Skip to content

Commit 8fcc454

Browse files
committed
add initial nvidia docs
1 parent a81ce58 commit 8fcc454

1 file changed

Lines changed: 82 additions & 0 deletions

File tree

docs/documentation/nvidia.md

Lines changed: 82 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,82 @@
1+
---
2+
title: "NVIDIA Drivers"
3+
---
4+
5+
# NVIDIA Driver Packages
6+
7+
After a [successful RFC approval by ALESCO](https://github.com/AlmaLinux/ALESCo/pull/8) we began building and shipping NVIDIA driver packages. These packages are based on the NVIDIA open source drivers and are secureboot-signed. The drivers are built using kernel kmods meaning upon installation/update there are no DKMS compilations required.
8+
9+
We build and ship drivers for AlmaLinux Kitten 10, AlmaLinux 10, and AlmaLinux 9. At this time we do not build or ship drivers for AlmaLinux 8, and support for AlmaLinux 8 is not expected at this time.
10+
11+
NVIDIA drivers are only available for x86_64 and aarch64 architectures (excluding AlmaLinux 10 x86_64_v2 builds for now).
12+
13+
## Installation of GPU Driver
14+
15+
Installation is very easy, requiring only 2 `dnf` commands - the first to enable the repository, and the second to install the drivers.
16+
17+
`sudo dnf install almalinux-release-nvidia-driver`
18+
19+
This will enable the AlmaLinux NVIDIA driver repository, AlmaLinux CRB repository, and NVIDIA's upstream CUDA repository. It will also install the [EPEL repository](https://docs.fedoraproject.org/en-US/epel/).
20+
21+
`sudo dnf install nvidia-open-kmod nvidia-driver`
22+
23+
This will install the minimal packages needed for the GPU driver to function.
24+
25+
For a more complete install you can use `sudo dnf install nvidia-open`.
26+
27+
At this point you can reboot to load the kernel module, or, if you're already running the latest kernel, `sudo modprobe nvidia_drm` will load the module.
28+
29+
### Verify
30+
31+
You can confirm the kernel modules are loaded as follows:
32+
33+
```
34+
[root@server]# lsmod |grep -i nvidia
35+
nvidia_drm 151552 0
36+
nvidia_modeset 2166784 1 nvidia_drm
37+
drm_ttm_helper 16384 1 nvidia_drm
38+
video 81920 1 nvidia_modeset
39+
nvidia 12988416 1 nvidia_modeset
40+
```
41+
42+
### nvidia-smi
43+
44+
Many people want or need nvidia-smi for information about their GPU(s). The `nvidia-smi` binary is provided by the `nvidia-driver-cuda` package and can be installed as such:
45+
46+
`sudo dnf install nvidia-driver-cuda`
47+
48+
You can then run it directly:
49+
50+
```
51+
[root@server]# nvidia-smi
52+
Mon Jul 21 14:23:02 2025
53+
+-----------------------------------------------------------------------------------------+
54+
| NVIDIA-SMI 575.64.03 Driver Version: 575.64.03 CUDA Version: 12.9 |
55+
|-----------------------------------------+------------------------+----------------------+
56+
| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |
57+
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
58+
| | | MIG M. |
59+
|=========================================+========================+======================|
60+
| 0 NVIDIA L4 Off | 00000000:35:00.0 Off | 0 |
61+
| N/A 38C P8 11W / 72W | 0MiB / 23034MiB | 0% Default |
62+
| | | N/A |
63+
+-----------------------------------------+------------------------+----------------------+
64+
65+
+-----------------------------------------------------------------------------------------+
66+
| Processes: |
67+
| GPU GI CI PID Type Process name GPU Memory |
68+
| ID ID Usage |
69+
|=========================================================================================|
70+
| No running processes found |
71+
+-----------------------------------------------------------------------------------------+
72+
```
73+
74+
### Installation of CUDA Components
75+
#### CUDA SDK
76+
`sudo dnf install cuda`
77+
#### GPUDirect Filesystems
78+
At this time NVIDIA GPUDirect components are not available.
79+
80+
## Common Problems
81+
* `modprobe: ERROR: could not insert 'nvidia_drm': No such device`
82+
* This most likely stems from your running kernel being different than the latest installed kernel. Rebooting should fix it if that is the case.

0 commit comments

Comments
 (0)