VENDOR_ID="${VENDOR_ID:-0x10de}"   # NVIDIA
# DEVICE_ID: exact PCI device id to match (e.g. 0x2b85 for RTX 5090). Leave
# empty to match ANY NVIDIA display controller by PCI class — this covers the
# whole RTX 50-series (5090=0x2b85, 5060 Ti=0x2d04, …) without a hardcoded list,
# in the same spirit as is_tb_tunneled below. Set DEVICE_ID to pin one board.
DEVICE_ID="${DEVICE_ID:-0x2d04}"
# PCI base class 0x03 = display controller (VGA 0x0300 / 3D 0x0302); selects the
# GPU function and excludes its HDMI-audio function (class 0x0403).
GPU_CLASS_PREFIX="${GPU_CLASS_PREFIX:-0x03}"

# nvidia-smi 

Mon Jul 27 15:15:18 2026       
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 610.43.03              KMD Version: 610.43.03     CUDA UMD Version: 13.3     |
+-----------------------------------------+------------------------+----------------------+
| GPU  Name                 Persistence-M | Bus-Id          Disp.A | Volatile Uncorr. ECC |
| Fan  Temp   Perf          Pwr:Usage/Cap |           Memory-Usage | GPU-Util  Compute M. |
|                                         |                        |               MIG M. |
|=========================================+========================+======================|
|   0  NVIDIA GeForce RTX 3060 ...    Off |   00000000:01:00.0 Off |                  N/A |
| N/A   50C    P0             26W /  115W |       0MiB /   6144MiB |      0%      Default |
|                                         |                        |                  N/A |
+-----------------------------------------+------------------------+----------------------+
|   1  NVIDIA GeForce RTX 5060 Ti     Off |   00000000:05:00.0 Off |                  N/A |
| 45%   36C    P0             15W /  180W |       0MiB /  16311MiB |      1%      Default |
|                                         |                        |                  N/A |
+-----------------------------------------+------------------------+----------------------+

+-----------------------------------------------------------------------------------------+
| Processes:                                                                              |
|  GPU   GI   CI              PID   Type   Process name                        GPU Memory |
|        ID   ID                                                               Usage      |
|=========================================================================================|
|  No running processes found                                                             |
+-----------------------------------------------------------------------------------------+

GRUB_DEFAULT=0
GRUB_TIMEOUT_STYLE=hidden
GRUB_TIMEOUT=0
GRUB_DISTRIBUTOR='tuxedo'
GRUB_CMDLINE_LINUX_DEFAULT="quiet splash"
#GRUB_CMDLINE_LINUX="iommu.passthrough=1 thunderbolt.host_reset=false pcie_aspm.policy=performance thunderbolt.clx=0 pcie_port_pm=off pci=resource_alignment=35@0000:00:01.0"
GRUB_CMDLINE_LINUX="iommu.passthrough=1 thunderbolt.host_reset=false pcie_aspm.policy=performance thunderbolt.clx=0 pcie_port_pm=off pci=resource_alignment=35@0000:00:05.0"

# update-grub2
Sourcing file `/etc/default/grub'
Sourcing file `/etc/default/grub.d/50-tuxedo-fix-nvidia-preserve-vram-suspend.cfg'
Sourcing file `/etc/default/grub.d/90-tuxedo.cfg'
Sourcing file `/etc/default/grub.d/tuxedo-theme.cfg'
Generating grub configuration file ...
Found theme: /usr/share/grub/themes/tuxedo/theme.txt
Found linux image: /boot/vmlinuz-6.17.0-122035-tuxedo
Found initrd image: /boot/initrd.img-6.17.0-122035-tuxedo
Found memtest86+ 64bit EFI image: /boot/memtest86+x64.efi
Warning: os-prober will be executed to detect other bootable partitions.
Its output will be used to detect bootable binaries on them and create new boot entries.
Adding boot menu entry for UEFI Firmware Settings ...
done

