Huge difference in performance between the "modes" qualification and race

Discussion in 'Technical & Support' started by DrivingFast, Oct 18, 2019.

  1. DrivingFast

    DrivingFast Registered

    Joined:
    Mar 1, 2018
    Messages:
    1,638
    Likes Received:
    1,083
    Why is there such a huge difference in performance between the "mode" qualification and the "mode" race ?

    With of course the same race conditions (rain, time of the day, time scale, clouds, .....), the number of AI (determining point) and the exact same graphic settings.
     
  2. McFlex

    McFlex Registered

    Joined:
    Feb 23, 2012
    Messages:
    1,031
    Likes Received:
    317
    There is no difference
     
  3. DrivingFast

    DrivingFast Registered

    Joined:
    Mar 1, 2018
    Messages:
    1,638
    Likes Received:
    1,083
    In offline, there is, 100% sure. 100% of 100%
     
  4. McFlex

    McFlex Registered

    Joined:
    Feb 23, 2012
    Messages:
    1,031
    Likes Received:
    317
    No there is still no difference. Even if you are 1000% sure
     
  5. davehenrie

    davehenrie Registered

    Joined:
    Jul 6, 2016
    Messages:
    7,453
    Likes Received:
    4,369
    In real life, Qualifying is always faster than race pace. Teams take extreme measures to just have enough fuel for the number of laps they can run for qualifying, they cheat on settings like camber that provide heat faster to the tire but would never last a full fuel stint. (these are some of the reasons Parc Ferme rules were created.)
    A-lot of rF2 users have stated the the AI, for some reason, use hard tires and thus are slower than would be expected.

    BUT.

    for the 1000th time.

    S397 is NOT, repeat NOT working on Solo/AI game issues. They just are not. They will not until the UI and Online Comp systems are matured and usable. Any AI work will not be addressed until those online gaming programs are done. Until someone like Christopher or Marcels comes on and states they are now investigating offline issues, I won't expect any solutions. None, Zip, Zilch, Nadda. NoWayJose.
    Don't argue that they should, don't argue that they are ignoring offline users, don't argue that your immersion is killed. S397 knows all that, they have chosen to focus 100% on the Competition system and UI.
     
  6. McFlex

    McFlex Registered

    Joined:
    Feb 23, 2012
    Messages:
    1,031
    Likes Received:
    317
    Are you sure he is complaining about AI performance? o_O
     
  7. stonec

    stonec Registered

    Joined:
    Jun 19, 2012
    Messages:
    3,399
    Likes Received:
    1,488
    Yeah I'm not sure, is this thread about graphical "performance", performance relative to AI, or absolute lap times? I have no idea.
     
    David O'Reilly likes this.
  8. davehenrie

    davehenrie Registered

    Joined:
    Jul 6, 2016
    Messages:
    7,453
    Likes Received:
    4,369
    language can be tricky, but I ASSUME humans are relatively consistent when qualifying and racing. My assumption was that the AI was displaying wide swings between race pace and qualifying. If I'm in error, my appologies. If he WAS asking about AI issues, I was trying to convey the futility of such complaints at this time.
    The notion of system performance, i.e. fps, never entered me noggin. I guess I should try to consider more meanings.
     
    Last edited: Oct 18, 2019
  9. DrivingFast

    DrivingFast Registered

    Joined:
    Mar 1, 2018
    Messages:
    1,638
    Likes Received:
    1,083
    I'm talking about graphic performance, if we can call it that.

    And it seems directly related to AI.

    - In qualifying, in exactly the same racing conditions + graphical parameters, it's almost smooth with 50 AI.
    - On the same track, exactly the same conditions and graphical parameters, AI visible 6 too, and only 10 AI, the performances graphical are clearly inferior.

    This is not new, it has always been like that but it is a problem.

    I'm talking about graphic performance between the qualifying session and the race session.

    Performance problem, related to the AI, but graphic problem (not laptime).
     
  10. Coutie

    Coutie Moderator Staff Member

    Joined:
    Oct 5, 2010
    Messages:
    3,779
    Likes Received:
    2,241
    Does it happen on a specific track? You can turn stuff on and off in different sessions, maybe there are a lot of added objects or something. I don't see any other reason it would happen.
     
  11. davehenrie

    davehenrie Registered

    Joined:
    Jul 6, 2016
    Messages:
    7,453
    Likes Received:
    4,369
    I understand now. Part of the problem is the amount of detail that tracks display on race day is often far greater than what is shown in qualifying. If you look at the grandstands in qualifying, they are often empty or nearly empty. But on race day the stands are full, the infield is packed with lots of increased graphics. There is just much more being displayed. Also the cars tend to be strung out more in qualifying and compressed on the track during a race, especially at the start.
    So lets try a slightly different test.
    Begin a qualifying session and immediately note the fps.(ctrl-f) Then drive at least 10 minutes or preferably longer. Note the fps again and do this while driving not sitting in the garage.
    Now start the race session.(same cars, no changes)
    Note the fps with all the cars on track at the start. Take another fps reading at the end of the first lap and then one more at the same time as your 2nd qualifying session.(in my example 10 minutes)
    This should give you a clear picture of the difference between the qualifying and race sessions. Also take a look at the track detail between qualifying and raceday. You should see some difference in the amount of trackside objects and grandstands. Take a screen shot of the track side at the same point in both qualifying and raceday.
    Should be interesting.
     
    Last edited: Oct 18, 2019
  12. Lazza

    Lazza Registered

    Joined:
    Oct 5, 2010
    Messages:
    12,345
    Likes Received:
    6,572
    You aren't doing logical tests. You can't say it's directly related to AI when you get worse performance in a race session with fewer AI. (why are you even changing AI?)

