Seth

thanks, will try it out and upload serial logs after i solder the wires on it. got a bit distracted since i'm currently cleaning up the other box from junk and trying to reflash and reinstall pihole.

links on first post.

hello, i got a new rk3328/3318 box and it is an MX10 box. board is MXQ-RK3328-D4_A ver: 2.0 date:20181219. i was wondering if anyone has the same board as mine and got armbian running on it. here are pics of the board. https://photos.app.goo.gl/ikSCNH7s59f2v5EE7 if serial logs are needed i can solder some wires to gather the info. thanks.

works like a dream @jock. here's my freshly installed/updated hk1 max box dmesg using latest libreelec hdmi patched 6.3.13 kernel. i hope mainline gets around to updating hdmi timing stuff. brcm 6334 2.4G and 5G works out of the box. using it currently as pi hole box. for those having problems with rtl 2734c (rtl8237cs). wireless module that gets detected but cannot connect to AP, here's a reuploaded nvram you can try out (no guarantees though), it solved my problem way back with the rounded h96 max i am using as klipper host. file is not mine, credits to the original uploader. dmesg.log nvram_2734c.txt

@FatalWorld there is a kernel oops about swapper being tainted. your board is the same as mine albeit with different wifi chip. these are the things i would try if that was my board: 1. backup original firmware and try flashing different firmwares on the emmc chip from the github page. 2. if running from sd card, i would try a different sd card just to test filesystem stability. 3. i would also try ram testing. 4. upload logs for different test image used. just do a "dmesg >> dmesg519.log", download that log using winscp and include it in your next post, it would help jock and others with debugging where your problem is coming from. got mine solved that way. cheers and good luck!

please indicate box type and board version, if posible attach top and bottom images of the board.

board version? sometimes those 4 pads do not work for me, what i sometimes do when i'm feeling lazy is i try connecting the ground on the board first then i try connecting the rx pin of usb serial to different pads of the board while booting to find its tx pin. you only need two pins to capture serial debug logs. i found mine at the back of the board. i think i posted the pics in this post at page 26. you get very low chances of bricking your board if you do it this way. other method i use is a low cost stm32 o-scope and a multitester which a is way safer method in my opinion.

very strange indeed. i was testing it on a uhd 55" samsung tv that has not been updated for about 2 years, it is plugged in to hdmi ARC port, so that maybe the source of my confusion too. the reason why i bought the same board as yours is to see if they would all behave the same way and indeed they have. i really need to start learning how to compile the kernel for these boxes, i do have experience compiling kernels for linux desktop and server but that was way back when 2.6 was bleeding edge.. dmesg-rk3328yx.log

hello everyone. i just got an hk1 max with yx_rk3328-v1.0 board and tested it with 5.15 kernel and 5.19 images. HDMI works great on 5.15 out of the box, 5.19 on the other hand has no hdmi display. kernel 5.19 with the libreelec patch has hdmi display out of the box.

hello. 5.19.15 kernel works both on rounded and square h96 max box v1.4 boards and so far they are both running stable with hdmi display for about 6 hours now. attached are the dmesg logs for both boxes. for the square box, i did a fresh install of @jock's full image and for the rounded one, i have installed the new kernel and dtb. i2cdetect -l needed sudo and the output is the same as @MattWestB's. dmesg-square.log dmesg-rounded.log

@jock awesome! it worked like a charm, unfortunately i was not able to upload a dmesg log. will upload it tomorrow after i test it on the rounded one too. it runs at 1.3GHz and 76-83 degrees celcius and is very usable as minimal desktop, it even has compositing enabled, with a little bit of active cooling and a better case with more ventilation, it can be nice mini workstation. usb is also stable for both usb2 and usb3, i tested it with usb3 hub and plugged in 4 devices(printer, keyboard, mouse and a usb microscope) and they all worked fine.

