Car handling different between offline and online

Discussion in 'Technical & Support' started by Marc Brink, Apr 29, 2019.

  1. stonec

    stonec Registered

    Joined:
    Jun 19, 2012
    Messages:
    3,399
    Likes Received:
    1,488
    If you want you can PM me with the files and I'll have a look. 1.5 seconds slower online on the same realroad doesn't make any sense to me, during all the time I drove in various leagues the difference observed was within margin of driving errors, i.e. tenths at most.
     
    bwana and Ari Antero like this.
  2. lordpantsington

    lordpantsington Registered

    Joined:
    Oct 5, 2010
    Messages:
    849
    Likes Received:
    79
    Confirmation Bias.
     
  3. Lazza

    Lazza Registered

    Joined:
    Oct 5, 2010
    Messages:
    12,346
    Likes Received:
    6,572
    If you are logging using a recent version of DAMPlugin there's a realtime loss channel which would indicate any physics timing issues. Although by nature it's a different beast to rF1 (as explained earlier) it might help highlight an issue.

    That's why the whole issue is muddy. Everyone I personally heard talk about this felt (in rF1) that they were faster offline. Online they felt like they had less grip, which I took to mean they had less feel or connection to the car and it was harder to find the limit.

    I had a similar feeling online in rF1, but my pc was barely able to run the game at that point. Online with other cars was pushing it and no doubt causing some sync issues either inside or outside the game itself.

    On the other hand to all this - I heard at least one person complain about changing grip levels from lap to lap, in rF1. I don't believe that was happening, but they were adamant. ("I did exactly the same as last lap, but this time I spun") Given how dynamic rF1 isn't, I think the human factor shouldn't be underestimated. And like setup changes, it's easy to convince yourself x is leading to y.
     
  4. stonec

    stonec Registered

    Joined:
    Jun 19, 2012
    Messages:
    3,399
    Likes Received:
    1,488
    Yep, same experience for me. I remember back in around 2007 when rF1 was new, it took a state of the art Nvidia 8800 GTX card to run high-resolution mods with large grids at good FPS. What happened in online races for me was that FPS often plummeted, which was the major reason for the "disconnect".

    Few year later when rF1 matured and the hardware caught up, I could quite often drive personal best laps in the official league qualifying sessions, which wouldn't be possible if there was any slowdown effect from online, at least not at the highest level of sim racing where drivers are within tenths from each other.
     
  5. Comante

    Comante Registered

    Joined:
    Nov 20, 2013
    Messages:
    1,694
    Likes Received:
    1,218
    I'm consistently slower both online and during races than practicing offline . My conclusions on the issue is that the lack of performance is psychological (In my case) I'm not a race pilot, and stress level and lack of focus and increased muscular tension all play against fast lapping.
     
    Eddy likes this.
  6. Comante

    Comante Registered

    Joined:
    Nov 20, 2013
    Messages:
    1,694
    Likes Received:
    1,218
    I've noticed that (as a personal level of course) more grip does not always mean faster laps. Yes, from green track up to a point, you go from ice skating to decent driving to "wooohhhaaaaahhh" moment, then performance start to deteriorate again.
    Btw I'm curious about the meaning of " same realroad condition online/offline. Are we talking about settings or about realroad saved state?
    Maybe the saving method of the realroad condition is not 100% accurate when saving from online server.
     
  7. mantasisg

    mantasisg Registered

    Joined:
    Aug 17, 2016
    Messages:
    2,926
    Likes Received:
    3,876
    Those few times that strange feel occurred to me it never felt anything related with grip, but it felt related with car responsiveness.
     
  8. Lazza

    Lazza Registered

    Joined:
    Oct 5, 2010
    Messages:
    12,346
    Likes Received:
    6,572
    Fair call, I don't think I've tried a direct comparison but I know loading the autosaved realroad on a server didn't match how it was earlier. That was a few years ago.

    However, I have done back to backs with a server and locally both loading the same rubber file, and it seemed fine. That's obviously probably not the way most users would do it, so it's a fair point to consider.
     
  9. Stefan_L_01

    Stefan_L_01 Registered

    Joined:
    Nov 18, 2012
    Messages:
    595
    Likes Received:
    386
    I said ambient, not surface temp. Ambient is air temp. There is a dominant cooling effect happening, if you read 70deg in the HUD you may have 110 for the rubber (inner layer) allready.
    If you set 15° air temp in the session settings you see 15° surface temp also but as you said this does not matter. If a weather plugin sets a weather to 15° where the session initially sets it to 29°C, surface temp shows 29°C but again this should not make a difference
     
