Project Mercury - CPU Priority Booster for heavy multitaskers

Discussion in 'Operating Systems' started by SvenBent, Apr 10, 2015.

  1. SvenBent

    SvenBent 2[H]4U

    Messages:
    2,485
    Joined:
    Sep 13, 2008
    Just a heads up

    to be frank if you only use you compatu to luanh a game you might not experience any great improvement if any at all. its mostly when the system is under a severe load that the biggest effect comes into place.
     
  2. SvenBent

    SvenBent 2[H]4U

    Messages:
    2,485
    Joined:
    Sep 13, 2008
    Just a small teaser

    I'm currently working on implating reduced I/O Priority on minimized programs. hopefully that means backgrounds that a heavy on the I/O should have less of a negative effect on your game/main application.
    first test show a bit improvemnts (+30%) in avgFPS while running two instancese of cipher wiping of mechanical drives. but both with and without PM the game was unplayable :(
     
  3. Meeho

    Meeho [H]ardness Supreme

    Messages:
    4,178
    Joined:
    Aug 16, 2010
    Any further progress with this?
     
  4. SvenBent

    SvenBent 2[H]4U

    Messages:
    2,485
    Joined:
    Sep 13, 2008
    Not a lot of new stuff has been added.
    Stabiltiy issues with my Ryzen setup has made it rather harder to fully test balacing load beterr CCX and SMT.
    and no bug fixes has been needed



    PRocesslasse has a lot of bells and whistles to it which means more overhead and there for a harder hit on your performance
    - it uses more memmory
    - it uses more CPU ressoruces

    Processlassy is not built withe the no loop monitring approcuhy mine has whih mean its consnntlso sitting and monitoring for things eating up cpu when you are just sitting in your game.
    When you are in your game my programs eats an absoly ( not found a single programs that was able to measure it) 0 cpu time

    Process lasso in its deafulst settings will fall flat on it ass under havey multi processions because it looking for a SINGLE process that is misbehaving

    examples witha 4 core system
    you have 8 process runnings eating up massive amount of cpu.
    they are each using 50% of a core
    process lassy does not seem to find anything that are using more than 80% ( or whatover the treshold is) so its declaring that the systme is running fine and does nothing
    you pc is still laggin

    Project Mercury will still give whatever application you are sitting and working in ( or gaming in) higher cpu priority so its less affected by the high CPU load and threby resolves some of the lag


    This is a very easy test to do, if you want to confirm the issue


    So in short:
    Process lass uses more resources to do a less effective work. but it looks fancy with all those graphs
     
    Last edited: Nov 22, 2018
  5. SvenBent

    SvenBent 2[H]4U

    Messages:
    2,485
    Joined:
    Sep 13, 2008
    I must admit I have a hard time believe it will prevent crashes directly. unless it for some reason priority you cpu to a thread with less cpu intensive workload and thereby lessed the intensity of the load on the cpu. but then its because you system is unstable
    however if you have any emperical data on it i would be interested

    I'm sorry do you know what vaporware is ? I having a hard time putting that tag on a released software, that is available for everyone to get.
    Why is updates impotent? I'm not interested in bloating my software. it has features i have tested and verified makes a difference.
    If you want to waste resources on bloat because it keeps a program updated then go ahead and use PL.

    but for any kind of technical viewpoint none of this seems to have any merits on it
     
    Last edited: Nov 22, 2018
  6. SvenBent

    SvenBent 2[H]4U

    Messages:
    2,485
    Joined:
    Sep 13, 2008
    just to put in some old resource numbers

    Code:
                      Work set     Commit size   CPU usage (Monitoring)
    Mercury 1.2.0.1    588-  616k   3336- 3336k  0%
     
    TopWinPrio 1      8276-11176k  29328-32228k  0.018%
     
    TipWinPrio B3     6880-9792k   47776-50268k  0.024%
     
    ProcessTamer      2400-2600k    3136- 3416k  0.05%
     
    ProcessLAsso           2108k         11748k  0.156%
    
    Memory usage was taken from taskmanagers detail mode
    CPU usage was taken from AverageCPUcycles http://www.boray.se/software/averagecpu/

    The difference are minuscule. But I don't think it optimal for a software there is supposed to optimized CPU resource are eating into it, when you can do it in a way that it does not.