@jock good news! i tested you edge build of 5.19 kernel and dtb and it booted great with hdmi output on the squared h96 max v1.4 board. i reckon it would work flawlessly on the rounded one too. sorry for the crappy image quality i tested this at work.

@jock the hdmi does not work on both rounded and squared one with 5.19 kernel. date markings are exactly the same "2020/06/29" as well as board version rk3318_v1.4 on both boards. the board version was under the heatsink, i just used my phone's flashlight to see and read the board version. also out of curiousity i opened up my dead tylex v88 mini board and the external crystal is 26MHz and uses a 3R3 inductor, seems the SPxxxx wifi chips all need external crystal and inductors. @MattWestB i do not know if i2c is used for communication with hdmi but both boards do behave the same, the both have hdmi output on 5.15 and both have no output on 5.19, the components around the hdmi looks exactly the same but i will take better pictures tonight when i get back home from work.

hello again. @jocki'll be taking pics of both boards tonight and extracting the dtb's. some vendors do give clues as to what wifi chip is placed on the board. my squared h96 max v1.4 board works fine on 2.4GHz and 5GHz and works on default distro of the 5.15 kernel build including hdmi, hdmi does not work on the 5.19 kernel though. my rounded one which i will disassemble seems to have the same board but has no 8 segment lcd. it works almost the same as the squared one with the exception of the nvram that needs replacing. edit: i seem to have identical boards for the rounded one and the square one, both are rk3318 v1.4 dated 2020/06/29. differences are square and rounded respectively. wifi chip: fx8934, sp2734c crystal beside wifi chip: none, 37.4MHz inductor beside wifi chip: none, 4R7 fd6551 and 8 segment: yes, no square pics:https://photos.app.goo.gl/U5LEJrav5FZMh6LEA rounded pics: https://photos.app.goo.gl/dUBqpmVfWpshq4PV9 i think @fabiobassa can chime in on the hardware part because it seems we can put in an fd6551 and an 8 segment on the rounded h96 max box easily and make openvfd work on it too. i also think we can transplant a similar but better wifi chip on it if ever we have those on hand.

there seems to be 2 types of h96 max on the market, one is the more rounded and smaller case which has an rk3318 and 2734c chip that is working only on 2.4GHz and the other one has a bigger case with square corners which has real 4334 wifi chip and supports both 2.4GHz and 5GHz bands. Some square box with lower specs for storage 2GB ram +16GB has 2734c but 4GB+32GB almost always has real 4334 chip. i bought both. the square one is about $3 more expensive than the rounded one. my square h96 max seems to have real rk3328 (verified by using hexdump) and am testing xfce 5.15 version at the moment, hdmi sound doesn't seem to work though. but overall stable at 1.3GHz with a little bit of active cooling. anyways, if you're gonna play with boxes like this, make sure you have an extra cable to the router ready always. my setup includes an always ready rig for uart, usb male to male, ethernet cable and 5v supply ready for quick hookup. the rounded one i have is currently running my 3d printer and will not be updated anytime soon. i'm planning on getting an hk1 max as that is @jock's testing platform, i hope i get lucky and get the same specs.

i got myself a new box, still an h96 max, same as what @paradigman has. it also has a 2734c wifi chip, his posted nvram from the manufacturer works. hdmi on edge kernel still doesn't work on my monitor/tv but i can access it via ssh after install so everything is good. thanks again @jock, @fabiobassa and @paradigmanfor the awesome community support.

yup, i thought mine was an isolated incident but now i saw 2 more cases. anyway, mine is ok now since i installed the weekly build from github 22.08 v5.15 kernel cli. i set cpu speed to 1.3GHz and ram to 660MHz in rk3318-config, so far it has been stable for days, i have the same box case as @williamfj board version 1.4.

this arrived about 5 minutes ago, "out of the box and into the fire pit". board id is MXQ_8058201_v1.3. seems i got an emcp. it came in with the wifi sdio shield removed, i'm guessing they changed the inductor on the lower left of the pic and didn't bother putting the emi shield back on. it's a broadcom 4336 chip the uart is very accessible. will edit later for further findings. https://photos.app.goo.gl/48jMUW5LP239yZdbA

