Linux on SER9.

support4

RandomName
Hi there
When in balanced mode . PC will adjust the power as to its power consumption .
but in performance mode. PC will consider performance first which needs high power consumption

old486whizz

@RandomName - thanks very much for that.

I would like to attach an “smbios.txt” file, captured from the efi shell by running “smbiosview > smbios.txt” - but apparently it’s too big.
I shall instead copy a snippet:
`SMBIOS Entry Point Structure:
Anchor String: SM
EPS Checksum: 0xD0
Entry Point Len: 31
Version: 3.7
Number of Structures: 54
Max Struct size: 169
Table Address: 0×78051000
Table Length: 2523
Entry Point revision: 0×0
SMBIOS BCD Revision: 0×37
Inter Anchor: DMI
Inter Checksum: 0×8A
Formatted Area:
00000000: 00 00 00 00 00 …..

=========================================================
Query Structure, conditions are:
QueryType = Random
QueryHandle = Random
ShowType = SHOW_DETAIL

=========================================================
Type=0, Handle=0×0
Dump Structure as:
Index=0,Length=0×57,Addr=0×78051000
00000000: 00 1A 00 00 01 02 00 F0-03 FF 80 98 0B 00 01 00 …………….
00000010: 00 00 01 0D 05 23 FF FF-20 00 41 6D 65 72 69 63 …..#.. .Americ
00000020: 61 6E 20 4D 65 67 61 74-72 65 6E 64 73 20 49 6E an Megatrends In
00000030: 74 65 72 6E 61 74 69 6F-6E 61 6C 2C 20 4C 4C 43 ternational, LLC
00000040: 2E 00 53 45 52 39 54 31-30 33 00 31 30 2F 32 38 ..SER9T103.10/28
00000050: 2F 32 30 32 34 00 00 /2024..
Structure Type: BIOS Information
Format part Len : 26
Structure Handle: 0
Vendor: American Megatrends International, LLC.
BiosVersion: SER9T103
BiosSegment: 0xF000
BiosReleaseDate: 10/28/2024
BiosSize: 16384 KB
BIOS Characteristics:
PCI is supported
BIOS is Upgradeable(FLASH)
BIOS shadowing is allowed
Boot form CD is supported
Selectable Boot is supported
BIOS ROM is socketed
EDD (Enhanced Disk Driver) Specification is supported
Bits 32:47 are reserved for BIOS Vendor
Bits 48:64 are reserved for System Vendor
BIOS Characteristics Extension Byte1:
ACPI supported
BIOS Characteristics Extension Byte2:
BIOS Boot Specification supported
Enable Targeted Content Distribution
UEFI Specification is supported
Bits 5:7 are reserved for future assignment
SystemBiosMajorRelease: 5
SystemBiosMinorRelease: 35
EmbeddedControllerFirmwareMajorRelease: 255
EmbeddedControllerFirmwareMinorRelease: 255
ExtendedBiosSize: 32 MB

=========================================================
Type=1, Handle=0×1
Dump Structure as:
Index=1,Length=0×61,Addr=0×78051057
00000000: 01 1B 01 00 01 02 03 04-00 02 00 03 00 04 00 05 …………….
00000010: 00 06 00 07 00 08 00 09-06 05 06 41 5A 57 00 53 ………..AZW.S
00000020: 45 52 39 00 44 65 66 61-75 6C 74 20 73 74 72 69 ER9.Default stri
00000030: 6E 67 00 44 65 66 61 75-6C 74 20 73 74 72 69 6E ng.Default strin
00000040: 67 00 44 65 66 61 75 6C-74 20 73 74 72 69 6E 67 g.Default string
00000050: 00 44 65 66 61 75 6C 74-20 73 74 72 69 6E 67 00 .Default string.
00000060: 00 .
Structure Type: System Information
Format part Len : 27
Structure Handle: 1
Manufacturer: AZW
ProductName: SER9
Version: Default string
SerialNumber: Default string
Dump Uuid
size=16:
00000000: 00 02 00 03 00 04 00 05-00 06 00 07 00 08 00 09 …………….
System Wakeup Type: Power Switch
SKUNumber: Default string
Family: Default string

