might be wrong about this) I think it meant that if the Host was disconnected the race meeting was affected. I run a room with deputy Hosts so that if I get disconnected one of them can run the meeting as normal in my absence.

    I think I'll need to experiment with this again to see how its changed.
     
    VBR likes this.
  13. PJTierney

    PJTierney

    Messages:
    3,288
    Location:
    Ireland
    If a Fixed Host disconnects the lobby shuts down.

    I've noticed that when I create a room it defaults to a 4 Star Connection, but when I change it to Fixed Host it drops to a 2 Star rating. If I then reduce the number of possible players that can enter, it gradually increases to 3 and 4 Star ratings.
     
    VBR likes this.
  14. Griffith500

    Griffith500

    Messages:
    8,536
    Fixed host is the normal client - server approach. You, as the host, receive information from all other connected players, then resend all that received data to all other players. Other players only send their own data to the host.

    The other mode is peer to peer, where all connected players send their own data only to all other players.

    So in a sixteen player fixed host room, the host receives 15 lots of one-car data, and sends 15 lots of 15-car data. The other players only send one lot of one-car data each, and receive one lot of 15-car data.

    In a sixteen player peer to peer room, all players send 15 lots of one-car data each, and receive 15 lots of one-car data. So the host is really an arbitrary distinction, hence the ease of swapping.

    Fixed:
    Up: 225; Down: 15 (host)
    Up: 1; Down: 15 (client)

    P2P:
    Up: 15; Down: 15 (all)

    So in the fixed host setup, you need markedly more upload capacity. Most connections are asymmetric, having much more download capacity.
    In practice, the P2P mode will have extra data flows for checks etc.


    The other issue is that if the fixed host connection is slow, everyone's game is laggy.

    But in the peer to peer situation, slow connections affect only the individual player - except for the teleporting everyone else has to witness, and the inevitable bumping. And those extra checks.

    EDIT: I have a feeling that's inaccurate, now I've written it. I know @Famine did lots of tests with GT5, and the reality he found might differ from the "theory" above.
     
    Last edited: Feb 7, 2015
    Sick Cylinder, PJTierney and IfAndOr like this.
  15. PJTierney

    PJTierney

    Messages:
    3,288
    Location:
    Ireland
    That's a good explanation @Griffith500.

    The league I'm currently running has an average of 11 drivers showing up each week, so far we have used P2P.

    Week 1 had no issues.
    Week 2 had 2 invisible cars initially but a new room eliminated that issue.
    Week 3 had 5 invisible cars initially and a new room didn't make a difference (same 5 drivers affected).

    I am considering using Fixed Hosting for the next event, would this sort of connection quality be good enough to host a 10-12 person lobby?

    [​IMG]

    [​IMG]
     
    Last edited: Feb 7, 2015
  16. Griffith500

    Griffith500

    Messages:
    8,536
    I honestly have no idea what bandwidth is required, because I don't know how much data the game relies on for communicating position, physics inputs, lag-prediction etc. Your ping is also important, but not to your ISP, or some speed test site, rather to the individuals you're playing with.

    The easiest way is for you to try it and let us know! :)

    One thing to bear in mind is that the bandwidth for upload in the P2P case scales linearly for all players with the number of players, is constant for a client connected to a fixed host, but scales as to the square of the number of players for the fixed host itself. So a small change in number of players can have a large effect on required upload bandwidth (e.g. 12 players is 121 lots of car data to send, instead of 225 for 16 players). When you start running out of bandwidth, you get the effect of increased ping and especially jitter (variance of ping).


    The "invisible cars" thing points to some kind of synchronisation between peers in peer-to-peer mode, which may mean that vastly different bandwidth (or, more precisely, latency / lag / ping) connections do not mix well in that mode. NAT type incompatibility also had similar effects in GT5; maybe NAT adds extra latency.


