The Ryzen AI 9 HX 370 on the SER9 is very new processor and needs some very recent parts to get running on Linux. The requirements are: Linux 6.10 or later. Mesa 24.2 or later Linux Firmware 20240811 ONLY (At the time of this writing any newer firmware causes amdgpu to hang). Sadly, manjaro, ubuntu 24.10, bazzite and others that roll the latest stuff come with newer firmware that won’t work with the 890m. You can specify nomodeset for the command line to get an install done, then downgrade to Linux Firmware 20240811. My successful build is using Mageia Cauldron with linux 6.11.4, mesa-24.2.5 and radeon-firmware-20240811. I was able to get the build working with Mageia 9 plus Linux 6.11.4 and mesa 2.4.4 but I did not have 3D acceleration. The firmware tarballs are here: https://git.kernel.org/pub/scm/linux/kernel/git/firmware/linux-firmware.git/

HI there , Thanks for sharing . Please let me know, if you have any query , i am glad to help. Your satisfaction is our top priority !

Same problem here. Tried the latest endeavouros doesn’t boot with the latest release without nomodeset. Also tried installing on a 6900HX on usb drive updated everything moved the usb drive to SER9 and hangs on boot. Is that a bios problem or amdgpu driver problem? And how come AMD doesn’t fix this problem? P.S. also tried ubuntu desktop 24.10 to double check with the same results.

It’s the firmware, if you check the site: https://git.kernel.org/pub/scm/linux/kernel/git/firmware/linux-firmware.git/ You’ll see that there are two firmwares after 20240811 and on both of those the system, or amdgpu more specifically, will hang, prior to 20240811, the GPU won’t initialize but at least you’ll get a text screen or a vesa display at 1024×768. I worked around the problem by looking at how mageia did their firmware and rebuilt the firmware package, in this case radeon-firmware-YYYYMMDD.rpm and used that. I’ve looked for a way to file a bug report with the firmware guys but I couldn’t find any way to do so. I saw that ETA prime and Techtablets on the YouTube were able to get Linux running which sent me down the rabbit hole for chasing the firmware issue. I don’t know about other distributions but you should, hopefully, be able to get an older firmware rpm/deb/apk that has 20240811. And you’re right, AMD should straighten out the firmware problem, pronto.

EndeavourOS is arch based. I installed a fresh copy and downgraded the linux-firmware to the august release and upgraded everything else. This worked for now but it’s a bypass of the problem. It’s not a solution. Let’s hope the next linux-firmware release will solve this problem.

Well at least you have a working system and it verifies that the firmware is the root cause of the problem. Thanks for following through and posting your results.

ximejo HI there You can email us with pictures of BIOS version and SN picture. Here`s how to check your BIOS version: https://mega.nz/#F!yuISGa4I!s1bQQajKwnsEdzjqq4nopQ See if updating BIOS would improve that situation .

I don’t know about uploading a picture but here’s the output from “dmidecode”: SMBIOS 3.7.0 present. Table at 0×78256000. Handle 0×0000, DMI type 0, 26 bytes BIOS Information Vendor: American Megatrends International, LLC. Version: STX.3xx.SER9.V103.P8C0M0C15.54.Link.GID Release Date: 09/10/2024 Address: 0xF0000 Runtime Size: 64 kB ROM Size: 32 MB Characteristics: It would be nice to be able to use a more up to date amdgpu firmware.

support4 I have updated the bios to the latest release from https://dr.bee-link.cn/ . It still doesn’t boot without nomodset. Have you tested the latest ubuntu desktop release (24.10)? Does it boot on your test machine without nomodeset? Is it our problem? or a general problem? What are the steps to solve this issue?