=========================================================
Type=2, Handle=0×2
Dump Structure as:
Index=2,Length=0×55,Addr=0×780510B8
00000000: 02 0F 02 00 01 02 03 04-05 09 06 03 00 0A 00 41 ……………A
00000010: 5A 57 00 53 45 52 39 00-44 65 66 61 75 6C 74 20 *ZW.SER9.Default *
00000020: 73 74 72 69 6E 67 00 44-65 66 61 75 6C 74 20 73 string.Default s
00000030: 74 72 69 6E 67 00 44 65-66 61 75 6C 74 20 73 74 tring.Default st
00000040: 72 69 6E 67 00 44 65 66-61 75 6C 74 20 73 74 72 ring.Default str
00000050: 69 6E 67 00 00 ing..
Structure Type: Base Board Information
Format part Len : 15
Structure Handle: 2
Manufacturer: AZW
ProductName: SER9
Version: Default string
SerialNumber: Default string
AssetTag: Default string
Base Board Feature Flags: Hosting board | Replaceable
LocationInChassis: Default string
ChassisHandle: 0×3
Base Board Board Type: Motherboard`

This has many “Default String” bits in there.
I urge anyone wanting to update the BIOS to capture their info before hand, with both “dmidecode > file.txt” and “smbiosview > smbios.txt”.
But, without the ability to restore these details it isn’t much help.

We need the “AMIDEEFI64.efi” file in order to save and restore the smbios info (and I am unable to find a copy of the file that works).

–
old486whizz

gukin

When either specifying cores 0-3, or 8-11 or using gamemoderun, I’ve seen those cores run up to 5.150 in gkrellm in single core speeds. In multi-core tasks such as compiling a kernel, the full cores will go to 4.6 Ghz while the compact cores top out at 3.3 Ghz.

Even so, building a kernel using “oldconfig” completes in about 12 minutes.

oddhack

Has anyone tried installing current Debian on the SER9 and can share their experience? I have an SER9 on order and will need to do this soon. I imagine that the issues are probably roughly equivalent to Ubuntu as discussed occasionally above, but most posters seem to be using other distros / lineages.

gukin

oddhack
You may find this as a bit of a challenge, the Ryzen AI 370 requires the 6.10 or later kernel, mesa 24.2 or later and linux non-free firmware 20240811 or later.

Sadly, Debian 12.9.0 is running kernel 6.1.123, mesa 22.3.6 and firmware 20230210-5, none of which will play nice with the SER9’s 890M video.

You would have a much happier time running Debian 12.9 on a SER8 8845HS.

oddhack

gukin Thanks, that’s helpful. At least testing is pretty far along and trixie should release by midyear, by historical cadence anyway. Perhaps I should try Ubuntu 24.10 instead, in the meantime? It’s probably the most familiar distro to me of the ones discussed here, though it’s been 6+ years since I was on Ubuntu.

old486whizz

oddhack
I’m using Fedora 41, which uses Kernel 6.12.11, mesa 24.3.4, and ‘linux-firmware’ 20241210-1 .
I’m also using the SER9 version 2 (which is a slightly different version being sold now than the original, with a different BIOS).

Mine worked out of the box.

–
old486whizz

RandomName

Ubuntu 24.10 worked out of the box for me, with the bios already on the device (version STX.3xx.SER9.V103.P8C0M0C15.68.Link.GID.Linux_kimi_v1)

oddhack

RandomName @old486whizz thanks. Mine is somewhere between Hong Kong and LAX ATM, so hopefully recently manufactured with BIOS version 2 as well. I might give trixie testing a try first just for completeness, but sounds like I’ll be on another distro for a while. Pity, I was hoping to just pop in the SSD from another machine running 12.9 and go.

RandomName

You could pop in that SSD and upgrade debian to Testing (which is at kernel 6.12, if I see it correctly) in text mode.
Debian stable is very very stable, but not really comparable to Ubuntu or Fedora’s update speed.

old486whizz

While I’m getting 90% working - SER9 V2 mobo and BIOS is still occasionally giving me a green screen video driver crash.
Green screen, with no recovery except rebooting.

–
old486whizz

support7

old486whizz
Here are the potential solutions:
①Please try to update the driver from kernel.org
②Please change another HDMI/DP cable.
③ Please go to display setting. Please adjust your display settings in the operating system or graphics driver control panel to the recommended settings.

old486whizz

With regards to the green screen - it is a very specific error that is going on here.
I’m playing Octopath Traveller 2, going into a specific room and moving around, and then after a period of time between 1 second and 5 minutes I am experiencing the game hanging, the screen turns off and then a green screen comes up. Before and after the system are perfectly fine - it’s not any cable or resolution settings.