@fukowaka hello, congrats for successful install. emmc optimization is only really to squeeze more performance out of the hardware if it supports it. since our tv boxes are literally frankenstein's monters and are put together from probably used parts or even re-labeled chips, i tend to stick with the the defaults and get it working first to see if it would fit my use case. the default speed of the emmc is already way faster than an sd card i wouldn't worry about it and go with the defaults.

hi @jock thanks again, i can finally do some tests on my scishion since my h96-max box is doing all the heavy lifting on my network and printer now and it's running smoothly. my board has an id MX4M-V11. i tried the led-conf1(mx4vm) and it did not detect wifi either, it works fine with the legacy build i tested about 2 hours ago. i will eventually go full on hardware mod on this and probably transplant an EMMC chip on this board if there is an unpopulated place for it underneath the NAND.

happy to report that the official CSC build from github works flawlessly on my rk3318 box that has ap6334. everything seems to be in order, i did have to rename box64 though to fix the slow dependency building, i also have the smartmontools problem on first boot, but not on the second boot, i have not checked succeeding reboot dmesg though since i had to rebuild my klipper environment. running it at 1.3GHz without any problems so far. uses very little resource when printing, 1-5% cpu usage and 14% ram. 2GB ram and 16G EMMC.

hello again. i downloaded the current 5.15.61 sid build on github for rk322x and tried it on my box which has rtl8723cs chip. the chip is detected by rk322x-config but no wireless interface is shown when i do ip link. looking at dmesg i get dmesg grep rtl8723cs output [ 9.848405] RTL871X: rtl8723cs v4.3.22.9_16440.20151221_BTCOEX20151214-0700 [ 9.848426] RTL871X: rtl8723cs BT-Coex version = BTCOEX20151214-0700 [ 9.848950] rtl8723cs: probe of mmc1:0001:1 failed with error -110 dmesg cfg80211 [ 2.404787] cfg80211: Loading compiled-in X.509 certificates for regulatory database [ 2.451272] cfg80211: Loaded X.509 cert 'sforshee: 00b28ddf47aef9cea7' [ 2.462651] cfg80211: loaded regulatory.db is malformed or signature is missing/invalid i have scishion v88 mini which has NAND so i used multitool to install jumpstart nand and it booted to the sd card. went through all the config just fine except for the wireless. i also attached the complete dmesg log. anyways, i am going to try out 5.15.25 bullseye build from this thread to see if it is going to work. it did work fine with the legacy build though. edit same error with 5.15.25 build. dmesg.log

same as my experience with 5.18 kernel except i didn't build armbian from source. thanks for the case @Aapo Tahkola will try it out. i got an rk3228 box running my printer at the moment occasionally testing rk3318 box as well, they both seem fine and on average use very little cpu and ram usage during printing, it seems these boxes can handle 2-3 simultaneous instances of klipper at the same time maybe even 4 with your case design.

Hello again. manual upgrade from 5.15 to 5.18 kernel and dtb works flawlessly. everything works except hdmi. im currently running my h96 max box with board version rk3318_v1.4. my procedure was install armbian with kernel 5.15 using multitool and sd card while plugged in to an hdmi monitor/tv since multitool and 5.15 kernel works with hdmi. after configuring with rk3318-config, i plug it in to lan and continue running the box headless, then i manually upgrade to 5.18 kernel and dtb. config was 1.1GHz, ddr and emmc is option 1 (typical speeds), led conf i used is led-conf1 (YK led conf i think). seems to work fine and stable apart from the hdmi. i have attached dmesg for 5.15 and 5.18 logs if anyone can spot differences. dmesg-5.15-kernel.log dmesg-5.18-kernel.log

https://github.com/paolosabatino/armbian-build/tree/rk3318