# cat /proc/cmdline 
BOOT_IMAGE=/boot/vmlinuz-6.17.0-122035-tuxedo root=UUID=a25de6da-74c3-49d2-9b12-7b9d960b2ad8 ro iommu.passthrough=1 thunderbolt.host_reset=false pcie_aspm.policy=performance thunderbolt.clx=0 pcie_port_pm=off pci=resource_alignment=35@0000:00:01.0 quiet splash nvidia.NVreg_PreserveVideoMemoryAllocations=1 nvidia.NVreg_TemporaryFilePath=/var/tmp vt.handoff=7
# nvidia-smi 
Mon Jul 27 14:52:12 2026       
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 610.43.03              KMD Version: 610.43.03     CUDA UMD Version: 13.3     |
+-----------------------------------------+------------------------+----------------------+
| GPU  Name                 Persistence-M | Bus-Id          Disp.A | Volatile Uncorr. ECC |
| Fan  Temp   Perf          Pwr:Usage/Cap |           Memory-Usage | GPU-Util  Compute M. |
|                                         |                        |               MIG M. |
|=========================================+========================+======================|
|   0  NVIDIA GeForce RTX 3060 ...    Off |   00000000:01:00.0 Off |                  N/A |
| N/A   47C    P0             24W /  115W |       0MiB /   6144MiB |      3%      Default |
|                                         |                        |                  N/A |
+-----------------------------------------+------------------------+----------------------+

+-----------------------------------------------------------------------------------------+
| Processes:                                                                              |
|  GPU   GI   CI              PID   Type   Process name                        GPU Memory |
|        ID   ID                                                               Usage      |
|=========================================================================================|
|  No running processes found                                                             |
+-----------------------------------------------------------------------------------------+

# systemctl disable aorus

#  sudo prime-select nvidia 
Info: selecting the nvidia profile
Deleting /lib/modprobe.d/nvidia-runtimepm.conf
Updating the initramfs. Please wait for the operation to complete:
Done


# grep "nvidia_drm" /etc/modprobe.d/*
/etc/modprobe.d/aorus.conf:#blacklist nvidia_drm
/etc/modprobe.d/aorus.conf:install nvidia_drm /bin/false
/etc/modprobe.d/aorus.conf:options nvidia_drm modeset=0 fbdev=0
/etc/modprobe.d/nvidia.conf:#blacklist nvidia_drm
/etc/modprobe.d/nvidia.conf.aorus.00:#blacklist nvidia_drm
/etc/modprobe.d/nvidia-graphics-drivers-kms.conf:# aorus-disabled: options nvidia_drm modeset=1
/etc/modprobe.d/nvidia-graphics-drivers-kms.conf.aorus.00:options nvidia_drm modeset=1
/etc/modprobe.d/nvidia-power-management.conf:# aorus-disabled: options nvidia_drm modeset=1
/etc/modprobe.d/nvidia-power-management.conf.aorus.00:options nvidia_drm modeset=1

# Blacklist display modules for compute-only
#blacklist nvidia_drm
blacklist nvidia_modeset
blacklist intel_snd_hp

# Override system softdep - only load nvidia-uvm, skip drm/modeset (compute-only)
#softdep nvidia post: nvidia-uvm

# Disable Resizable BAR (resize fails with small bridge windows)
# aorus-disabled: options nvidia NVreg_EnableResizableBar=0

# Disable dynamic power management (prevents D3cold issues over Thunderbolt)
# aorus-disabled: options nvidia NVreg_DynamicPowerManagement=0

# Force eGPU mode (bypasses Thunderbolt bridge detection for TB4/TB5 enclosures)
#options nvidia NVreg_ForceExternalGpu=1
#options nvidia Vreg_EnablePCIeGen3=3
#options nvidia NVreg_EnableGpuFirmware=0

#options nvidia NVreg_OpenRmEnableUnsupportedGpus=1
#options nvidia NVreg_RmMsg=":"
#