  10. Stefan_L_01

    Stefan_L_01 Registered

    Joined:
    Nov 18, 2012
    Messages:
    595
    Likes Received:
    386
    Back to the topic: I´m still thinking whether it is lack of grip or a slowdown of physics engine. Lack of grip usually makes you oversteer, this is not allways the case.
    However, in case the physics engine is triggered with 400hz, and if you run straight it would make you advance by 1/400 sec according to your speed, you could think about slowdowns when the REAL rate is 399.8hz and others will go faster with true 400.1 hz. . A loss of one frame here and there would be enough to give you a disadvantage.
    So a first basic test we should have is for example a 10km straight with a time measurement after 5km to 10km. With same settings everybody should get same result +- 0.1sec for the 5km. In case times differ, there is an issue with the internal timing or controller .
    oh and I had some races online with unexplainable slow lap times, after race was over I rejoined server with same settings and was able to lap significantly faster. So the tests must be done in a full field
     
    Last edited: Apr 30, 2019
  11. stonec

    stonec Registered

    Joined:
    Jun 19, 2012
    Messages:
    3,399
    Likes Received:
    1,488
    This is why there is the Ctrl+C physics graph to easily be able to track the physics loop occupancy. The graph in normal cases is only about 50% used, which means there is another 50% CPU margin available. It's only when the physics graph starts nearing maximum that there is any likelihood of a missed calculation. In my experience with using an old i5, maxed out physics have only ever occurred on a "bad" car mod or track.
     
  12. Stefan_L_01

    Stefan_L_01 Registered

    Joined:
    Nov 18, 2012
    Messages:
    595
    Likes Received:
    386
    you would not see a missing Frame at 400khz in the display, nor would you ever see that your System Clock runs at 400.1 hz.
    I encounterd so many strange situations regarding lap times online that I do not trust it anymore. Sometims i could lap 0.8sec faster for one lap (0.4/min), quite often being alone, mostly it was the other way around, being in a race with many around.
    At the Moment we do not even know what grip Level is present. Only visual Impression (Darkness) is there. Too bad (confirmed by others) that you join a Server with advanced rubber look and it can take you up to 10 laps, including some box Exits, until you can feel full grip in terms of sensemaking laptimes.
    I mean you script a practise session with 15°C and light Clouds for all 5 time sectors, join offline rf2 and what you see are 5 Icons tellling you something completly different, 25°C with rain and so on. Do you really trust a program regarding core timings and network interactions if even some obvious simple visual gui items are totally wrong since years?
     
  13. stonec

    stonec Registered

    Joined:
    Jun 19, 2012
    Messages:
    3,399
    Likes Received:
    1,488
    The UI is irrelevant because ISI literally didn't give a damn about it, they were all about developing gMotor. The physics engine used in rF2 has been in development for over 20 years and as far as I understood, is partly borrewed from rF Pro, which most definitely must be running real time with maximal certainty. The way a real time system works is that there are safety measures in place in order to avoid a situation of a "frame skip", as you describe it. Much like an airplane software has certain rules and restrictions to guarantee that no hick-up occurs, physics engines of professional sims have similar kinds of safety measures. It basically comes down to the point that a sim physics engine has to perform equal calculations on every type of hardware. A graphics engine can scale with performance and "skip frames", but this would be detrimental in a physics engine.

    Marcel described it more in detail somewhere, but essentially there are certain timeframes within which each physics operation has to complete. In between that, there is an inbuilt overhead, so if one physics operation gets slightly delayed, the engine is designed such that it won't cause a "frame skip" under normal conditions. This is a bit speculation from my side, but I'd argue that the only way you'd ever be able to introduce a physics "frame skip" is if the sim runs out of real time, which basically screws up the entire lap time.

    If there was an easy way to make the physic engine skip, say one calclation, this could easily be used for cheating, because who is to say that a skipped calculation will make you slower? It could do the opposite as well.
     
  14. Lazza

    Lazza Registered

    Joined:
    Oct 5, 2010
    Messages:
    12,346
    Likes Received:
    6,572
    You would, because when it drops behind realtime the histogram sweeps from side to side; you don't need to see the horizontal bar hitting the right hand edge for a split second.

    If you use DAMPlugin you'll also have a log of total realtime loss, so any small freezes have nowhere to hide. (accurate to telemetry rate, 100 Hz so 0.01s, for the life of the log)

    As for minor frequency differences, games are made to cope with that. Any aberrations would be detected and dealt with long before you'd be able to notice.
     
  15. Stefan_L_01

    Stefan_L_01 Registered

    Joined:
    Nov 18, 2012
    Messages:
    595
    Likes Received:
    386
    Seriously, I do not care why rf2 is off sometimes. But when we go through practise, quali and warmup I just expect that without Change in Environment (as we use real weather, so temp changes not that much) that I can lap in a race with clean air as good as 1 hour before in practise with less grip. I mean in practise when your lap times get stable as it takes some time until you feel the grip you see, or lets say until you get the laptimes which correspond to the grip Level. Others (few however) can do! The Server is available all time the days before so you know what is possible the first 10 laps and beyond if you try.

    And if your Clock or hpet is off it IS off also for your logs, so you would need to set an IO (e.g. RX of Serial line) and measure by a scope the frequency to be sure. However I agree that the phenomen I experienced can not be explained by a Clock Deviation, they should run with sufficient precision today
     
    Last edited: May 1, 2019

Share This Page