Hardware Reference
In-Depth Information
sdram_freq=500
Just like with
arm_freq
, you'll need to reboot your Raspberry Pi for this to take effect.
Increase GPU Frequency
Your last major overclocking option is the GPU components, the frequencies of which
are all defined by
gpu_freq
and default to 250 MHz.
gpu_freq
is a sort of
super setting
. Setting it assigns the same value to the
core_freq
(GPU processor core frequency),
h264_freq
(hardware video block frequency),
isp_freq
(image sensor pipeline block frequency), and
v3d_freq
(3D block frequen-
cy). If you have a GPU-intensive task, you might get some extra performance by in-
creasing the
gpu_freq
to 325. You can do this by adding this line to
/boot/config.txt
:
gpu_freq=325
That said, we don't recommend changing the
gpu_freq
value, because it will take per-
formance away from the CPU. Instead, you might try just changing the
core_freq
value.
If you do this, it is important to keep all of the GPU frequencies (listed previously)
either the same or different by a factor of an integer multiplier. If you do not do this,
the GPU components will receive a mixture of incompatible pulses and things will stop
working very quickly.
However, because the
core_freq
value also includes the L2 cache and some of the
SDRAM memory clock cycles, increasing just that value could give the ARM CPU a
performance boost. Multiply the default value by 2 (the largest integer that will really
work) and set the value to 500 in
/boot/config.txt
like this:
core_freq=500
Note that this might not work. Some people report success, while others report
failure. If you try to mix this
core_freq
change in with the other overclocking fea-
tures, it might work only when they are set low (or left at the default).
We cannot emphasize this enough: sometimes, when overclocking fails, it does so
in less-than-obvious ways. Reliable programs become buggy, hardware devices
stop working at random, and the system might just reboot for no good reason.
When you do overclock, you'll want to have a quantifiable test case that you can run
over and over again to see what gives you the best performance for the workload that
you care about on your specific Raspberry Pi. Do not simply download a canned
benchmark and trust it. A benchmark designed to show GPU performance will not
help you optimize your overclocked Raspberry Pi system for tasks that are CPU-
bound.