    Take out AI completely and compare sessions. If you get different performance it's nothing to do with AI (obviously) and probably about session specific track detail.


    Change one thing. Test.


    If you change multiple things, and get different results, you can't draw any conclusions. Nothing at all.
     
    Daniele Vidimari likes this.
  13. DrivingFast

    DrivingFast Registered

    Joined:
    Mar 1, 2018
    Messages:
    1,638
    Likes Received:
    1,083
    This happens more on certain tracks like Monaco ePrix or Nurburgring.

    I tried to drastically lower the graphics, and I still get this difference in graphical performance between qualification and race.

    I'm pretty sure it relates to AIs and their impact on performance.

    I just acquired an extremely powerful PC with the exception of the GPU which is still a GTX 1080.

    rF2 and Windows 10 reinstalled.

    I'm trying to make the AI visible to a minimum.

    I thought it was a CPU problem, I wonder if this is not a VRAM problem.
     
  14. DrivingFast

    DrivingFast Registered

    Joined:
    Mar 1, 2018
    Messages:
    1,638
    Likes Received:
    1,083
    I did not notice, in 7 years of rF2, that there were more more objects on the track in race than qualification, to be honest.


    @Lazza

    I did not specify the complete test procedure.

    Whether for qualifying or for the race, here are the identical points :

    - Same number of AI (I actually tested with less in race afterwards).
    - Same multiclass racing with the same cars.
    - Tested with for example 10, 25, 50 AI.
    - Same time of day (5:00 AM).
    - Same track (Nürburgring combined).
    - Rain with mostly cloudy.
    - Recording FPS racing from ~3 minutes (maybe I should do it from 5/7 minutes ?).
    - Same graphic parameters.
    - Same number of visible AIs.
    - Same Porsche GT3 Cup car.
    - etc. etc.

    I would be glad there was no problem and I just misunderstood something.

    Indeed, in the race the cars are in package : would that be a sufficient explanation ?
     
  15. McFlex

    McFlex Registered

    Joined:
    Feb 23, 2012
    Messages:
    1,031
    Likes Received:
    317
    Of course it does :rolleyes::rolleyes::rolleyes::rolleyes::rolleyes::rolleyes::rolleyes::rolleyes::rolleyes::rolleyes:

    Case solved. Next.
     
  16. DrivingFast

    DrivingFast Registered

    Joined:
    Mar 1, 2018
    Messages:
    1,638
    Likes Received:
    1,083
    FPS are halved (approximately) with only 6 AI visible.

    Do you think that this extremely high drop in performance is normal, in this situation ?

    I agree that the performances are logically inferior, but as much ?

    I specify that I have no particular knowledge in terms of video game, graphics engine or normal/abnormal performance in a video game.
    I am a normal user without specific technical knowledge.

    I'm just trying to understand, and put forward a problem that I thought existed.
     
  17. Lazza

    Lazza Registered

    Joined:
    Oct 5, 2010
    Messages:
    12,345
    Likes Received:
    6,572
    Can you put 1 AI car in, follow it in qualifying and in the race, and compare performance?

    Or, if you're feeling talented, try to follow more cars but the same number in both sessions.

    You're starting to state a performance difference between qual and race in other threads, but I don't think you've established it as fact yet.
     
  18. DrivingFast

    DrivingFast Registered

    Joined:
    Mar 1, 2018
    Messages:
    1,638
    Likes Received:
    1,083
    I have already stated many times that there is a difference between the two.

    As I said before, exactly the same test, same number of AI, even multiclass, same car, same weather, same clouds, same hour of the day, same track, same time scale etc etc etc etc.

    During my tests I have always tried to have a maximum of AI around me, whether in qualifying or racing.

    And in both cases with only 6 AI visible, and graphics strictly identical.

    I found a solution to play in acceptable conditions with very good graphics in VR with 61 AI (10 GT3) on the Nürburgring combined, if it confirms, I have no more reasons to waste my time doing tests, even if you do not believe me.

    In addition my PC is disconnected during the next days.

    There is only one thing that could explain the huge difference in performance IMO, but I doubt that the explanation is sufficient is that in the race there is more AI to display at the same time, but as I use only 6 AI visible, I doubt that it explains everything.

    I noticed a problem that has existed for a long time, at S397, to check whether there is a problem and to take the appropriate measures.

    If they do not, at least they will have been warned.
     
  19. Lazza

    Lazza Registered

    Joined:
    Oct 5, 2010
    Messages:
    12,345
    Likes Received:
    6,572
    It's not about believing you or not; in a race, especially at the start, cars are running nose-to-tail. In qualifying that is more rare. You didn't say you tested with a single AI which is why I suggested it - it takes away the variability of (AI) car spacing etc.
     
  20. DrivingFast

    DrivingFast Registered

    Joined:
    Mar 1, 2018
    Messages:
    1,638
    Likes Received:
    1,083
    With 0 AI, I tried a race and just after a qualification with exactly the same conditions (at 100%).

    And indeed I saw a real difference.

    The problem is that I have tested once, I am surprised at the result, but as I told you it should retest, and I do not do it again :

    Because I spent dozens of hours since I started on rF2 trying to understand the physics engine (2000 hours of play), I'm fed up now.

    On rF2, using the day/night cycle uses excessively the resources I think.

    Same for the use of AI, especially in multiclass (CPU + GPU).

    The rain should either be more realistic graphically or consume fewer resources.

    Other than that, the performances are good ..... but if you want to do endurance races with day/night cycle + rain, it's hell.

    I hope new optimizations in these areas.
     

Share This Page