Finding the max (or near max) overclock for multiple RPi3B's

[CVivian]

Hi all,

I recently purchased a large number of Raspberry Pi 3 model B devices, and I am no stranger to overclocking them. However, I know that some boards can overclock more than others (sometimes quite substantially) and I am looking for a way to get close to the maximum overclock for each individual board.

Is there a way, rather than just slowly increasing the values for each setting on each board, to find the maximum overclock settings that a particular board can handle?

Any input is greatly appreciated.

[MaxK1]

If there is, I haven't found it. I have 4 Pi 3's - 3 of which overclocked to ~1.35GHz. The newest one overclocked to 1.43GHz. All have heatsinks and fans and I try to keep the CPU temp below about 45' while running a mainframe emulator and my weekly full backup of the entire system. That pushes the CPU usage to about 65%-85% and if it is still running after a week, I consider it "stable". But each one has slightly different settings so they each need to be hand tuned to get there.

PS - If you have ever watched "Reefer Madness" (the 1936 version) you'll understand the "overclockers problem"

[henryww]

Hi all,

I recently purchased a large number of Raspberry Pi 3 model B devices, and I am no stranger to overclocking them. However, I know that some boards can overclock more than others (sometimes quite substantially) and I am looking for a way to get close to the maximum overclock for each individual board.

Is there a way, rather than just slowly increasing the values for each setting on each board, to find the maximum overclock settings that a particular board can handle?

Any input is greatly appreciated.

Perhaps create a script that runs on boot and stress tests the pi for a reasonable amount of time if the test passes it will edit the config.txt slightly increasing the overclock and then reboot and repeat until the test fails or the pi fails to boot where it will not edit config.txt.
Once you have a pi that won't boot or fails the stress test, roll back to the last successful clockrate manually (probably a good idea to have the script append a log of successful clockrates).

If you make several images of Raspbian that does this you could save a lot of time finding the potentially more overclockable boards and maybe you will win the silicon lottery