Armbian images are now available for Rock 5b!

[fraz0815]

@whitefox

I ran into same problems. When 'sensors' reports only 

in0: 5.00 V (min = +5.00 V, max = +5.00 V)
curr1: 0.00 A (max = +0.00 A)

emmc + nvme stopped working reliable together.

 

I solved most of my problems with buying an new PSU AND a certified 60W USB Type C to Type C Cable.

Random old Type C to Type C cable was unreliable. Some old non-Type C 5V/2A PSU did work, but only as long as i did not use emmc, sd + nvme at the same time, sensors reported same values 

 

Here is a incomplete list about PSU's and their possible boot options: https://wiki.radxa.com/Rock5/5b/power_supply

 

I bought a PSU from ALLNET with: USB-C 5V/3A, 9V/3A, 12V/3A, 15V/3A, 20V/2,25A (45W)

https://shop.maker-store.de/single-board-computer/zubehoer/weiteres-zubehoer/3763/rock-pi-4-zbh.-power-adapter-pd-netzteil-45-watt-eu-plug + 

+ 60W cable

https://www.amazon.de/gp/product/B09YHHNNCG/ref=ppx_yo_dt_b_asin_title_o03_s00?ie=UTF8&psc=1&language=en_GB

 

in0:          20.00 V  (min = +20.00 V, max = +20.00 V)
curr1:         2.25 A  (max =  +2.25 A)

 

So far I am very impressed about performance vs rk3399, this is a beast 🙂

And would also consider the legacy kernel in a better state than expected.

 

[specs]

On 11/22/2022 at 2:58 AM, lurk101 said:

How do I get it to boot in console mode when I mostly don't need the desktop? The usual

Edited Tuesday at 03:04 AM by lurk101

Normally I'd say it is either ubuntu-desktop or ubuntu-minimal, but here my first choice would be removing xfdekstop4.

 

When in doubt, try "dpkg -l | grep desktop"  and remove whatever you'd find, like "apt remove  xfdesktop4".

Then you follow with an "apt autoremove". In this stage read carefully what is to be removed!

[PigLover]

Is there any chance we could get kernel support for open-iscsi built into this kernel (I believe this requires module iscsi_tcp)?  I'm doing some Kubernetes testing and this is required for Longhorn.

 

I know I could build my own kernel and probably will if this is a request that will take a long time.  But I believe this will be a common use case for this board and it would be nice to have it in the base kernels.

 

Thanks!

[Werner]

4 hours ago, PigLover said:

Is there any chance we could get kernel support for open-iscsi built into this kernel (I believe this requires module iscsi_tcp)?  I'm doing some Kubernetes testing and this is required for Longhorn.

 

I know I could build my own kernel and probably will if this is a request that will take a long time.  But I believe this will be a common use case for this board and it would be nice to have it in the base kernels.

If somebody is willing to a) test this scenario and b) then provide a PR that enables the requested module across all configs, then yes, we could ship it by default.

[PigLover]

Thank you.  I built it last evening.  It appears stable, iscsid now starts (fixed the key symptom) and does allow Longhorn to work.  I also included the Ceph kernel modules due to future testing plans.

 

I'd like to run with it for a few days to see if there are any problems and, barring any problems, will submit the PR in a week or so.

[Dann]

Anybody managed to getting Chromium or Firefox to YouTube over GPU hardware accelerated?

[williamv]

hi, got production Rock5b with 16GB, passive cooling, Rock A8 wifi6 card, Jammy with Cinnamon desktop, updated SPI so have SSD boot now, nice. Thank you Armbian team!

Even with hardware acceleration sw installed still mis 10% of frames in youtube both Chromium and FF, tab crashes now and then in Firefox.

Wifi works but bluetooth, no devices found ..

Bought new 1TB WD Black SN850X gen4 for euro 105.

@gounthar looks like SN850X is a mini bit faster than @mzbs Samsung 970 EVO Plus.

Looking at my earlier testing is on par with my AMD 2400G and 2-4x faster than my Nanopi4a RK3399, not bad!

 

hdparm -tT --direct /dev/nvme0n1

/dev/nvme0n1:
 Timing O_DIRECT cached reads:   3770 MB in  2.00 seconds = 1886.21 MB/sec
 Timing O_DIRECT disk reads:     5888 MB in  3.00 seconds = 1962.44 MB/sec

 

fio ... (summary)

Run status group 0 (all jobs):
  WRITE: bw=3157MiB/s (3311MB/s), 3157MiB/s-3157MiB/s (3311MB/s-3311MB/s), io=185GiB (199GB), run=60041-60041msec

Disk stats (read/write):
  nvme0n1: ios=0/379492, merge=0/1939, ticks=0/14926632, in_queue=14926644, util=99.81%

 

Edited December 13, 2022 by williamv
typos

[gounthar]

Thank you.

[williamv]

@Dann for HW acceleration instructions see https://www.armbian.com/rock-5b/

[Dann]

@williamvyou mean ppa:liujianfeng1994 and kodi instructions? Done all that. I am talking Chromium YouTube which remains software/CPU rendering. From @Gnattu https://forum.radxa.com/t/armbian-images-are-now-available-for-rock-5b/11896/109 we get:

 

To sum up what you will need to perform hardware decoding using chromium from scratch:

• Rockchip’s mpp library for the userspace api to access rockchip’s media processing blocks. Look for packages contains librockchip-mpp, librockchip-vpu and rockchip-mpp.
• Rockchip’s libv4l as a v4l2 plugin so that chromium can use it to access the mpp library. Look for packages contains libv4l and v4l2. You probably need v4l-utils as well.
• The patched chromium itself. You could find an x11 version in radxa repo.

I have yet to succeed, though

[Dann]

Anybody playing with Steam able to getting it to work? Either native ARM or over box64/86 or over wine, even?

[fraz0815]

When adding the ppa's for 3D accleration, there a more steps involved, which are printed out once and can be missed.
So far desktop and chromium are running great, even wayland.

Spoiler

You have to create the following udev rules to enable mpp and rga hardware acceleration:

 

KERNEL=="mpp_service", MODE="0660", GROUP="video"
KERNEL=="rga", MODE="0660", GROUP="video"
KERNEL=="system-dma32", MODE="0666", GROUP="video"
KERNEL=="system-uncached-dma32", MODE="0666", GROUP="video" RUN+="/usr/bin/chmod a+rw /dev/dma_heap"

 

To enable libv4l-rkmpp for chromium, you have to install libv4l-rkmpp, v4l-utils and chromium-browser in this ppa:

sudo apt install chromium-browser=$(apt-cache show chromium-browser|grep Version|grep rkmpp|cut -d " " -f2) chromium-codecs-ffmpeg-extra=$(apt-cache show chromium-browser|grep Version|grep rkmpp|cut -d " " -f2) libv4l-rkmpp, v4l-utils

 

Then run the following commands:

sudo ln -s /lib /usr/lib64
cd /usr/lib64/ && sudo ln -s aarch64-linux-gnu/libv4l2.so.0.0.0 libv4l2.so

 

And add the following lines to /etc/rc.local:

echo dec > /dev/video-dec0
chown root:video /dev/video-dec0
chmod 0660 /dev/video-dec0
echo enc > /dev/video-enc0
chown root:video /dev/video-enc0
chmod 0660 /dev/video-enc0

 

Add chrome flags "--use-gl=egl" to /etc/chromium-browser/default:

CHROMIUM_FLAGS="--use-gl=egl"

 

Edited December 14, 2022 by fraz0815

[archetech]

fraz  you should check your journal log when running wayland.  I always get yellow kernel warnings.  I get total freezes or just a screen with a cursor on KDE.  Gnome seems to handle wayland better than KDE at this point of panfork dev.

Edited December 15, 2022 by archetech

[fraz0815]

Of course it's far from perfect, but can be used. I remember the start of rk3399 boards, that was pretty bad in comparison to today.

ARM64 and panfrost really made some serious progress.

 

JFYI, wayland progress on gnome was always more advanced than KDE's (and only choice if you wanted working wayland DE besides sway etc)

[Dann]

@fraz0815Thanks pointing out instructions which I followed and repeatef with care. The last step --use-gl=egl makes overall browser experience (even mouse pointer movement) unbearably slow and I am forced to run without this flag. Yet I don’t believe YouTube video is hardware accelerated. The video decoders show up under about:gpu, yet, Developer Tools Media still says hardware decoding NO. Also, CPU usage is significant. I don’t have dropped frames as CPU powerful enough, yet, I don’t believe it’s GPU accelerated. I don’t see any obvious debugging messages other than loads of Vulkan extension failures

[Dann]

I agree Armbian on Rock 5B is highly useable. Yet is all non-mainline: Android, Linux and Mesa (Panfork not Panfrost). That's worrying as it's hard to be confident we'll see proper bug free support any time soon. For me Panfrost (not Panfork) with Vulkan would be priority yet I think mainline kernel is also necessary 

[PigLover]

On 11/28/2022 at 10:43 PM, Werner said:

On 11/28/2022 at 5:54 PM, PigLover said:

Is there any chance we could get kernel support for open-iscsi built into this kernel (I believe this requires module iscsi_tcp)?  I'm doing some Kubernetes testing and this is required for Longhorn.

 

I know I could build my own kernel and probably will if this is a request that will take a long time.  But I believe this will be a common use case for this board and it would be nice to have it in the base kernels.

If somebody is willing to a) test this scenario and b) then provide a PR that enables the requested module across all configs, then yes, we could ship it by default.

Now that the crazyness of the last few weeks have settled down I can finally pay attention to this again.

Is there any chance you could point me to an example of a good/successful PR to enable a feature that I could use to model against?  Would rather get it right the first time rather than wasting time doing it wrong, responding to PR comments, doing it again, rinse, repeat...

[PeteL]

I start Chrome/Vivaldi with these switches.   No anti-aliasing, but the browser runs really well.
--enable-features=UseOzonePlatform --ozone-platform=x11 --use-gl=egl

 

If I want to get anti-aliasing working properly, this compromise works with a slight performance hit.
--enable-features=UseOzonePlatform --ozone-platform=x11 --use-gl=egl --disable-gpu-compositing

 

[Werner]

8 hours ago, PigLover said:

Is there any chance you could point me to an example of a good/successful PR to enable a feature that I could use to model against? 

Kind a like this: https://github.com/armbian/build/pull/4611

 

 

 

[PigLover]

16 hours ago, Werner said:

On 1/5/2023 at 2:01 PM, PigLover said:

Is there any chance you could point me to an example of a good/successful PR to enable a feature that I could use to model against? 

Kind a like this: https://github.com/armbian/build/pull/4611

 

Guess I was a bit slow to the punch - but at least somebody else was on top of it!

[Marco Schirrmeister]

Is there a list with limitations even there are now supported images available for the Rock 5?

ZFS from the Armbian repo is not working on the Jammy images.

[Igor]

7 hours ago, Marco Schirrmeister said:

ZFS from the Armbian repo is not working on the Jammy images.

I can assure you ZFS from Arbmian repo works perfectly fine. We are using it on two production grade servers.
 

odroidhc4:~:# modinfo zfs
filename:       /lib/modules/6.0.12-meson64/updates/dkms/zfs.ko.xz
version:        2.1.6-0york1~22.04

 

[Marco Schirrmeister]

For Odroid yes. I have it as well running fine on a HC4.

But this thread is about Rock 5 and I see problems on the Rock5 (rk3588). Thats why I wanted to check if there is a list of known issues or limitations, before going to deep into the weeds.

 

[Vitalicus]

I downloaded Lunar version for 5B. Armbian_23.04_Rock-5b_lunar_current_6.2.0_gnome_desktop.img.xz 

On SD It seems to work ok, but I can't start it from NVMe. 
rkspi_loader.img  not work with this version?

Edited May 23, 2023 by Vitalicus

[Marco Schirrmeister]

Looks like you use an image from balbes150. You may want to ask in this thread here, which is about his images.

 

But before asking, read the important comments a few times.