difference schedule up migrate and group up migrate in hardcore tuning

Jaga LilinJaga Lilin Level 1
edited March 12 in ROG Phone 5

Please tell us some basic information before asking for help:

  1. Model Name: asus rog phone 5
  2. Firmware Version: 2112.210
  3. Rooted or not: not rooted

guys i tried to change value in advance harcore tuning

but i dont know the difference between "/proc/sys/kernel/sched_upmigrate" and /proc/sys/kernel/sched_group_upmigrate

can any one tell what the diference i tried to search in the internet but only found this one


/proc/sys/kernel/sched_downmigrate

/proc/sys/kernel/sched_upmigrate

/proc/sys/kernel/sched_group_downmigrate

/proc/sys/kernel/sched_group_upmigrate

Kernel Run Job Schedule

If the job is running more busy than the upmigrate number (percentage), schedule the job to the big core or the super core

If the job running busy is lower than the dowbmigrate number (percentage), schedule the job to the big core or the small core

Comments

  • Anyone can help me??

  • Anyone can help me??

    can anyone ??

  • Indeed, the doc is almost non-existent for groups, it seems that this is related to CPU clusters and task numbers.

    With upmigrate, a task using for example 80% of a 1.8 CPU for a given time will switch to CPU 2.42, if it still uses 80% it will switch to 2.84 or go down to 1.8 if it uses -60%.

    For Groups, it would be the same but only if 2, 3 or 4 tasks would be eligible, example if 3 tasks use 80% of the 1.8 cores they would pass on the 2.42 cores

    The values ​​given are only examples and other parameters such as temperature and time are taken into account

    2.84GHz Single-Core Kryo 680

    2.42GHz Triple-Core Kryo 680

    1.8GHz Quad-Core Kryo 680

    It is risky to risk this setting without a perfect knowledge of the kernel, the game software concerned and the analysis and monitoring tools, not to mention that the possible gain (on a kernel with good default values) is negligible. compared to simpler operations such as the deletion of numerous logs, which in addition do not generate an increasing temperature, which is already difficult to control.


    Good luck in your research

  • thank you for your explanation and 1 more thing 


    lower clock meaning lower power usage so, is it better to lower the clock speed so the cpu utilization can be high ?


    in my mind higher clock speed but low cpu utilization is like use a big truck but with low cargo

  • In absolute terms you are right, the higher the frequency, the higher the consumption (and the temperature) will be.

    In practice it is different, the cores are not blocked at their maximum frequency so the 2.84 core will not consume more than one of the 1.8 cores if it is used at this frequency.

    For the comparison with the truck, it is necessary to take into account the power requested at a time T and the number of times necessary because there is latency at each change of CPU group.

    Concretely, if 1.8 is used at 70% and frequently asks for 85 or 90%, a change of group at 80% is not the right solution, you have to test at 70% (it will stay more often on 2.42 or 85-90% it will stay on 1.8). There is no miracle solution, it depends on each software, its programming, the resulting temperature, this last point is the most critical on mobile.

    To go further it takes long hours of analysis with software like https://www.gamebench.net/ but above all an effective cooling system and to date there is not really one

    There is a watercooling system but it is restrictive and on ROG 5, the fixing comes on the triggers

    The coolest thing would have been that ASUS was inspired by this cooler to make its own by including it in the ROG Kunai 3 (no need to buy it the Rog 5 is too long)

    https://www.watch.impress.co.jp/donya/DN-915903.html

This discussion has been closed.