Help for choosing a compatible power supply

I ordered VisionFive V2 board via Kickstarter. Upon my query, I was notified that a power supply will not be included but I was informed about the specifications of a power supply.

(Added the screenshot because I can’t find a way to link to this message but this is from the Kickstarter comments.)

According to this message, the any power supply that outputs 5V/3A, 9V/3A, 12V/3A should work without problems.

But the VisionFive V2 Datasheet (Section 4.1), it says that the power requirements are 9V/2A, 12V/2A, 15V/2A, 20V/2A.

I now assume that the power requirements mentioned in the VisionFive V2 Datasheet is what I should follow.

So once my board arrives, I will get a power supply that adheres to the mentioned specifications. Is there a way I will be able to test if the board is receiving enough power?

I am looking for an indication that lets me know if a power supply is problematic like the warning shown on the Raspberry Pi.

I’m using this power supply for my VisionFive 1 right now: CanaKit 3.5A Raspberry Pi 4 Power Supply (USB-C)

I also checked the specifications for the VF 1 to see if it had that same discrepancy the VF 2 has, and it does. Even though it does not say that it supports 5V, the power supply I linked only supports that and it works perfectly. Therefore, I would expect it to work for the VF 2 as well.

Hey @bucketoffun thanks for replying! Good to know that the 5V support is present. I assumed it was since it is the lowest common denominator in most power supplies.

Are you aware of something similar to this that is present on the Raspberry Pi?

That is typically a feature of the PMIC (Power Management Integrated Circuit), most modern PMIC chips should have that feature, basically in addition to their primary function they have a multiplexed ADC which the host can read voltages via I2C.

e.g. Dialog Semiconductor DA9063-A was used in the Hifive Unmatched board, which contained four SiFive U74 cores and one SiFive S7 core in it’s SoC, a SiFive Freedom U740 (similar in some ways to a JH7110, but with far fewer features and only clocked at 1.2GHz). A DA9063-A can provide 12 A continuous output from 0.3 V to 3.3 V (3.6 Watt to 39.6 Watt), and it has a 10-bit ADC with nine channels and configurable alarm thresholds. I’m sure that a similar but different part was used on the VisionFive 2 board, but none of the images online are of a high enough resolution to discern the part number. The RPi hardware currently uses a custom made PMIC, a MXL7704 which contains a multiplexed ADC to return voltages via I2C.

The bit that might be missing is the software, but my guess would be that the hardware probably has that feature even if not currently being accessed. The way to know for sure would be to find out the exact part number and read the datasheet programming registers. The PMIC chip is on the right hand side on the top surface of the board, about half way between the USB-C power port and the two Ethernet ports.

PMIC on Star64 (which I’d bet comes from a reference design from StarFive and is either a descendent or a sibling to VF2 - but that’s only a guess) is AXP15060.

The code that manages this thing on startup is pretty gnarly just because that 12V input into a load of different voltage/current combos for all the devices that have to be switched on and off in specific order with specific setup/hold times and individually controllable in the name of power management. The WiFi module, if present, might be later in the power on sequence than eMMC or SD since you need the latter to boot.

To Bucket’s point, I also use the Canaberry supply on VF1. More importantly, I can power it from an Apple PS or Macbook port because it includes the CC1/CC2 pullups that were missing on Pi. (A competing board of similar vintage failed to boot from those same sources, but I didn’t recognize it as “the Pi4 Problem” then.

I haven’t exercised USB-C power on Star64 yet; I’ve been mostly feeding it from the 12V/3A hose, though the board will boot and communicate just from the 5V power on the USB/serial connection. I’ve little doubt that if you lit up all four cores, PCI video card, attached stuff to USB, ignited the various radios, etc. that the board would enter a sad, dark, slumber.

There’s still exploration/documentation to do once we have more working systems.


Thanks, looking at an image of the chip on a VF2 board, I’d say that you are right that it is a AXP15060. I was initially looking at it umop apisdn (upside down) and could not make head nor tail of it.

And looking at the datasheet it’s interrupts and registers can detect under voltage on the outputs of the six DC2DC buck converters. The input voltage is not monitored, but if any of the outputs from the DC2DC converters are low it would be a strong indication of a problem with the power supply (or a fault after the buck converter).


Can anyone confirm booting from a higher voltage usb-c. I’m using a raspberry pi 3.5A 5V adapter and the visionfive2 boots; but doesn’t do a great deal. Wondering if the voltage could have anything to do with that? I have alternative USB-C power supplies for my project voltera and another for my Rock Five B.

You can get more information about VisionFive 2 power supply in the VisionFive2_DatasheetSheet.



I’m about 95% sure that it will never request > 5V via USB power delivery. I’ve had mine (OK, a Star64) on an analyzer and never seen it, but I haven’t been able to light up all four cores because we still don’t really have a working distro for them. A single core, without a PCI add-in, just doesn’t need a lot of power.

Define “doesn’t do a great deal” and how you expect potentially over-volting it to help. The rules of USB-C power delivery say that if a device WANTS more than 5V, it has to ask the power supply to ratchet it up after the 5V connection is established.

1 Like