are you guys using a repurposed usb power supply? If so, the power supply might not properly detect the load. Try connecting the d+/d- line on the cable leading to the power supply, on some supplies this enables high current charging mode(this is how android phones do it. Since not all power supplies handle this equally, consider shorting them through a resistor(10-50Ohms). Make sure you do NOT short the power lines!
Rock 5B power through GPIO pins
Hey there, I am powering the Rock 5b through GPIO Pins 4 and 6. It works on my device. But even though I am using a 5V 4A supply (not sure whether those Pins are meant to carry 4A!! at 5V) it (ir)regularly demands more power than the supply delivers, which results in a sudden restart. (no nvme, only sd-card + usb keyboard + usb mouse + hdmi & usb powering a small touchscreen)
If I power the touchscreen separately there are less of those sudden restarts. Also setting the governor to performance seems to be more stable paradoxically (maybe the power supply I use can’t ramp up as fast as necessary. The power issues got worse since I switched to faster settings for the on_demand governor)
It seems that for these settings from Thomas Kaiser I would need a better power supply or less load / more headroom on the supply. ( https://github.com/ThomasKaiser/Knowledge/blob/master/articles/Quick_Preview_of_ROCK_5B.md#important-insights-and-suggested-optimisations )
/etc/sysfs.d/tk-optimize-rk3588.conf:
devices/platform/dmc/devfreq/dmc/upthreshold = 25
module/pcie_aspm/parameters/policy = default
devices/system/cpu/cpufreq/policy0/ondemand/io_is_busy = 1
devices/system/cpu/cpufreq/policy4/ondemand/io_is_busy = 1
devices/system/cpu/cpufreq/policy6/ondemand/io_is_busy = 1
devices/system/cpu/cpufreq/policy0/ondemand/up_threshold = 25
devices/system/cpu/cpufreq/policy4/ondemand/up_threshold = 25
devices/system/cpu/cpufreq/policy6/ondemand/up_threshold = 25
devices/system/cpu/cpufreq/policy0/ondemand/sampling_down_factor = 10
devices/system/cpu/cpufreq/policy4/ondemand/sampling_down_factor = 10
devices/system/cpu/cpufreq/policy6/ondemand/sampling_down_factor = 10
devices/system/cpu/cpufreq/policy0/ondemand/sampling_rate = 200000
devices/system/cpu/cpufreq/policy4/ondemand/sampling_rate = 200000
devices/system/cpu/cpufreq/policy6/ondemand/sampling_rate = 200000
My findings from running the Rockpi 5b over GPIOs:
If you power the board through USB froma sufficient charger and then switch over to GPIO power it works fine.I am guesing there some software loop depending on some signal from the USB or voltage to boot up.
Hi,can you please explain what the config is for and whats it doing?
The above suggested settings only affect cpufreq behaviour with I/O workloads and have been adopted by Radxa recently.
As for powering using 5V at 4A is nonsense since you get either/or but never both at the same time with this kind of cabling. Ohm’s law still is in place and both cable and contact resistance will result in underpowering (most probably undervoltage).
You can use my sbc-bench to monitor the input voltage in a way similar to here: Realtime power usage
If it drops below 5V already at idle you know your cabling is at fault.
2 Likes
LOL, the guy trying to insult others as ‘turds’ tries to behave this time? 
Since you’re still focused on the irrelevant number (4A) to the basics first: https://en.wikipedia.org/wiki/Ohm’s_law
A 5V/4A power brick can be considered almost a scam since at only 5V cable and contact resistance are that high that you never get 5V at 4A at the device in question. You either get 5V (with low loads) or 4A (with voltage dropped below sane levels) but never both at the same time. There’s a reason why none of the USB-C and USB PD power modes exceeds 3A at 5V: since it can’t work due to resistance too high -> voltage drop.
That’s why the following advise has already been given:
With tiny Dupont wiring and at 5V undervoltage might occur below 2A or even 1A! And undervoltage is defined as the tolerable voltage drop until some protection mechanism kicks in. The buck converter the GPIO pins are fed into is rated at 4.5V - 26V so as soon as the input voltage drops below 4.5V you will see random reboots.
And depending on how long and thin those Dupont cables are and if only two pins are used (contact resistance adds to the problem) at 5V you’ll have almost no chance of stable operation under any significant load since the 5V at the other end of the cable drops way too low on the board.
Is there any moderation on this forum? Users who call others names should at least be warned. Or grow up.
4 Likes
What you do really doesn’t matter at all.
