CPU usage vs Main_t value

Discussion in 'Assetto Corsa' started by Crispin Williamson, Jul 5, 2017.

  1. Crispin Williamson

    Crispin Williamson
    Washed up Supermoto racer in an MDF box in York

    Messages:
    167
    Ratings:
    +59
    Hi,

    I'm playing around with my PC altering graphics settings and played with overclocking to see if it made much difference, it didnt. When playing with a full grid my Main_t value gets near 100% and I get frame rate drop, when I have MSI running however it shows that my CPU (all 4 cores) are nowhere near maxed out, my graphics card (980ti) is plodding along.

    Am I missing something here?

    Cheers
     
  2. kunos

    kunos
    Kunos Simulazioni

    Messages:
    240
    Ratings:
    +973
    yes, the meaning of the value (MAIN_T) in AC.
    The number shows the relationship between the time spent inside the game code in a frame and the total time of a frame.
    No relationship whatsoever with a CPU occupancy value, and it was renamed just with the hope that people will stop treating it as a CPU occupancy indicator and getting confused about it.
     
    • Like Like x 2
    • Love Love x 1
  3. Crispin Williamson

    Crispin Williamson
    Washed up Supermoto racer in an MDF box in York

    Messages:
    167
    Ratings:
    +59
    Thank you, so my performance is limited by this factor and not my CPU or GPU? Can I do anything to improve performance?
     
  4. kunos

    kunos
    Kunos Simulazioni

    Messages:
    240
    Ratings:
    +973
    A high value of main_t suggests the game could be CPU bound... but it could really be other things from RAM speed to BUS speed, PCs are complicated beasts.
     
  5. Crispin Williamson

    Crispin Williamson
    Washed up Supermoto racer in an MDF box in York

    Messages:
    167
    Ratings:
    +59
    My CPU values are quite low. Anyone got any tips?
     
  6. KToMmi

    KToMmi

    Messages:
    7
    Ratings:
    +0

    Hi, did you ever manage to lower your MAIN_T values?
    I have been experiencing the same issues for a couple of weeks even though my CPU isn't bad at all (i7-6820HQ) and even though it has always worked perfectly fine. I have not found a freaking solution yet.
     
  7. dud

    dud
    Premium

    Messages:
    433
    Ratings:
    +66
    Just because no specific CPU core is maxed out during an observation period that does not mean that there isn't a single-threaded segment holding up the show. A single thread can be freely rescheduled to run on different cores. Looking at that thread in isolation and at that complete process would reveal that there is 100% CPU usage of one CPU during that period, however which specific hardware core provided that execution unit can change during the observation period. But that is hard to observe for verification from the outside, you generally want source code and debugging for this.
     
  8. RasmusP

    RasmusP
    AC Stuff and G27 help Staff

    Messages:
    3,372
    Ratings:
    +1,214
    From my experience it's just your CPU.
    But as said, it's complicated!

    What you can do about it:
    Overclock the CPU and/or your RAM. If you can't do that (any higher) you're left with changing settings like:
    - disable all apps/hud
    - disable the apps you don't need in the main menu
    - put reflections to static
    - reduce shadow settings
    - run fewer AI

    If you're running any limiter and not vsync, lower the fps limit to something, you can maintain more consistently.

    The root problem is that you just cannot split some things of the program any further after some point so as stated above, there is one thread that loads one core to 100% for a split second you won't see on any CPU load measurement. But it's there!

    Hopefully the next generation of Sims will have better multicore support. Or better said, "splitability" of threads.
    Funny thing where my knowledge ends is that hyper threading, more cpu cores etc increase the fps although it's a single thread problem.
     
    • Love Love x 1
    • Agree Agree x 1
  9. dud

    dud
    Premium

    Messages:
    433
    Ratings:
    +66
    As always the best test to see whether you are in any way CPU capped (multithreaded or not) is to downclock the CPU. Downclocking is a lot easier to do (just set the CPU divider lower, which almost all BIOSes allow) than overclocking, and of course safe.

    You should disable turbo mode for those experiments, though. So that your downclocked CPU doesn't end up running more turbo, messing up the result.
     
    • Agree Agree x 1
  10. Fernando Deutsch

    Fernando Deutsch
    Premium

    Messages:
    681
    Ratings:
    +347
    Interesting post.

    Here are some samples of screen captures where MAIN_T is high and seems no CPU/GPU is intensive used. RivaTunner stats are GPU temp/%/FB/BUS/MEM.

    First view the 3D scenery is complex as it includes main straight and Eau Rouge. Check the value of SCENE and TRI are quiet big.
    acs_2018_02_26_15_42_42_771.jpg

    Another 3D scenery of the straight Eau Rouge/Les Combes. Scene is not that complex but TRI is kind of high due to high LOD cars.
    acs_2018_02_26_15_43_43_722.jpg

    This is your "regular" camera view with no complex 3D scenery
    acs_2018_02_26_15_44_33_550.jpg

    CPU/GPU stats shows they are not really stressed. Same could be said from FB and BUS usage. I am using a 2500K @ 4.2 Ghz, GeForce 1060 3GB. PCIe 16x 2.0.

    PD, @kunos please use coma separators on the stat numbers, my eyes hurts trying to read if they are millions,billions,gazillions :cry::D