    These are all taken with the software running in its out of box settings.[/CODE]
     
    Last edited: Nov 23, 2018
    Meeho likes this.
  7. SvenBent

    SvenBent 2[H]4U

    Messages:
    2,485
    Joined:
    Sep 13, 2008
    checking in to see if NemesisX had opdated with how release software could be considred vaporware but sadly it seems like he delted his post when counter argued.


    Insteed om posting these olf benchmark results that show how PRoject Mercury improves gammign excperience under heavy multitasking

    No load = no background load
    7-zip Load = 7zip benchmarks is running in hte backgroudn gobblign up CPU
    7-zip +PM= 7zip benchmarks is running in hte background gobblign up CPU + Project mercury is running to improve performance

    ?u=https%3A%2F%2Ftse1.mm.bing.net%2Fth%3Fid%3DOIP.8xgo7BhywMoGees2hJu_twAAAA%26pid%3D15.jpg

    ?u=https%3A%2F%2Ftse1.mm.bing.net%2Fth%3Fid%3DOIP.LCblsd1VVLc7oNy9kOAs9QHaH_%26pid%3D15.jpg


    You will notice some of these have 0 improvements.
    I do not cherry pick my results but show the full test data of that system ( i had more but cant find them)
     
  8. Meeho

    Meeho [H]ardness Supreme

    Messages:
    4,178
    Joined:
    Aug 16, 2010
    Have you tested it on some of the newer better CPUs?
     
  9. SvenBent

    SvenBent 2[H]4U

    Messages:
    2,485
    Joined:
    Sep 13, 2008
    I had but unable to locate them atm.
    but it doesnt matter how new your CPU the issue remains the same. it just bigger FPS/points numbers on all 3 sides.
     
  10. SvenBent

    SvenBent 2[H]4U

    Messages:
    2,485
    Joined:
    Sep 13, 2008
    Just a quick test i did today on a Ryzen 1700x (stock speed) + gtx970

    Code:
    Super position
    1080 medium, DirectX
    
    Normal   Load   PM   PL
    5274   684   5237   748 
    5289   682   5020   568
    5262   593   5024   675
    The load was made by running 2 instance of my png optimizing batches witch runs a pngout.exe for each cpu logical cores for a total of 32 cpu heavy processes.
    both Project mercury (PM) and Process lasso (PL) was run with this load as well and with defaults settings.
    as it can see under heavy multitasking with a ton of CPU heavy process the method Process Lasso uses are simply inefficient.

    Here is what happens with PL. PL finds 32 process that are eating up a lot of CPU and start adjusting the priority
    then once it gone through some of them the new lowered ones are no longer getting CPU time because the remaining to be adjusted are eating it all.
    process lasso the adjsut the priority back up since they had a low cpu load. The CPU load shoot back up and process starts over.
    continuous adjusting CPU priorities up/down in an epileptic seizure


    I mean just look at the log
    Image1.png



    Now unlike the maker of process lasso I don't try to make ppl test with my own home brewed test software.
    You can do the test with any CPU hungry real world application. just start up 2 processes for each cores and PL just can't handle it

    and even if you are not multitasking to the point of PL breaking down in panic it isstill using more ressources then my simple program.


    -- edit ---
    if you loo kat the bottom of the picture you will notice the same process ( loo kat the PID) was put bakc to normla priroty and then 2 secs later it was lowered. this process was always going to aeat as much cpu power as available that means PL is roughlt 2 secs about responding. constantly leaving cpu heavy process in for 2 secs before adjusting them back down. plenty of time to ruin the smoothies of a game
     

    Attached Files:

    Meeho likes this.
  11. SvenBent

    SvenBent 2[H]4U

    Messages:
    2,485
    Joined:
    Sep 13, 2008
    just dumping this here


    The performance enhancement is no joke. I am also running with Vsync enabled for gsync, so it's possible that I would have gone even higher than 143 fps since that's where my sync should be holding it anyway. I still did drop to 130 ish here and there, but on the previous settings where I capped out at 117 fps, it would drop to 100 fps as well.


    22% performance boost. :happy: