search

owncloud-archive / pi-image

Scripts and configurations for Raspberry PI disk images
29 stars 8 forks source link

Overclocking #18

Open WaaromZoMoeilijk opened 8 years ago

WaaromZoMoeilijk commented 8 years ago

Hey guys, I would like to know your thoughts about overclocking the RPI. Next I will share my findings.

OS: Ubuntu Core 15.04 apt-get package manager Hardware: RPI2B

The following settings void warranty. I have been using this setup for over 2 months, its runs smooth without any errors. It does seriously make a difference running for example apt-get update & apt-get upgrade and other general load times of the front end. I will try to post the benchmarks with and without later on. I do have 3 heat sinks installed on the RPI2B and a CPU Fan.

These are the most extreme settings I could get out of the RPI2, with higher values its unable to boot.

Settings: gpu_mem=16 boot_delay=1 force_turbo=1 initial_turbo=30 temp_limit=80 arm_freq=1100 core_freq=500 sdram_freq=500 over_voltage=8 over_voltage_sdram=8

A safer and better alternative is to use 'raspi-config' which also allows you to do various other useful things, which of course could also be automated... https://github.com/asb/raspi-config https://www.raspberrypi.org/documentation/configuration/raspi-config.md

How to install on linux: https://github.com/snubbegbg/install_raspi-config

Be aware that overclocking may reduce the lifetime of your Raspberry Pi. If overclocking at a certain level causes system instability, try a more modest overclock. Hold down shift during boot to temporarily disable overclock.

A small indication:

image

The overclocking options here do not void warranty, are supported by the RPI2B dev's and do speed up the RPI2B.

These are the options:

Overclocking Settings

ARM, SDRAM and GPU each have their own PLLs and can have unrelated frequencies. The GPU core, h264, v3d and isp share a PLL, so need to have related frequencies. pll_freq = floor(2400 / (2 * core_freq)) * (2 * core_freq) gpu_freq = pll_freq / [even number]

The effective gpu_freq is automatically rounded to nearest even integer, so asking for core_freq = 500 and gpu_freq = 300 will result in divisor of 2000/300 = 6.666 => 6 and so 333.33MHz.

Standard Profiles: arm_freq core_freq sdram_freq over_voltage

None            700            250             400              0
Modest          800            300             400              0
 Medium          900            333             450              2
 High            950            450             450              6
 Turbo           1000           500             500              6

The options below (not default) explain a bit more about all the different options

https://github.com/ezraholm50/BerryCloud/blob/master/config.txt

Please let me know how you guys think about this!

Kind regards,

Ezra

tflidd commented 8 years ago

The moderate overclocking, which doesn't void the warranty, could be interesting. Do you have any numbers on the performance improvement? Does it help during heavy file operations?

WaaromZoMoeilijk commented 8 years ago

Sorry for the delayed reply, been busy.

No hard numbers, only thing I have is when the overclocking is enabled it saves me 20 minutes or more installation time (My BerryCloud repo) of total 60 mins without overclocking.

Not that into benchmarking etc yet so its hard for me to get some solid figures posted. Could use some help on this one.