First, visit the official Ubuntu website and download the latest version of Ubuntu ISO file. The installation method is to use a USB drive as bootable media.
The capacity of the Ubuntu ISO image is approximately 6 GB, so prepare a USB drive with a capacity of 6 GB or more. We recommend a USB drive of at least 8 GB.
Next, use a free tool called Rufus to burn the downloaded ISO file to the USB drive. This handy tool turns the USB drive into bootable media, allowing the computer to boot directly from this USB.
The Rufus screen will ask you about partitioning and the target system during installation, but if you have no particular reason to do so, the default settings are fine. This is recommended for novice users, as Ubuntu will automatically apply the settings it deems most appropriate. when writing an Ubuntu ISO image to a USB drive, all data in that USB drive will be erased. If you have important files stored on the USB, back them up beforehand or use another USB drive.
In order to boot your computer from a USB drive, you must first enable USB booting in the BIOS (or UEFI) settings. When booting the computer, you can usually enter the BIOS settings screen by pressing the “Delete” key or the “F2” key. In the settings screen, find the “Boot” menu and set the USB device to the top of the boot sequence. The computer will now attempt to boot from the USB drive first.
When installing Ubuntu, the desktop version allows you to select Japanese as the language setting during the installation process. However, the server version does not provide an option to select Japanese during the installation process. In order to get the latest updates during the installation process, I selected Wi-Fi and connected to the Internet. It is possible to install Ubuntu without an Internet connection, but if you are connected, you can simultaneously update your system and install necessary packages during the installation. We left most settings at default for the installation, but since this computer has a GPU, we also decided to install third-party drivers and software. if you want to remove Windows altogether and use Ubuntu on its own, It is common practice to erase the existing Windows partition, including the Windows Boot Manager, before installing Ubuntu. This way, the system will be simpler and you will have fewer boot options. In this case, first completely delete the disk and then install Ubuntu. At this point, you can choose the option of using LVM (Logical Volume Manager) for more advanced disk management; LVM gives you the flexibility to manage your disks, including adjusting disk partition sizes and making future expansion easier. If disk space is limited, the advantages of using LVM are not as great. If you have less disk space, it will be difficult to take full advantage of LVM’s flexible management and expansion capabilities, and it may be easier and more efficient to choose a simpler partition configuration for management.
After the installation is complete and the reboot button is pressed, some systems will automatically attempt to boot from the internal disk, but to ensure that the next boot will be from the internal disk, it is recommended that the installation media (USB drive) be removed. This avoids the risk of accidentally starting the installation process again. In my case, the system did not boot from the USB drive without removing it. After rebooting, the system may reboot itself several times. This is generally because when a new operating system boots for the first time, some initial settings, driver installations, and updates are automatically applied. Multiple reboots may be necessary, especially when there are major updates or when new hardware settings are applied to the system.
One of the first important checks to make after installing Ubuntu is to ensure that the NVIDIA display drivers are properly installed. There is an icon called Additional Drivers, which can be found here.
You can also open the terminal and run the following command to view information about the installed NVIDIA drivers and the current state of the GPU
nvidia-smi
According to this command, it shows CUDA 12.2. To further verify that the CUDA Toolkit is installed, enter the following command
nvcc --versionThe following message is displayed when you actually try it.
Command ‘nvcc-‘ not found. Did you mean:
command ‘nvcc’ from deb nvidia-cuda-toolkit (12.0.140~12.0.1-3)
Try: sudo apt install
nvcc is a command for the NVIDIA CUDA compiler. However, the command “nvcc- v” is incorrect and will produce an error. You can get the version information from the error message. Actually, the correct usage of nvcc is “nvcc –version”.
I believe you can install it with this command, but the official NVIDIA page does not show it, probably because Ubuntu 24.04 has just been released.
The CUDA version indicated by nvidia-smi in Ubuntu represents the latest version of CUDA supported by that NVIDIA driver, but the version of the CUDA Toolkit included in the Ubuntu repository is not necessarily the latest.
Also, the following information is complicated and should be noted.
The complicated part about the difference between CUDA and CUDA Toolkit.
- CUDA (Compute Unified Device Architecture):.
- A programming model and technology developed by NVIDIA for GPUs.
- It is the underlying technology that enables parallel computation using GPUs and is necessary for many applications to take advantage of the power of GPUs.
- CUDA Toolkit:.
- A software package that provides a CUDA development environment.
- The Toolkit includes a compiler (
nvcc), libraries, debugging tools, and other utilities related to NVIDIA’s GPUs for developing CUDA programs. - The CUDA Toolkit also includes sample projects, which you can use to learn the basics of CUDA programming and try out specific implementation examples.
Furthermore, the CUDA version displayed by the nvidia-smi command indicates the version of CUDA supported by that NVIDIA driver, but that does not mean that the CUDA Toolkit is actually installed. This version only represents the maximum version of CUDA that the driver supports, and it requires a separate check to see which CUDA Toolkit is actually installed or has been installed.
To check if the CUDA Toolkit is actually installed or not, you can take the following steps in addition to the command
- Checking the CUDA compiler: Type
nvcc --versionin your terminal to check the version of the CUDA compiler. This will show you the version of the CUDA Toolkit that is actually installed. - Check libraries: Access the
/usr/local/cudadirectory to check the installed libraries. This directory is the general path to where the CUDA Toolkit is installed. - Check for samples: Since the CUDA Toolkit includes sample projects, it is also useful to check if these samples exist and can be compiled and executed.
For #1, as mentioned earlier, no such directory was found; checking #2, no such directory existed; if CUDA Toolkit is installed, then /usr/local/cuda exists as a symbolic link, and the latest or specific CUDA Toolkit version installation directory (e.g. /usr/local/cuda-12.2 ). If this directory does not exist, it indicates that CUDA Toolkit is not installed.
On Windows, the CUDA version indicated by nvidia-smi is the version of CUDA supported by that NVIDIA driver, but the actual installed version of CUDA Toolkit must be determined from the version number in the installation directory The actual installed version of CUDA Toolkit must be determined from the version number in the installation directory.
For example, if nvidia-smi indicates that CUDA 12.4 is supported, but the directory C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.8 exists, the actually installed CUDA Toolkit version is 11.8. This means that the previous version is still installed and has not been upgraded.