This is my dimdecode output: # dmidecode 3.6 Getting SMBIOS data from sysfs. SMBIOS 3.7.0 present. Table at 0x78256000. Handle 0x0000, DMI type 0, 26 bytes BIOS Information Vendor: American Megatrends International, LLC. Version: STX.3xx.SER9.V103.P8C0M0C15.55.Link.GID.90 Release Date: 09/26/2024 Address: 0xF0000 Runtime Size: 64 kB ROM Size: 32 MB Characteristics: PCI is supported BIOS is upgradeable BIOS shadowing is allowed Boot from CD is supported Selectable boot is supported BIOS ROM is socketed EDD is supported ACPI is supported BIOS boot specification is supported Targeted content distribution is supported UEFI is supported BIOS Revision: 0.0 if(window.hljsLoader && !document.currentScript.parentNode.hasAttribute('data-s9e-livepreview-onupdate')) { window.hljsLoader.highlightBlocks(document.currentScript.parentNode); }

To try to get a little more information about this matter I did the following steps: I removed radeon-firmware (Mageia packages it separately from linux-firmware) I booted my system normally, and got the expected error messages: [Fri Oct 25 08:11:47 2024] amdgpu 0000:65:00.0: amdgpu: Fetched VBIOS from VFCT [Fri Oct 25 08:11:47 2024] amdgpu: ATOM BIOS: 113-STRIXEMU-001 [Fri Oct 25 08:11:47 2024] amdgpu 0000:65:00.0: Direct firmware load for amdgpu/psp_14_0_0_toc.bin failed with error -2 [Fri Oct 25 08:11:47 2024] [drm:amdgpu_device_init.cold [amdgpu]] ERROR early_init of IP block <psp> failed -19 [Fri Oct 25 08:11:47 2024] amdgpu 0000:65:00.0: Direct firmware load for amdgpu/dcn_3_5_dmcub.bin failed with error -2 [Fri Oct 25 08:11:47 2024] [drm:amdgpu_device_init.cold [amdgpu]] ERROR early_init of IP block <dm> failed -19 [Fri Oct 25 08:11:47 2024] amdgpu 0000:65:00.0: Direct firmware load for amdgpu/gc_11_5_0_pfp.bin failed with error -2 [Fri Oct 25 08:11:47 2024] [drm:amdgpu_device_init.cold [amdgpu]] ERROR early_init of IP block <gfx_v11_0> failed -19 [Fri Oct 25 08:11:47 2024] amdgpu 0000:65:00.0: Direct firmware load for amdgpu/sdma_6_1_0.bin failed with error -2 [Fri Oct 25 08:11:47 2024] [drm:amdgpu_device_init.cold [amdgpu]] ERROR early_init of IP block <sdma_v6_0> failed -19 [Fri Oct 25 08:11:47 2024] amdgpu 0000:65:00.0: Direct firmware load for amdgpu/vcn_4_0_5.bin failed with error -2 [Fri Oct 25 08:11:47 2024] [drm:amdgpu_device_init.cold [amdgpu]] ERROR early_init of IP block <vcn_v4_0_5> failed -19 [Fri Oct 25 08:11:47 2024] amdgpu 0000:65:00.0: Direct firmware load for amdgpu/gc_11_5_0_mes_2.bin failed with error -2 [Fri Oct 25 08:11:47 2024] amdgpu 0000:65:00.0: amdgpu: try to fall back to gc_11_5_0_mes.bin [Fri Oct 25 08:11:47 2024] amdgpu 0000:65:00.0: Direct firmware load for amdgpu/gc_11_5_0_mes.bin failed with error -2 [Fri Oct 25 08:11:47 2024] [drm:amdgpu_device_init.cold [amdgpu]] ERROR early_init of IP block <mes_v11_0> failed -19 [Fri Oct 25 08:11:47 2024] amdgpu 0000:65:00.0: amdgpu: VPE: collaborate mode false [Fri Oct 25 08:11:47 2024] amdgpu 0000:65:00.0: amdgpu: Fatal error during GPU init [Fri Oct 25 08:11:47 2024] amdgpu 0000:65:00.0: amdgpu: amdgpu: finishing device. The system then came all the way up into multi-user mode and as expected sddm didn’t start but I could access the system from ssh. I opened two ssh sessions, one I did a “dmesg -Tw” and the other I left for entering commands. Next, I installed the latest radeon-firmware, then did a “rmmod amdgpu” followed by “modprobe amdgpu” at which point the output from the SER9 went black and the following output from “dmesg -Tw” came through from my remote session: [Fri Oct 25 08:14:58 2024] [drm] amdgpu kernel modesetting enabled. [Fri Oct 25 08:14:58 2024] amdgpu: Virtual CRAT table created for CPU [Fri Oct 25 08:14:58 2024] amdgpu: Topology: Add CPU node [Fri Oct 25 08:14:58 2024] [drm] initializing kernel modesetting (IP DISCOVERY 0×1002:0×150E 0×1F66:0×0030 0xC1). [Fri Oct 25 08:14:58 2024] [drm] register mmio base: 0xDC400000 [Fri Oct 25 08:14:58 2024] [drm] register