# aorus-disabled: options nvidia NVreg_DynamicPowerManagement=0
# aorus-disabled: options nvidia NVreg_EnableGpuFirmwareLogs=2
# aorus-disabled: options nvidia NVreg_EnablePCIeGen3=1
# aorus-disabled: options nvidia NVreg_EnableResizableBar=1
# aorus-disabled: options nvidia NVreg_EnableStreamMemOPs=1
# aorus-disabled: options nvidia NVreg_InitializeSystemMemoryAllocations=1
# aorus-disabled: options nvidia NVreg_PreserveVideoMemoryAllocations=1
# aorus-disabled: options nvidia NVreg_ResmanDebugLevel=2
options nouveau modeset=0
blacklist nouveau

Ensuite :

# sudo systemctl restart systemd-modules-load.service
# sudo systemctl status systemd-modules-load.service
● systemd-modules-load.service - Load Kernel Modules
     Loaded: loaded (/usr/lib/systemd/system/systemd-modules-load.service; static)
     Active: active (exited) since Mon 2026-07-27 14:20:36 CEST; 2s ago
       Docs: man:systemd-modules-load.service(8)
             man:modules-load.d(5)
    Process: 55332 ExecStart=/usr/lib/systemd/systemd-modules-load (code=exited, status=0/SUCCESS)
   Main PID: 55332 (code=exited, status=0/SUCCESS)
        CPU: 5ms

juil. 27 14:20:36 tuxedo-17 systemd[1]: Starting systemd-modules-load.service - Load Kernel Modules...
juil. 27 14:20:36 tuxedo-17 systemd-modules-load[55332]: Module 'loop' is built in
juil. 27 14:20:36 tuxedo-17 systemd-modules-load[55332]: Module 'i2c_dev' is built in
juil. 27 14:20:36 tuxedo-17 systemd-modules-load[55332]: Module 'loop' is built in
juil. 27 14:20:36 tuxedo-17 systemd-modules-load[55332]: Module 'fuse' is built in
juil. 27 14:20:36 tuxedo-17 systemd[1]: Finished systemd-modules-load.service - Load Kernel Modules.

• 610.43.03 : Linux x64 (AMD64/EM64T) Display Driver 610.43.02 | Linux 64-bit
• Tue Jul 07, 2026