    I think what @Famine found was some reliance on low latency in the P2P case, which complicates things further. Fixed host is more robust for latency; only those with distant connections to the host have problems, everyone else depends on their own connection to the host only.

    In GT5, a single distant (high-ping) connection tended to disrupt the whole P2P scheme, somehow, as all players are connected to all others, and everyone is dependent on everyone else's ping (possibly synchronisation, and those mysterious "checks" I suspect exist).

    Remember the bandwidth setting ("race quality") in the game sets the frequency of data updates. If synchronisation is an issue, due to latency, then going above a certain frequency (related to round trip speed) might break things for that one player.


    EDIT: If everyone is playing on a LAN together, use the fixed host setting. The total bandwidth required is contained on the same bit of ethernet, rather than distributed over lots of hardware in the wider web, and P2P is a monster in that regard. A 16 player game requires 15 times the total bandwidth in P2P mode than it does in fixed host mode - it is effectively cubic instead of quadratic.

    In that regard, anyone who can play online happily in a 10-16 player room in P2P mode can host a minimum of 4 players (including themselves) happily in fixed host mode, at the same quality*. Note that your ping is still important.

    * If you can play in a P2P room with n total players happily, then the minimum number of players, including yourself, that you can host in fixed host mode is √(n-1) + 1. E.g. for n = 16 -> √15 + 1 = 4.87... ; for n = 10 -> √9 + 1 = 4. The reduction in total number of connections (e.g. 16 to 5) reduces packet overhead, and you can squeeze a bit more out of your upload.
     
    Last edited: Feb 7, 2015
  17. Sick Cylinder

    Sick Cylinder

    Messages:
    4,817
    Thanks for extra detail Griffith - I'm not a very tech savvy person so I can't say I fully understood that. Is this the right interpretation? That PJTierney would be better not using the Fixed Host unless he has a high upload speed?

    I have a download of about 5.5meg and an upload of about 0.8 meg and I find a public room runs ok with up to 16 players, but Fixed Host doesn't work well - presumably I don't have enough uplaod speed for that option?
     
  18. Griffith500

    Griffith500

    Messages:
    8,536
    Yes, exactly right. Additionally, there is no point in using P2P on a LAN, even with up to 4 or 5 "external" players.

    The race quality sets the update frequency, and the connection / host type determines the way the bandwidth requirements are distributed amongst the different players.

    The total bandwidth required (per player) is a function of the per-update data size and the update rate. It's relatively easy to convert between the per-update data sizes of each type of connection.

    For n players, per-update bandwidth required is proportional to:
    Code:
    Type     Up    Down
    ----------------------
    Host     n²    n
    Client   1     n
    P2P      n     n
    Ignoring overheads, which will be proportionally constant for data.


    It's harder to figure out what update rate would be equivalent in fixed host mode to the P2P mode, in terms of bandwidth, for the same number of players without knowing overheads (non-data communications between players), so you'd have to do that via trial and error.

    You can probably host a larger grid, but only at very slow update rates ("race quality" settings), which is somewhat obvious. We may be looking at least as low as a quarter of the rate - it's hard to know whether the rate 1 star represents is a quarter of what 4 stars represents. The recommended settings will be more useful there, because they'll be coded into the game based on the actual underlying data rates, and on development tests.


    Ping is still a massive spanner in the works. You may have the bandwidth to host all by yourself, but if your ping to all players is too high (e.g. because they're all on a different continent), then they will have a poor experience, because they can only see each other through a round trip (there and back) communication with the host. From your end, the problem seems only half as bad, because the delay is only half (one-way).

    It may be difficult to tell between a marginally inadequate bandwidth and high-ping; a substantially inadequate bandwidth will be obvious due to jitter (sometimes the updates come in fine, sometimes huge delays). I don't know whether low-quality (low update rate) racing looks any different from high-ping racing, either. You can't tell very much as the host, though, you need an accomplice to join as a client.


    Ultimately, the only way to be sure about any of this, is to test it.