hexdump

@Ruediger - do you really see 4gb ram? i'm asking as the allwinner h6 can only properly address up to 3gb, so usally all 4gb h6 boxes only have 3gb useable in reality ... regarding wlan i cannot say much, but there seem to be different versions of the tx6 out with different wlan chips - https://linux-sunxi.org/Tanix_TX6

the hdmi support on the amlogic s8xx socs is still in its very early stages, so things like this might happen - maybe @balbes150 has some more details or ideas?

you might have a look at https://github.com/mripard/sunxi-mali for the legacy mali blob solution on mainline or at https://gitlab.freedesktop.org/lima/web/blob/master/README.md for the future ready open source mali/mesa gpu driver which is included in the latest mainline kernels and mesa releases and works already well for some things but not for all yet ... but do not expect any wonders on this hardware, especially running on x11 ...

it is sometimes not shown if you have a too small screen - for instance i have this on my thinkpad x61s with its 1024x768 screen - which is my everyday machine ... what helped was to zoom out in firefox (something like 80%) for posting in that case ...

@RMG - i think this is the wrong thread for this question as those boxes have a rk3328 and no rk3288 ... technically it is possible but it is very far from easy as you have to build a custom u-boot and write it in a complicated way (as you only write part of it) using the rockchip tool via usb to emmc and even then it is not guaranteed to work - at least i have a h96 max h2 which is now booting following the extlinux.conf file from either usb and sd card but it was a long journey to get there ... i have it on my todo list to document this procedure, but i'm not sure when i'll find the time for this ...

@s812us - i think "apt-get install build-essential" should give you all you need and if still something is missing, the error messages should tell you what

@Nuno Cruz - i just found an old note from me regarding this box: "serial port might be the three test points next to each other with two resistors - yes it is :)" - maybe it helps you finding the serial port - i do not have mine here right now ... in order to get anything going it is required to get it recognized by the rkdeveloptool again ... iirc there were even two test points on the backside of or near the emmc chip to put it into maskrom mode (i think one of them had to be grounded, but i might be wrong) good luck - hexdump

i think @balbes150 was experimenting with usb booting on the h6 iirc - so maybe he would be interested ...

@s812us - you might have a look at this dts file: https://github.com/hexdump0815/linux-mainline-and-mali-amlogic-kernel/blob/master/misc.m8x/dtb/meson8m2-m8s-dvfs.dts - it is basically an m8s dts with frequency scaling added - the one in the kernel tree only has this one fixed frequency of 1.2ghz ... the dts is for v5.6 - for the latest kernel from martin (which balbes150 is using for his images i think) you may try https://github.com/hexdump0815/linux-mainline-meson8-kernel/blob/master/misc.m8x/dtb/meson8m2-m8s-dvfs.dts - in both repos you can also find a file called readme.m8x which should give you enough information about how to compile a kernel (you might just need to adjust some paths maybe) good luck and best wishes - hexdump

@Beppe - there seem to be boxes which are configured in a way that they simply do not boot from sd card at all and i guess yours is one of them ... it is technically possible to get them working by building a special second stage u-boot which then boots from sd/usb and write it to the right position on emmc, but this is far from trivial and often needs to be adjusted for each box (i did it for a h96max rk3328 h2 which also does not boot from sd by default) - as a result i would recommend you to get another box if that is an option for you and better avoid rk3318 at all as those are really lowest end tv boxes and not really easy to deal with ... good luck and best wishes - hexdump

looks to me like it initialized properly but only has 2pp cores instead of 4 defined in the dtb maybe?

for some real world workloads i already saw better performance of an 1.8ghz s905x2 than of an 2ghz s905x3 although the benchmark numbers look different

the actual blender requires opengl 3.3 which is not supported by any arm soc (except maybe the nvidia jetson if you want to count it in) - otherwise you will be limited to blender 2.79 ...

@WanJiXianSheng - then it can be that the box simply only has 1gb of ram although it is sold as 2gb box - see

[note to the moderators: this should maybe split into its own thread?] @WanJiXianSheng - decompile your dtb with "dtc -I dtb -O dts meson-gxl-s905w-tx3-mini.dtb > meson-gxl-s905w-tx3-mini.dts" and then edit the dts file: change "reg = <0x0 0x0 0x0 0x40000000>;" in the memory@0 section to "reg = <0x0 0x0 0x0 0x80000000>;" and compile it back into a dtb: "dtc -I dts -O dtb meson-gxl-s905w-tx3-mini.dts > meson-gxl-s905w-tx3-mini.dtb" good luck - hexdump

by chance i did experiment with panfrost on s912 right now using a 5.6.2 kernel (with panfrost as a module) and 20.0.4 mesa (all compiled myself) and ran into problems as well - what made it work in the end for me was to compile a fresh libdrm (https://gitlab.freedesktop.org/mesa/drm) and use the xorg.conf from the lima project (https://gitlab.freedesktop.org/lima/web/blob/master/README.md) with "meson" as display drm driver. everything else is regular ubuntu 18.04 and now it works very well. good luck and best wishes - hexdump

this dtb is very old and will most probably no longer work with the latest images - just try all the h6 dtbs which are part of the images - the qplus is quite a standard box, so it should work with one of them, but on the other side the qplus hardware seems to vary quite a bit so it might also not work at all due to memory initializing problems or something like this good luck and best wishes - hexdump

just a guess: maybe the gpu is enabled now due to the panfrost driver? but it miht be a wrong guess of course ...

@jock - that approach with erasing the emmc might work quite well for some boxes, but not for others: i have an rk3318 box where i was not able to get mainline u-boot+atf running at all and was relying on the first stage of the original bootloader ... there are allwinner boxes where mainline u-boot is not able to get the memory timing right - the only option there is to build a hacked non-redistributable u-boot using an allwinner blob and on amlogic it is even more complex: here u-boot uses binary blobs for the memory timing etc. which are often specific to a certain box, so that you'll have to first extract the original boot blocks from the box, extract those sections from it and and build a mainline u-boot with them following the for sure not trivial build procedure for the amlogic u-boot - it can be done if one is used to doing such things but i think it will fail as a generic solution for everyone ... best wishes - hexdump

@Miguel Moyano - give this one a try, it should work i think: make sure to read the instructions of the first post carefully to not skip some important step good luck and best wishes - hexdump

@balbes150 - how exactly is this working technically? isn't in theory the mbr part occupied by the bootloader as well on s905? a lot of thanks in advance and best wishes - hexdump

@Italo Felipe - somewhere in this thread i also posted a t9 dtb for mainline, which should work for 5.x - not sure if i already had wifi working with it but i think i had ... good luck and best wishes - hexdump

@Italo Felipe - i think there is a newer version of the t9 box which has a rk3318 instead of the rk3328 in the original one - maybe your box is one of those? i think the old one came with android 8 and the new one with android 9 ... if you have the newer one, you may try the things mentioned in this thread and other on this forum about getting rk3318 boxes working ... good luck - hexdump

i guess this is on s905 - right? this is due to bad design of the s905 - if booting from emmc it expects its boot sectors at the beginning of the disk (compared to if booting from sd card where it expects them to start later) - you can find more information about this from here on in that thread: i think there is no really easy way around it on s905 best wishes - hexdump

@Nuno Cruz - it is sometimes possible to extract some information from the original dtb and use it to adjust a mainline one, but its not easy and not always possible ...