royk

While playing a video press the letter "O" to check if you're using hardware or software decoding. For hardware decoding you need to compile Kodi with FFmpeg that supports rkmmp hardware decoding and with gless. For 10 bit decoding you should run Kodi from GBM. You'll also need to change a setting in Kodi under video, set render method to "Direct to plane". So install FFmpeg as described at: https://github.com/nyanmisaka/ffmpeg-rockchip/wiki/Compilation Build Kodi as described at: https://github.com/xbmc/xbmc/blob/master/docs/README.Linux.md#41-configure-build with as cmake command for example: cmake ../kodi -DCMAKE_INSTALL_PREFIX=/usr/local -DCORE_PLATFORM_NAME="wayland gbm" -DAPP_RENDER_SYSTEM=gles

You could check the UUID with blkid /dev/mmcblk0p1: UUID="b23e98d0-700e-4a09-88b6-95675ee524f5" Then check if this is the same in /boot/armbianEnv.txt and /etc/fstab

@Mr.Tree Yes you can find the compiled one also on the OneDrive. If your application checks /sys/kernel/realtime you'll need a small patch from 6.12 which you can also find on the OneDrive. But it needs the following config: - CRASH_DUMP [=y] && ARCH_SUPPORTS_CRASH_DUMP [=y] && KEXEC_CORE [=y] - PROC_KCORE [=y] && PROC_FS [=y] && MMU [=y]

@BOFFBOY Just boot from SD, enter in terminal: sudo armbian-install Choose 1. Boot from MTD flash - System on .... After that it asks you to also install the bootloader to MTD flash and choose yes

@laurentppol It might be that the links have not been correct in /boot cd boot sudo ln -sf uInitrd-6.1.75-rt23 uInitrd sudo ln -sf vmlinuz-6.1.75-rt23 Image And of course the dtb from the 6.1.75 kernel must be used. Perhaps downgrading the kernel also helps, in case you don't use real-time applications that might be the better option: https://forum.armbian.com/topic/47090-radxa-rock5c-pentahat-and-emmc/#findComment-206251

- Using the panthor driver might help, the 6.1.75 kernel with the panthor overlay enabled. - If Xorg is the trouble maker, try wayland with Sunshine server and moonlight-qt as client. - Disable hw acceleration in de xorg config file, although here it never crashed for months

@Inis Funny that it's likely caused by a "fix" of Radxa themselves for the 5 itx board. But great that you found the problem

If it's not the u-boot you could check for differences between the boot directory of the sd-card and emmc with: diff /dir-of-sd/boot /dir-of-emmc/boot

@Inis I think this should work: Follow nr 2 and use the device name of your emmc in case it's different. To be sure since it works correctly when you boot from SD, you didn't update the kernel after installing to emmc?

@Inis If that's the case you also won't be able to see the emmc memory and pentaHat at the same time after booting from SD From what I can see is that SPI flash and emmc share the same pins, so the bootloader should be written to the emmc

Probably the bootloader, when you boot from SD it uses the bootloader from the SD and when you boot from emmc it loads the bootloader from spi flash or emmc. So update the bootloader and see if it helps

Which kernel? The 5.10 kernel has known issues with USB audio, the 6.1 vendor kernel should work correctly. Further a realtime kernel might help and set pulseaudio at rt priority.

@laurentppol Perhaps you've tried the 6.1.75 legacy kernel which doesn't work with nvme on the Orange Pi 5 like Igor mentioned? Here I have a RT patched kernel with that commit reverted which works with nvme, https://drive.google.com/drive/folders/1r76sUsfG_F8pq0pkzzPgEyzqdRG6fGcz?usp=drive_link The current image should be with a kernel from before that commit but I cannot confirm that.

@Pine Li No I didn't, perhaps it's without thermal limit and it overheats? I had it running continuously for months without issues. Isolating core 5 and 7 and the realtime task on core 7 gave the best result with LinuxCNC. With the 6.1.75 kernel only isolating core 7 is sufficient. I think that the 6.1.75 kernel is better either way. For USB audio the 5.10 kernel gives a lot of (known) problems too. In case you had problems with booting from nvme with the 6.1.75 kernel, Igor pointed to a (revert) commit that created that problem, removing that solves it.