• Linux x64 (AMD64/EM64T) Display Driver 610.43.02 | Linux 64-bit

[  558.622194] RIP: 0010:gpuGetKernelFifoShared_IMPL+0x43/0x60 [nvidia]
[  558.623096] Code: 74 0b 41 80 bc 24 81 04 00 00 00 74 12 49 8b 84 24 d8 1c 00 00 4c 8b 65 f8 c9 c3 0f 1f 40 00 4c 89 e7 e8 20 6c 0f 00 49 89 c4 <49> 8b 84 24 d8 1c 00 00 4c 8b 65 f8 c9 c3 66 2e 0f 1f 84 00 00 00
[  558.623102] RSP: 0018:ffffced1047bf658 EFLAGS: 00010246
[  558.623110] RAX: 0000000000000000 RBX: ffff88d1c301c0c8 RCX: 0000000000000020
[  558.623114] RDX: 0000000000000001 RSI: ffffced1047bf5fc RDI: 0000000000000002
[  558.623119] RBP: ffffced1047bf668 R08: 0000000000000020 R09: 0000000000000000
[  558.623123] R10: 0000000000000000 R11: 0000000000000000 R12: 0000000000000000
[  558.623127] R13: ffffced1047bf6e8 R14: ffffced1047bf710 R15: ffffced108f7d008
[  558.623132] FS:  0000000000000000(0000) GS:ffff88e04a562000(0000) knlGS:0000000000000000
[  558.623138] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[  558.623143] CR2: 0000000000001cd8 CR3: 0000000c08a40005 CR4: 0000000000f72ef0
[  558.623149] PKRU: 55555554
[  558.623153] note: llama-server[4082] exited with irqs disabled
[  558.623208] Fixing recursive fault but reboot is needed!
[  558.623212] BUG: scheduling while atomic: llama-server/4082/0x00000000
[  558.623217] Modules linked in: ccm rfcomm snd_seq_dummy snd_hrtimer evdi(OE) nvidia_uvm(OE) nvidia(OE) qrtr cmac algif_hash algif_skcipher af_alg bnep typec_displayport snd_hda_codec_intelhdmi typec_thunderbolt xe sunrpc snd_sof_pci_intel_tgl snd_sof_pci_intel_cnl snd_hda_codec_alc269 snd_sof_intel_hda_generic snd_hda_scodec_component soundwire_intel snd_hda_codec_realtek_lib gpu_sched snd_sof_intel_hda_sdw_bpt snd_hda_codec_generic drm_gpuvm snd_sof_intel_hda_common drm_gpusvm_helper snd_soc_hdac_hda drm_exec drm_suballoc_helper snd_sof_intel_hda_mlink snd_sof_intel_hda soundwire_cadence snd_sof_pci snd_sof_xtensa_dsp snd_sof snd_sof_utils snd_soc_acpi_intel_match snd_soc_acpi_intel_sdca_quirks soundwire_generic_allocation snd_soc_acpi intel_rapl_msr soundwire_bus intel_rapl_common snd_soc_sdca snd_hda_codec_nvhdmi intel_uncore_frequency crc8 intel_uncore_frequency_common snd_hda_codec_hdmi intel_tcc_cooling x86_pkg_temp_thermal snd_soc_avs intel_powerclamp snd_soc_hda_codec snd_hda_intel coretemp snd_hda_ext_core
[  558.623332] pcieport 0000:04:03.0: pciehp: Slot #3 AttnBtn- PwrCtrl- MRL- AttnInd- PwrInd- HotPlug+ Surprise+ Interlock- NoCompl+ IbPresDis- LLActRep+
[  558.623341]  snd_hda_codec iwlmvm uvcvideo snd_hda_core snd_seq_midi videobuf2_vmalloc kvm_intel uvc snd_soc_core snd_intel_dspcfg snd_seq_midi_event videobuf2_memops snd_compress videobuf2_v4l2 ac97_bus cmdlinepart snd_rawmidi mac80211 videobuf2_common btusb snd_intel_sdw_acpi ee1004 tuxedo_nb02_nvidia_power_ctrl(OE) snd_hwdep snd_pcm_dmaengine spi_nor btrtl kvm binfmt_misc uniwill_wmi(OE) i915 snd_seq tuxedo_io(OE) clevo_wmi(OE) btintel videodev libarc4 snd_pcm mei_pxp mtd mei_hdcp irqbypass tuxedo_keyboard(OE) i2c_i801 snd_seq_device btbcm iwlwifi tuxedo_compatibility_check(OE) asus_wmi spi_intel_pci i2c_smbus btmtk rapl nls_iso8859_1 mei_me platform_profile led_class_multicolor wmi_bmof intel_cstate spi_intel mc i2c_mux drm_buddy snd_timer bluetooth cfg80211 mei drm_display_helper snd intel_pmc_core cec soundcore pmt_telemetry rc_core pmt_discovery i2c_algo_bit pmt_class intel_pmc_ssram_telemetry intel_hid intel_vsec acpi_pad sparse_keymap acpi_tad input_leds joydev serio_raw mac_hid sch_fq_codel vmwgfx
[  558.623499]  drm_ttm_helper ttm msr parport_pc ppdev lp parport dm_crypt nvme_fabrics efi_pstore nfnetlink dmi_sysfs ip_tables x_tables autofs4 btrfs blake2b_generic raid10 raid456 async_raid6_recov async_memcpy async_pq async_xor async_tx xor raid6_pq raid1 raid0 linear usbhid cdc_ether usbnet r8152 mii ucsi_acpi typec_ucsi typec hid_multitouch r8125(OE) hid_generic nvme uas usb_storage r8169 polyval_clmulni ahci nvme_core ghash_clmulni_intel thunderbolt intel_lpss_pci realtek libahci i2c_hid_acpi intel_lpss nvme_keyring i2c_hid idma64 nvme_auth hid video wmi pinctrl_tigerlake aesni_intel
[  558.623616] CPU: 2 UID: 0 PID: 4082 Comm: llama-server Tainted: G      D W  OE       6.17.0-122035-tuxedo #35~24.04.1tux1 PREEMPT(voluntary) 
[  558.623624] Tainted: [D]=DIE, [W]=WARN, [O]=OOT_MODULE, [E]=UNSIGNED_MODULE
[  558.623626] Hardware name: TUXEDO TUXEDO Polaris Intel Gen3 (TGL)/GMxTGxx, BIOS N.1.07A03 09/24/2021
[  558.623629] Call Trace:
[  558.623632]  <TASK>
[  558.623637]  dump_stack_lvl+0x76/0xa0
[  558.623646]  dump_stack+0x10/0x20
[  558.623650]  __schedule_bug+0x64/0x80
[  558.623659]  __schedule+0x685/0x7a0
[  558.623667]  ? vprintk+0x18/0x50
[  558.623674]  do_task_dead+0x4a/0x60
[  558.623683]  make_task_dead+0x136/0x140
[  558.623691]  rewind_stack_and_make_dead+0x16/0x20
[  558.623699] RIP: 0033:0x75b36aa98d71
[  558.623704] Code: Unable to access opcode bytes at 0x75b36aa98d47.
[  558.623706] RSP: 002b:000075b1117cb670 EFLAGS: 00000246 ORIG_RAX: 00000000000000ca
[  558.623710] RAX: fffffffffffffdfc RBX: 00006089d8e42dd0 RCX: 000075b36aa98d71
[  558.623713] RDX: 0000000000000000 RSI: 0000000000000189 RDI: 00006089d8e42df8
[  558.623716] RBP: 000075b1117cb6b0 R08: 0000000000000000 R09: 00000000ffffffff
[  558.623719] R10: 000075b1117cb800 R11: 0000000000000246 R12: 0000000000000000
[  558.623721] R13: 0000000000000000 R14: 00006089d8e08a10 R15: 00006089d8e42df8
[  558.623728]  </TASK>
[  558.623955] nvidia 0000:05:00.0: enabling device (0000 -> 0003)
[  558.624301] nvidia 0000:05:00.0: vgaarb: VGA decodes changed: olddecodes=io+mem,decodes=none:owns=none