mmio size: 524288 [Fri Oct 25 08:14:58 2024] [drm] add ip block number 0 <soc21_common> [Fri Oct 25 08:14:58 2024] [drm] add ip block number 1 <gmc_v11_0> [Fri Oct 25 08:14:58 2024] [drm] add ip block number 2 <ih_v6_1> [Fri Oct 25 08:14:58 2024] [drm] add ip block number 3 <psp> [Fri Oct 25 08:14:58 2024] [drm] add ip block number 4 <smu> [Fri Oct 25 08:14:58 2024] [drm] add ip block number 5 <dm> [Fri Oct 25 08:14:58 2024] [drm] add ip block number 6 <gfx_v11_0> [Fri Oct 25 08:14:58 2024] [drm] add ip block number 7 <sdma_v6_0> [Fri Oct 25 08:14:58 2024] [drm] add ip block number 8 <vcn_v4_0_5> [Fri Oct 25 08:14:58 2024] [drm] add ip block number 9 <jpeg_v4_0_5> [Fri Oct 25 08:14:58 2024] [drm] add ip block number 10 <mes_v11_0> [Fri Oct 25 08:14:58 2024] [drm] add ip block number 11 <vpe_v6_1> [Fri Oct 25 08:14:58 2024] amdgpu 0000:65:00.0: amdgpu: Fetched VBIOS from VFCT [Fri Oct 25 08:14:58 2024] amdgpu: ATOM BIOS: 113-STRIXEMU-001 [Fri Oct 25 08:14:58 2024] amdgpu 0000:65:00.0: amdgpu: VPE: collaborate mode false [Fri Oct 25 08:14:58 2024] Console: switching to colour dummy device 80×25 [Fri Oct 25 08:14:58 2024] amdgpu 0000:65:00.0: amdgpu: Trusted Memory Zone (TMZ) feature disabled as experimental (default) [Fri Oct 25 08:14:58 2024] [drm] vm size is 262144 GB, 4 levels, block size is 9-bit, fragment size is 9-bit [Fri Oct 25 08:14:58 2024] amdgpu 0000:65:00.0: amdgpu: VRAM: 512M 0×0000008000000000 - 0×000000801FFFFFFF (512M used) [Fri Oct 25 08:14:58 2024] amdgpu 0000:65:00.0: amdgpu: GART: 512M 0×00007FFF00000000 - 0×00007FFF1FFFFFFF [Fri Oct 25 08:14:58 2024] [drm] Detected VRAM RAM=512M, BAR=512M [Fri Oct 25 08:14:58 2024] [drm] RAM width 128bits LPDDR5 [Fri Oct 25 08:14:58 2024] [drm] amdgpu: 512M of VRAM memory ready [Fri Oct 25 08:14:58 2024] [drm] amdgpu: 15550M of GTT memory ready. [Fri Oct 25 08:14:58 2024] [drm] GART: num cpu pages 131072, num gpu pages 131072 [Fri Oct 25 08:14:58 2024] [drm] PCIE GART of 512M enabled (table at 0×0000008000300000). [Fri Oct 25 08:14:58 2024] [drm] Loading DMUB firmware via PSP: version=0×09000500 [Fri Oct 25 08:14:58 2024] [drm] Found VCN firmware Version ENC: 1.22 DEC: 9 VEP: 0 Revision: 1 [Fri Oct 25 08:14:58 2024] amdgpu 0000:65:00.0: amdgpu: reserve 0×8900000 from 0×8010000000 for PSP TMR [Fri Oct 25 08:14:58 2024] amdgpu 0000:65:00.0: amdgpu: RAS: optional ras ta ucode is not available [Fri Oct 25 08:14:58 2024] amdgpu 0000:65:00.0: amdgpu: RAP: optional rap ta ucode is not available [Fri Oct 25 08:14:58 2024] amdgpu 0000:65:00.0: amdgpu: SECUREDISPLAY: securedisplay ta ucode is not available [Fri Oct 25 08:14:58 2024] amdgpu 0000:65:00.0: amdgpu: SMU is initialized successfully! [Fri Oct 25 08:14:58 2024] [drm] Seamless boot condition check passed [Fri Oct 25 08:14:58 2024] [drm] Display Core v3.2.291 initialized on DCN 3.5 [Fri Oct 25 08:14:58 2024] [drm] DP-HDMI FRL PCON supported [Fri Oct 25 08:14:58 2024] [drm] Wait for DMUB auto-load failed: 3 [Fri Oct 25 08:14:58 2024] [drm] DMUB hardware initialized: version=0×09000500 The system remained accessible from ssh but my “modprobe amdgpu” never finished it ran away at 100% CPU. If I boot the system with the newer firmware installed, it eventually comes up, after several minutes, in multi-user mode but the screen output is black and getting in via ssh shows that “udev-worker” is running 100%, X did not initialize. Reverting back to the 20240811 firmware the system comes up in a functional state; I guess we’re very lucky we have 20240811. I don’t know if this is unique to the beelink SER9 but I did see one poster said that he was unable to get any distribution working on a AI HX 370, which I assumed was probably due to the firmware issue. I have seen that the system seems to have another issue: If it’s not kept busy, it spontaneously reboots, I am assuming that it goes into a very low power usage mode that Linux doesn’t handle and causes a reboot. Enough for this post.

