I’m attempting to use a RTX 5090 on Fedora 43 in LockDown Mode, aka Secure Boot.
Ollama and other apps fail to use the card, due to what Gemeni feels is their inability to send 32 gb of data directly to the card.
Here is a log of what I’ve tried, thus far.
Gemeni has come to the conclusion that the only way to proceed is to disable Secure Boot.
=====
nvidia-smi .. nvidia
| NVIDIA-SMI 580.142 Driver Version: 580.142 CUDA Version: 13.0 |
| 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 5090 Off | 00000000:01:00.0 Off | N/A |
| 0% 25C P8 18W / 600W | 2MiB / 32607MiB | 0% Default |
| | | N/A |
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
| No running processes found |
nvidia-smi -q -d PERFORMANCE
Timestamp : Sat Mar 28 13:34:03 2026
Driver Version : 580.142
CUDA Version : 13.0
Attached GPUs : 1
GPU 00000000:01:00.0
Performance State : P8
Clocks Event Reasons
Idle : Not Active
Applications Clocks Setting : Not Active
SW Power Cap : Not Active
HW Slowdown : Not Active
HW Thermal Slowdown : Not Active
HW Power Brake Slowdown : Not Active
Sync Boost : Not Active
SW Thermal Slowdown : Not Active
Display Clock Setting : Not Active
Clocks Event Reasons Counters
SW Power Capping : 218861 us
Sync Boost : 0 us
SW Thermal Slowdown : 0 us
HW Thermal Slowdown : 0 us
HW Power Braking : 0 us
Sparse Operation Mode : N/A
=====
nvidia-smi -l 1
nvidia-smi -q | grep “Architecture”
=====
To see if Fedora has successfully assigned the correct drivers to each, run:
switcherooctl list
cat /proc/driver/nvidia/version
NVRM version: NVIDIA UNIX Open Kernel Module for x86_64 580.142 Release Build (dvs-builder@U22-I3-AD05-51-5) Tue Mar 3 19:14:46 UTC 2026
GCC version: gcc version 15.2.1 20260123 (Red Hat 15.2.1-7) (GCC)
cat /sys/devices/system/clocksource/clocksource0/available_clocksource
tsc hpet acpi_pm
env | grep -iE “CUDA|NVIDIA|PALLADIUM”
nothing returned
journalctl -u ollama --since “1 hour ago” | grep -i “pci”
nothing returned
=====
vi start-ollama.sh
export PATH=/usr/local/cuda/bin:$PATH
export LD_LIBRARY_PATH=/usr/local/cuda/lib64:$LD_LIBRARY_PATH
export OLLAMA_LLM_LIBRARY=“cuda_v12”
export CUDA_VISIBLE_DEVICES=0
export GGML_CUDA_FORCE_CUBLAS=1
export OLLAMA_FLASH_ATTENTION=1
export ROCR_VISIBLE_DEVICES=-1
export OLLAMA_DEBUG=1
export OLLAMA_TMPDIR=$HOME/.ollama/tmp
mkdir -p $OLLAMA_TMPDIR
ollama serve
=====
./start-ollama.sh
Starting Ollama on RTX 5090…
…
releasing cuda driver library
level=INFO source=types.go:130
msg=“inference compute”
id=GPU-5164997c-870c-87da-33d6-7532dc534ed5
library=cuda variant=v12
compute=12.0
driver=13.0
name=“NVIDIA GeForce RTX 5090”
total=“31.4 GiB”
available=“30.9 GiB”
.. will post log in a follow up ..
=====
ps aux | grep ollama
[uid] 29318 0.2 0.1 15096472 135956 pts/0 Sl+ 15:22 0:00 ollama serve
[uid] 29340 0.2 0.0 1770604 29008 pts/1 Sl+ 15:23 0:00 ollama run qwen3
[uid] 29369 808 4.9 9533832 6399428 pts/0 Sl+ 15:23 46:56 /usr/lib64/ollama/bin/ollama runner --model ~/.ollama/models/blobs/sha256-a3de86cd1c132c822487ededd47a324c50491393e6565cd14bafa40d0b8e686f --ctx-size 8192 --batch-size 512 --n-gpu-layers 37 --threads 16 --parallel 2 --port 45209
[uid] 53012 0.0 0.0 6560 2448 pts/2 S+ 15:29 0:00 grep --color=auto ollama
=====
groups
wheel
usermod -aG video,render $USER
=====
vi /etc/modprobe.d/nvidia.conf
options nvidia-current NVreg_OpenRmEnableUnsupportedGpus=0
options nvidia NVreg_EnableGpuFirmware=0
OR
options nvidia-current NVreg_OpenRmEnableUnsupportedGpus=1
options nvidia NVreg_EnableGpuFirmware=0
OR
options nvidia NVreg_OpenRmEnableUnsupportedGpus=1
options nvidia NVreg_EnableGpuFirmware=0
dracut -f
reboot
cat /sys/module/nvidia/parameters/NVreg_OpenRmEnableUnsupportedGpus
no such file or directory
=====
journalctl -u ollama
Nothing returned
=====
dnf config-manager addrepo --from-repofile=https://developer.download.nvidia.com/compute/cuda/repos/fedora43/x86_64/cuda-fedora43.repo
dnf clean all
dnf install cuda-toolkit
=====
ldconfig -p | grep cublas
1209 libs found in cache `/etc/ld.so.cache’
none where cublas
@after dnf install install cuda-toolkit
ldconfig -p | grep cublas
1285 libs found in cache `/etc/ld.so.cache’
libcublasLt.so.13 (libc6,x86-64) => /usr/local/cuda/targets/x86_64-linux/lib/libcublasLt.so.13
libcublasLt.so (libc6,x86-64) => /usr/local/cuda/targets/x86_64-linux/lib/libcublasLt.so
libcublas.so.13 (libc6,x86-64) => /usr/local/cuda/targets/x86_64-linux/lib/libcublas.so.13
libcublas.so (libc6,x86-64) => /usr/local/cuda/targets/x86_64-linux/lib/libcublas.so
=====
dmesg | grep -i “lockdown”
kernal is in lockdown from EFI Secure Boot mode;
=====
modinfo -F signer nvidia
signer: sage_1774667474_4502a9e6
modinfo -F signer nvidia_uvm
signer: sage_1774667474_4502a9e6
modinfo nvidia | grep -E “OpenRm|GpuFirmware”
parm: NVreg_EnableGpuFirmware:int
parm: NVreg_EnableGpuFirmwareLogs:int
parm: NVreg_OpenRmEnableUnsupportedGpus:int
=====
grep . /sys/module/nvidia/parameters/* | grep -i “OpenRm”
grep: /sys/module/nvidia/parameters/*: No such file or directory
=====
loginctl show-session $(loginctl | grep $(whoami) | awk ‘{print $1}’) -p Type
Type=tty
Type=unspecified
Type=tty
Type=tty
======
In GRUB
Add to the linux or linuxefi line …
nvidia.NVreg_OpenRmEnableUnsupportedGpus=1
nvidia.NVreg_EnableGpuFirmware=0
=====
cat /proc/cmdline
nvidia.NVreg_OpenRmEnableUnsupportedGpus=1
nvidia.NVreg_EnableGpuFirmware=0
Only appear when manually added during edit in Grub.