Hardware Reference
In-Depth Information
age during normal operation. When the switch changes, the capacitance of the tran-
sistor and the voltage applied affect how quickly the switch output changes. Config-
uring a circuit to use higher voltage (“overvolting”) allows the circuit to react faster,
which permits you to overclock the hardware further than what would normally be
possible.
The Raspberry Pi firmware exposes some configurable voltages, which map up with
the following values in
/boot/config.txt
:
•
over_voltage
(core)
•
over_voltage_sdram_c
•
over_voltage_sdram_i
•
over_voltage_sdram_p
If you do
overvolt
your Raspberry Pi by changing any of these settings, it might
permanently set a fuse in your BCM2805 system on chip. That means that the
vendor will know if you overvolt the hardware, it burns out, and you try to return it
as defective. We shouldn't have to say that it's not OK to return things as defective
when you were responsible, but you should be aware that this is warranty-voiding
behavior.
The biggest change comes from adjusting the
over_voltage
value, which is the core
voltage for the ARM CPU and GPU in the BCM2835. The possible values for
over_volt
age
run from -16 (0.8 V) to 8 (1.4 V), with default value at 0 (1.2 V). Each integer above
(or below) 0 steps the voltage by 0.025 V. You cannot go over 6 without also setting
force_turbo=1
(note that this will probably trip the “warranty voided fuse”).
The
over_voltage
configuration setting is a
super-setting
; changing it applies the value
to the
over_voltage_sdram_c
(SDRAM controller voltage),
over_voltage_sdram_i
(SDRAM I/O voltage), and
over_voltage_sdram_p
(SDRAM physical voltage) settings.
It is possible to set those settings independently, but you are far more likely to get
them wrong (or mismatched) and end up with memory corruption, so we strongly
recommend that you use the
over_voltate
super-setting instead.
If you decide to overvolt, just set these configuration options in
/boot/config.txt
, and
then reboot.
When you're overvolting (or overclocking as well), monitoring the voltage levels of the
components you've bumped up suddenly makes more sense. These methods can
nudge out a tiny bit more performance from the hardware, but you're trading that
extra bit of performance for a reduction in hardware lifetime (and possibly stability as
well).