Linux on SER9.

ximejo

I think the issue should investigated farther.
Not been able to boot the ubuntu 24.10 iso with 3D acceleration is a serious enough issue that demands official troubleshooting.
Using older Drivers is not an acceptable solution.
I’m not using my SER9 for this reason.
I’m still using my older machine.
If this issue is not resolved I will probably sell it.
Windows is no go for me.

MamaLilla

I think its not an issue with the 890m in general not working with newer firmware packages. Its because Beelink SER9 bios includes AMD binaries/AGESA Version from june 2024. Thats even prior to launch of the 890m. If they would update that to something reasonable it would also work with the latest firmware file for linux. So i guess we have to wait for a new bios.

ximejo

thanks for that.
The problem is that we paid 999 bucks and we cannot use our brand new computer.

gukin

I just successfully installed Ubuntu 24.10 and managed to get full hardware acceleration.

Here are the steps I took:
Boot the ubuntu 24.10 but use safe graphics.
Complete the install
On the first boot, poke the “e” button and add “nomodeset” to the line with “linux”
Poke F10 to boot, you’ll wind up with an 800×600 screen.
Go get the linux-firmware-20240811.tar.gz from the link elsewhere on this page.
Then do this:
tar fxva linux-firmware-20240811.tar.gz linux-firmware-20240811/amdgpu/dcn_3_5_dmcub.bin
sudo cp linux-firmware-20240811/amdgpu/dcn_3_5_dmcub.bin /usr/lib/firmware/amdgpu/

Reboot and you should be in full resolution with hardware acceleration.

I know it’s frustrating having to futz around with this stupid firmware file but even though we paid $1000 US for this computer, mini computers don’t always get the latest or even any BIOS updates. Hopefully MamaLilla is right about it being an AGESA problem and Beelink will get a BIOS update out that fixes the problem.

One of my other mini-computers is a GMKtec K8 and they put out a BIOS update that bricked several units and the AGESA for that system dates back to January and I bought it in May 2024. Fortunately it works fine with any distribution I throw at it. My favorite distribution is Mageia 9 and only their development system works the the SER9 and it’s not terribly stable.

I do hope you can get your system working because it really is a sweetheart of a computer and Strix-Point is a heckuva an APU.

ximejo

Have you tried windows 11 with the latest AMD drivers?
The NPU is missing.
Again you have to install older drivers in order to work.
This is unacceptable for a 1000 dollar machine.
I don’t bother with crippled hardware.
I will wait for a new BIOS.
And the most frustrating part is that beelink support doesn’t bother to post in this thread.
Thanks for your efforts.

gukin

As of November 10 2024, new Linux firmware has been released: linux-firmware-20241110 and like all firmware after the 20240811 release (but not including 20240811), the dcn_3_5_dmcub.bin file DOES NOT work with the SER9.

