2g on the asus starts fluctuating a bit wildy if i set it down to 17dBm to match the aw7916. test isn’t very clean, i am going through atleast 3 walls, and standing maybe 30 or 40m from where the routers are. with tons of interference. 
thank you for making that test. Now it is more clean for others to understand and compare results.
May the gods of irony be damned, there was a brief power failure here at home, the R4 got stuck on the boot prompt. For an internet gateway, this is unacceptable. Well, time to find a replacement router.
Now back on topic, via Wifi, on openwrt, ssh gets very slow sometimes. Respective bug
Got similar - when it crash it stuck on boot prompt or it goes to recovery mode. That’s why I set in NAND to reboot into “production system”
fw_printenv | grep bootmenu
From the list, I pick:
(...)
bootmenu_2=Boot production system from NAND.=run boot_production ; run bootmenu_confirm_return
(...)
bootmenu_default=0
To set default:
fw_setenv bootmenu_default 2
To verify:
fw_printenv | grep bootmenu_default
bootmenu_default=2
So many times I agree with you @Casulo - it is just a router that can not trust fully. Unexpected reboots, it did not return back to main system. The whole situation with 2.5G PoE version…
I have wondered this a bit, with the “official” case;
The basis of the case might be grounded from the board on attachment screws, but what about the slideable top plate? Because of surface treatment, doesn’t look like it would be electrically connected to the rest of the case.
Before I go ahead and cook up a threaded “corner block” inside the case - and associated counter-sunk screw holes on the frame side and top cover plate - has anyone already tried if it’d make any change grounding the top lid plate properly?
@frank-w Maybe this can be the default on your image?
Anyway, i got this:
root@banana1 /root $ fw_printenv
Configuration file wrong or corrupted
My images using uEnv.txt which imho is better than the binary file/partitions used with printenv/setenv
So, how to avoid being stuck on boot prompt after a crash?
Why should you hang at a bootprompt because of environment in file after crash?
No idea, but that’s what happened here.
Can we keep this thread on topic please, this is about WiFi signal issues
2 Likes
if you have that error, it seems you restored backup made on SDCARD and restored on NAND or opposite etc., right? Normally, on NAND it should be like:
root@BPI-R4:~# cat /etc/config/ubootenv
config ubootenv
option dev '/dev/ubi0_1'
option offset '0x0'
option envsize '0x1f000'
option secsize '0x1f000'
option numsec '1'
config ubootenv
option dev '/dev/ubi0_2'
option offset '0x0'
option envsize '0x1f000'
option secsize '0x1f000'
option numsec '1'
or
root@BPI-R4:~# cat /etc/config/ubihealthd
config ubi-device 'ubi0'
option device '/dev/ubi0'
option enable '1'
Can not do what is on sdcard/emmc now.
I’m using Frank images and kernel. But i do have another R4 serving as a secondary AP running Openwrt. The gateway is the most important one tough and it failed.
Start a new topic and share some boot/uart log.
1 Like
Tried this, no difference in signal, noise, temperature, iperf3, etc. Only difference was power decreasing by 1 watt. 10 vs 9.
I think 10 watts is high for this kind of device.
1 Fan
1 BE14 with 6 antennas, 2.4G and 5G enabled, 6G disabled.
1 SFP
I got a mini pc:
Intel 1340p
2 NVME disks
64GB of RAM (2 sticks)
2 intel 2.5g nics (both connected)
Running proxmox, 1 windows VM, 1 debian VM, both idle.
8-11 watts
The BPI-R4 is indeed quite power-hungry. My NAS, a CWWK U300E with dual 10GbE ports and 4 NVMe drives running in PCIe 3.0 4.0 x4 mode, consumes around 17W when idle or under low-load conditions.
For me the ASPM activation only made a difference using the offical OpenWRT build with their mt76 wifi driver. Running OpenWRT with the MTK feed and their mediatek driver, ASPM activation had no effect in noise and power consumption. This is one of the reasons I stick to the offical OpenWRT, despite the MTK feed enables Wifi7 with 320mhz support on the 6G band - wich is not working on the offical OpenWRT build.
Here we go!
This is my R4 together with my Flint 2 that acts as an access point:
Here I show the thermal pads (Arctic TP-3) of the AsiaRF miniPCIe cards (below. Above is a Quectel RM520N-GL modem):
And here the tests:
GL-iNet Flint 2 4x4:4 Wi-Fi 6 (Mediatek):
AsiaRF AW7915-NP1 4x4:4 Wi-Fi 6 (Mediatek):
The tests have been performed with a 1000/500 Mb/s GPON connection and an iPhone 15 Pro Max. Wi-Fi in the 5GHz band, channel 36, 80MHz (no DFS), 23dBm (ES). The distance is less than 10 meters with only one separation door (living room->room) in an environment free of interference and with the same snapshot version of OpenWrt. The power adapter used is a Chicony 19V 3.42A (65W).
I tell you, with the AsiaRF, the performance is very mediocre, but above all, unstable. It degrades very easily with distance or varying the position in the same place.
Let’s see how you and those who have the BE14 see it. I have a doubt whether a BE14 with the correct EEPROM would be able to perform better…
bro, sadly that’s true 
, ther other one have okey wifi (but only 1Gbps peak) and broken 2.5GbE.
details here: Banana Pi BPI-BE1900 Wifi7 Module design - #45 by tutugreen
that is their first try to make a wifi router…