A couple of months ago, I replaced this credit card sized computer "Banana Pi M64" because of persisting instability issues (random restarts) and replaced it with a Raspberry Pi 4B which is solid rock stable.
Before binning the old Banana Pi M64 completely, I wanted to investigate what was the problem. A stress test (stress --cpu 4) usually restarted it like within 5 seconds. What was worse, also rendering some webpages in Chromium and some other "normal" user activities.

1. Replacing the OS with another one (completely different kernel) didn't help
2. Attaching a heat sink (a small piece cut from an old Pentium II heat sink - see the attachment) put the CPU temp some 15° or 20°C down but it also didn't help
3. Replacing the power supply had no effect
4. Setting the CPU scaling governor to "on demand" policy and setting the upper bound frequency like 3 steps below the maximum 1152 MHz quite helped but the problem didn't disappear completely. And it was lazier of course.

The solution has been a bit strange. The preferred and only supported powering method is the dedicated round connector. Powering it through OTG USB isn't recommended - or shouldn't even work. However attaching a 5V/2.1A mobile phone recharger to that OTG USB port ensures flawless operation even at the maximum frequency. Tried long lasting stress tests and no more restarts since the different powering connector choice. There must be something wrong with the recommended round connector.
At the moment I'm using it as a "desktop" computer (it indeed rests on my desk with latest ARM64 Armbian OS and FVWM2 window environment (customised by me). Surprisingly it can render QHD (2560x1440) resolution (FHD 1920x1080 is the officially supported maximum) and it's fast enough for normal home tasks (browsing the internet, writing documents, editing photographs...). I haven't removed the heat sink so the CPU temperature keeps below 50°C (122° F) for most of the time.

Would you like me to link this comment to CPU-World forum there to see what they have to say?

If you think we could find an answer there, Robert...
I think it's just a faulty component in the dedicated power supply connector as powering through the OTG USB power is fine.

The board is quite 'exotic', they even stopped selling this brand here. Compared to Raspberry Pi4B, this Banana Pi M64 is about 50% computational power, but its A53 CPU has (unlike A72 in RPi4B) 'aes' instruction enabled, making this frequently used encryption significantly faster.

There, I posted at https://www.cpu-world.com/forum/viewtopic.php?p=307263#307263

To me it really looks like supply problem, when from one input it works and from the other won't.
Have you inspected the connector in detail, mainly on dirt, corrosion, contact pressure, just all things that use to be responsible for a bad contact...
What voltage is on the "round connector"? Is it 5V or something higher?
If higher, then the converter from that higher voltage down to whatever the board uses is likely bad, when from 5V OTG it is OK.
Then you may take an oscilloscope and look on all the regulated supply branches, how they behave. Mainly looking for ripple, ringing and the level itself, eventually whether it isn't collapsing.
Could be bad solder ball on the power management IC (frequent failure, as these tend to run quite hot) or some auxiliary components around.

Then checking thr power input even from the USB and comparing it with what is to be expected (a known good board would be the best) may help: It could be something is drawing too much current, so the supplies derived from the round connector collapse and the USB just happen to have a bit extra power capability.

Thanks...
The original power supply (recommended by the manufacturer) is 5V/2A. In my experience, some of switching power supplies seem to loose voltage after years of use, so it was one of suspects. I've even cut the cable and soldered a USB connector on the other end to be used with that 5V/2.1A mobile phone recharger but it didn't help. Using the same recharger with a cable attached to OTG made the difference though.
I haven't spotted any corrosion or dirt (it was used in a living room after all) but the insufficient contact pressure could possibly be a problem. I haven't got another round connector with the same diameter at the moment but could try using it when I find one.
The OTG powering is fine, I just found this problem 'strange enough to share' and to learn from.

Hi, maybe is bad usb connector?
I run in similar issues with a bad cable and asus tinkerboard.
Can you try using totally different cable from other manifacturer?
Also, it is a good idea if you can measure 5v voltage on gpio connector.

I use a few Raspberry Pi's myself, and have seen the Banana Pi's for sale when looking for a Raspberry Pi(didn't bother to check them out figuring they might not be fully compatible)

Certainly an odd/interesting issue you have there. One thing I have heard with Raspberry Pi's is that a standard 5v 2a powersupply isn't quite enough, and that you're better going with one that does atleast 2.4a.
Wondering if something like that might be the issue here, and for some reason it only shows going through the round connector? Maybe the traces between CPU and USB are just a bit shorter and/or thicker?

@marcopete87: No, the original power supply cable did not have any USB connector. The problem must be in the round connector either on the board or in the plug at the end of the cable.


Original power supply from manufacturer

@xmaslightguy: Yes, my original Rpi 4B power supply is indeed 5V/3A.
BananaPi was shipped with (optional) 5V/2A supply and my 2.1A mobile/cell phone recharger is definitely enough while (and only while) powering it via USB OTG port.
Yes, it's a Chinese clone with no really functional OS image from the manufacturer, but latest Armbian seems to operate pretty well here. I bought it in Rpi3 times when Rpi3 was 1GB RAM only and this BPi M64 had 2GB memory option available.

Forgot to mention: Connecting a USB device (like a keyboard) while running also restarted it or even switched it off. Also a keyboard with backlighting made it an infinite cycler because the restart was happening during the boot up phase when backligting came on.
No problem with that since powering through USB OTG.

Ok, so i think is likely a problem on connector.
Have you tried checking 5v rail?

Edit: i've read schematic
https://drive.google.com/file/d/0B4PAo2nW2KfnUE5tMGItNkZjMzg/view?usp=sharing&resourcekey=0-RppDw0b9JVcD4iXXGvIQYg

If you power all over OTG (page 4), you will power via VBUS over U6, have you tried using CN1 connector (micro USB) instead of CN2 (round connector)?

https://drive.google.com/file/d/1nRbP1T02KlNZBbTJx4BHW5wVWvQPmDxC/view

VBUS seems to be an alternative to ACIN1 and ACIN2 (page 17 of 2nd datasheet), but it have a double resistance, maybe a different current limit and can be disabled via software.

There, I posted at https://www.cpu-world.com/forum/viewtopic.php?p=307263#307263

Apparently, there was a response from that site Seems that it may be a power supply issue that cause an issues on the said board

Thanks, guys. As already said in earlier posts, there 's absolutely no problem with the power supply, if the board is powered via OTG (micro USB connector).
The same power supply causes problems if it powers the board via the round connector though.

I can perfectly live with OTG powering, no need to fix the problem. Just found it an interesting issue.

I'm happy to hear that you found a workaround with this issue, Adam