It must be hitting some sort of specific call or memory map which causes the card to perform some odd actions.

I did for a few attempts get an error message “DXGI_ERROR_INVALID_CALL”, in the d2d11query.cpp file..

The only reason I link it here and mention it is due to the green screen, and the music still playing.. I have yet to try and do any proper debugging on this issue - although I did totally wipe the system and re-setup Steam (so that it couldn’t be down to something else putting in bad libraries).
I will be looking at detailed debug logs to see if there’s anything being produced during the crash.

old486whizz

FIX<< I have a fix for the greenscreen / random crashing.
So, after a great amount of tracking bugs down, it appears it’s down to a scheduling issue with shaders. I was able to muck about with the “amdgpu” driver to enable the mes scheduler, and disable power management etc - those reduced the occurrences butthey still happened in my specific test case.
I even tried changing the PBO (Precision Boost Override) settings since I read that over/undervolting may cause issues with timing (as well as clocking).. And allowing for a minor overvolting and allowing the CPU to sustain higher clocks for longer did reduce the times it happened - but it still hapened.
…. until I found the mesa debug variables page, describing that we can use
RADV_DEBUG=shadersync
If you add that as an environment variable before the program (or into /etc/environment…. or in Steam you can add it before %command% in the properties window) then the MESA drivers will sync the shaders.
… You’ll lose ₂% of performance (from what I can see during testing).. But I would rather have that lower performance vs a system that crashes!

support7

Here are the potential solutions:
Please try to update the driver
Please change another HDMI/DP cable.
Please go to display setting. Please adjust your display settings in the operating system or graphics driver control panel to the recommended settings.

old486whizz

So, My fix is confirmed to work (for me).
The setting of “RADV_DEBUG=syncshaders” before any program or in the steam settings prevents any crashing (I think this is probably the AMD drivers and Vulkan - but it could also be a base AMD hardware issue).

I would like to share a tip with everyone too on performance.
I can get in some cases a ₃₀% gain in performance if I set these kernel parameneters (they go into the grub configs - I could go further into details but due to different distro’s it is best to check that yourself):
amdgpu.mes=1 amdgpu.mes_kiq=1 amdgpu.uni_mes=1

These settings use the newer MES Hardware scheduler, which seems to be much better than the default scheduler in the hardware.
I monitored using “radeontop” and “sensors” (which showed me the GPU performance usages, power use and temperatures) and I didn’t notice any significant changes in the use before or after - but I was getting a significant leap in FPS in several “vkmark” results (in some cases a leap from 21000->31000 - which is ₃₀%!!).

I’m sharing these bits in the hope that others may benefit from the information.

–
old486whizz

support7

Thank you for your feedback.

oddhack

Data point: I just installed Debian Trixie (testing) from the latest ISO, and it is working, bringing up gdm, runs glxgears / vkgears with the AMD Mesa drivers. Have not tested to see whether HW acceleration is happening aside from noting that vkgears -present-mode -fullscreen runs at 6700 FPS on a 4K display, which feels like a lot for a software renderer despite how simple the image being rendered is. More encouraging than I was expecting and perhaps I won’t need to migrate to Ubuntu after all.

oddhack

Also to note, the Windows 11 install was borked up. Initially it came up in Win 11 and was extremely unhappy with me because I didn’t have a network connection, which forced a power cycle. Then trying to shrink the 950 GB partition on the first SSD seemed to succeed, but then installing trixie in the free space didn’t seem to install the bootloader properly and it returned to Windows when rebooted - followed by failing to fix the disk problems and eventually saying I’d have to reinstall. At that point I didn’t even know where the local restore partition was - I got an SER9 with 2TB SSD, which turned out to be two separate SSDs each with a huge NTFS partition on them - and I gave up and let the Debian installer completely format NVM0. So I am probably going to have to reinstall Windows 11 from an ISO, and will need the product key. I gather this requires catching the attention of support here, then PMing them a picture of the product ID sticker and the BIOS boot screen? Not sure of the process.

support10

oddhack

Hello there,

Could you please send us a picture of SN code at the bottom of machine ?

Then we will send you the corresponding the key.

Have a nice day!

« Previous Page Next Page »