#  apt install -y nvidia-cuda-toolkit
Lecture des listes de paquets... Fait
Construction de l'arbre des dépendances... Fait
Lecture des informations d'état... Fait      
nvidia-cuda-toolkit est déjà la version la plus récente (12.0.140~12.0.1-4build4).
0 mis à jour, 0 nouvellement installés, 0 à enlever et 0 non mis à jour.

#  export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/cuda/lib64:/usr/local/cuda/extras/CUPTI/lib64
...
# export PATH=$PATH:$CUDA_HOME/bin

# npm install -g npm@latest

# cmake -B build -DGGML_CUDA=ON  -DCMAKE_CUDA_COMPILER=`which nvcc` -DLLAMA_CURL=ON -DCMAKE_CUDA_ARCHITECTURES=86

# systemctl stop llama-swap

# cmake --build build -j$(nproc)

# cd build

# make install

6.30.913.194 E CUDA error: unspecified launch failure
6.30.913.200 E   current device: 0, in function ggml_backend_cuda_buffer_set_tensor at /home/XXXX/Documents/llama.cpp/ggml/src/ggml-cuda/ggml-cuda.cu:719
6.30.913.201 E   cudaStreamSynchronize(((cudaStream_t)0x2))
/home/XXXX/Documents/llama.cpp/ggml/src/ggml-cuda/ggml-cuda.cu:104: CUDA error