However, replacing dcn_3_5_dmcub.bin from the 20240811 release allows the SER9 to work correctly with full hardware acceleration.

support1

gukin Hi there,

Where do you get the linux-firmware-20241110 and dcn_3_5_dmcub.bin ?
Please send us a link for download .
We will test it.

gukin

I filed a bug report with Mageia and after confirming that only dcm_3_5_dmcub.bin from 20240811 works with the SER9, they released their firmware with that file and now Mageia Cauldron (eventually Mageia 10) works without issue on the SER9.

Sadly, no other Linux distribution will work on the SER9 unless dcm_3_5_dmcub.bin is replaced with the file from 20240811.

Perhaps Beelink Support could answer the question of whether this is a purely firmware problem or whether it could be related to a BIOS issue?

support7

hello, about this, pls go to linux, thanks

ximejo

support7
Really?
Thank god competition is here and runs linux normally.
Have look at this video at 10:35 mark

support7

Could you please verify with the youtuber that it is really the latest linux build ? thanks

MamaLilla

There is also a discussion in the linux-firmware git. The developers confirmed there that the issue can’t be reproduced on their device which is a Zenbook S16 which is also based on Strix Point like the SER9. So it looks more and more like a problem specific to the SER9. This is now further confirmed with the Minisforum Device from the Video.

See the discussion here:
https://gitlab.freedesktop.org/drm/amd/-/issues/3744

gukin

support1

You get the firmware from here:

https://git.kernel.org/pub/scm/linux/kernel/git/firmware/linux-firmware.git/
or just do a wget https://git.kernel.org/pub/scm/linux/kernel/git/firmware/linux-firmware.git/snapshot/linux-firmware-20241110.tar.gz

From there you do a tar fxva linux-firmware-20241110.tar.gz linux-firmware-20241110/amdgpu/dcn_3_5_dmcub.bin
Then you just sudo cp linux-firmware-20241110/amdgpu/dcn_3_5_dmcub.bin /usr/lib/firmware/amdgpu/dcn_3_5_dmcub.bin

Presuming you are running linux 6.10+ and have mesa 24.2+ you should be rewarded with a black screen and a non-booting SER9

MamaLilla

Here you can all linux_firmware releases:
https://git.kernel.org/pub/scm/linux/kernel/git/firmware/linux-firmware.git/

The one from august contains the working dcn_3_5_dmcub.bin all afterwards contain the broken one.

ximejo

All the links are in the thread and might be distro dependent.
The dcn_3_5_dmcub.bin file is part of every linux-firmware release and is located in the /usr/lib/firmware/amdgpu/ folder.
Only the file from the 20240811 release is working with SER9.
You can always start your testing with the latest ubuntu 24.10 iso.

kanzan

@support1 from: https://cdn.kernel.org/pub/linux/kernel/firmware/linux-firmware-20241110.tar.gz
extract, then you’ll find: ./amdgpu/dcn_3_5_dmcub.bin

support1

gukin Thanks , our R&D will test it.

gukin

Follow similar instructions on the linux-firmware-20240811 to get the working dcn_3_5_dmcub.bin.

The thread above on gitlab.freedesktop.org seems to indicate that it’s a BIOS bug; at least we have a workaround.

gukin

Gentle support entities, Mario L. from AMD suggested that you contact them regarding this problem, they have been very responsive to me and I would say that both of you would be interested in finding a solution to this problem.

Here is the thread: https://gitlab.freedesktop.org/drm/amd/-/issues/3744#note_2666260

Oligon

Just was offered a new Firmware via the Update Manager in Linux Mint.
Updated to Kernel 6.8.0-49.49 and Firmware 20240318.git3b128b60-0ubuntu2.5

And it works without the need to downgrade “dcn_3_5_dmcub.bin” again to the older and so far only working solution.

gukin

Oligon

I doubt you have hardware acceleration, the 890m requires the 6.10 or later kernel and mesa 24.2 or later.

Can you run “vkcube”? It won’t run on my system without hardware acceleration complaining about DRI3 missing.

If you run "glxinfo | grep renderer, do you get something like this:
OpenGL renderer string: AMD Radeon Graphics (radeonsi, gfx1150, LLVM 19.1.3, DRM 3.59, 6.12.0)
or do you get something like this:
OpenGL renderer string: llvmpipe (LLVM 15.0.6, 256 bits)

The former is hardware acceleration but the latter is only software acceleration which is a disappointment considering the formidable power of the 890m video adapter.

« Previous Page Next Page »