At least you failed the second time in this thread to get the basics and these basics are still Ohm’s law. When powering with tiny cables, connecting with just two tiny contacts and all of this is at a very low voltage like 5V it can’t work. Amperage still doesn’t matter.
@RadxaYuntian for those of your users lacking any understanding of electronics basics (like this guy here who tries to insult others all the time instead of using his brain in a better way) you might want to add a small section to Wiki and product page that explains that 5V with more than 3A doesn’t work well in general and what the challenges are when powering with tiny cables and more than minimum loads.
Dumb people seem to be immune to explanations and seek for authority so it might be the vendor’s job to fix their misunderstandings since otherwise these people flood the forums with even more BS than they already do
1 Like
Hi all,
The issue seems to be with the Radxa OS release instead of voltage drop issues with the supply.
The latest Android 12 OS release provided by Radxa seems to have no boot issues while powering up with a supply or a custom hat with a 5V 4A buck converter.
Power Adapter 5V 4A hope this helps.
To be honest, while I don’t understand tangents about “stupid people”, I agree with @tkaiser at pushing 4A to power the board seems like a bad idea. Even if there are no power issues now, they may let themselves be known as soon as one connects something.
Sure, it is the capacity of the power supply, but ohm’s law holds, so you get a higher resistance, higher energy losses and voltage drops. Also, it seems that this board is prone to transient power spikes and with 5V you’ll have less headroom to deal with them than with a higher voltage (well, unless you use 9V2A for example).
I mean, you get less energy loss when powering with 12V 1.5A than with 5V3.6A, for example, if the board’s power is 18 watt
1 Like
I really don’t know what you mean by “euroturd” and I have no interest.
By “pushing 4A” I meant pushing this kind of current through a cable. Which is one of the key limiting factors here I suppose. As it has a resistance of its own.
The best is to use the USB C power point because it offers also some protection to the board. However, Dupont wires can be used on a Rock 5 board, but it’s important to make sure that the wire gauge is appropriate for the amount of current being carried. Thicker wires (such as 22AWG) can carry more current than thinner wires (such as 26AWG). Check what you have/buy.
In general, a 22AWG Dupont wire can carry up to 2.5 amps of current, while a 26AWG Dupont wire can carry up to 1.5 amps of current. However, it’s always best to consult the specifications of the specific wire you’re using to ensure it can handle the current you need. The above is valid for a short wire (150mm).
It’s also important to ensure that the Dupont connector is properly connected and not loose on the Rock pin, as this can lead to resistance and overheating. Again ohm law applies here.
If you’re experiencing current issues, it may be helpful to connect multiple pins in parallel to increase the current-carrying capacity. For example, you could connect pins 2 and 4 in parallel for positive, and connect two ground pins (such as 34 and 39) in parallel as well. This will offer a larger cross-sectional area for the wires and reduce resistance, allowing for more current to flow without overheating. As said in the above posts, ohm law applies, thicker is the wire, more current you can carry. However you are still limited by the Dupoint connector and the size of the pin/copper on the board.
The above has been tested and works.
Recently I am using an old laptop power brick with 19V 4A to power the rock and it runs stable. But unfortunately the usb-c port is blocked now.
And well … I just found out that it seems to be possible to power the rock with 5V 6A via GPIO:
@RadxaYuntian you should really spend some time/efforts on providing some basic knowledge about the relationship of voltage, current and resistance somewhere in your wiki since what happens here in the forums is just nuts.
Your consumers should be educated about these simple facts but maybe this already happens somewhere in a hidden/smelly corner of Discord or Telegram or wherever else community fragmentation happens…
Folks powering through GPIO pins experience undervoltage issues due to Ohm’s law.
This is cable + contact resistance.
5V at any current higher than 3A is just… BS!
2 Likes
As I said in my previous post. Use a 150 mm AWG 22, double it using the pin in parallel and the voltage drop will cause the 5 volts still to be usable with a minimum loss when draining more than 3 A. However I doubt you will get 6A out of it unless you put in in the freezer 
In your previous post you wrote about a ‘current issue’ while it’s a voltage/resistance issue in reality.
And for consumers using SBC it’s crucial to understand this since they think 5V would be 5V regardless of situation and if they just use a beefier power brick with more amps it will work in the end.
As explained already a month ago the GPIO power pins are routed to a buck converter that’s rated 4.5V-26V as such the tolerable voltage drop is just 0.5V.
And of course what you wrote (the shorter and thicker the cable the better, the more contact area the better) is true but only since it reduces cable/contact resistance and helps with the voltage issue everything powered with such low voltages as 5V is affected by.