Home » Blog » Understanding CUDA and NVIDIA GPU Driver – Solve the version conflicts between the Nvidia driver and CUDA toolkit Understanding CUDA and NVIDIA GPU Driver – Solve the version conflicts between the Nvidia driver and CUDA toolkit

Dealing with Nvidia GPU drivers and CUDA development software can sometimes be challenging. Upgrading CUDA versions or updating the Linux system can potentially lead to issues such as GPU driver corruption. When faced with such situations, we often encounter questions for which we need to search online for solutions. Finding the right solution may take some time and effort.


Some questions related to Nvidia driver and CUDA failures.

A)The following packages have unmet dependencies:

 cuda-drivers-535 : Depends: nvidia-dkms-535 (>= 535.161.08)

                    Depends: nvidia-driver-535 (>= 535.161.08) but it is not going to be installed


B)UserWarning: CUDA initialization: CUDA unknown error – this may be due to an incorrectly set up environment, e.g. changing env variable CUDA_VISIBLE_DEVICES after program start. Setting the available devices to be zero. (Triggered internally at ../c10/cuda/CUDAFunctions.cpp:108.)

Reboot after installing CUDA.


C)NVIDIA-SMI has failed because it couldn’t communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running


After going through this time-consuming process, I came to realize that having a deeper understanding of the intricate relationship between CUDA and Nvidia drivers could have enabled me to resolve the driver corruption issue much more swiftly. This realization underscores the importance of acquiring comprehensive knowledge about the interplay between software components and hardware drivers, which can greatly streamline troubleshooting processes and enhance system maintenance efficiency. In this post, I will try to clarify the concepts of GPU driver and CUDA version, and other related questions.


What is CUDA?

One of the significant successes of NVIDIA GPU is due to its CUDA platform. CUDA stands for Compute Unified Device Architecture. It’s a parallel computing platform and application programming interface (API) model created by NVIDIA. CUDA allows developers to utilize the computational power of NVIDIA GPUs (Graphics Processing Units) for general-purpose processing tasks beyond just graphics rendering.


Key components of CUDA include:


CUDA Toolkit: This is a comprehensive development environment provided by NVIDIA for building GPU-accelerated applications. It includes libraries, development tools, compilers (such as nvcc), and runtime APIs.


CUDA C/C++: CUDA extends the C and C++ programming languages with special keywords and constructs that allow developers to write code for both the CPU and the GPU. This enables developers to offload parallelizable portions of their code to the GPU for execution, thereby achieving significant speedups for many types of applications.


Runtime API: CUDA provides a runtime API that allows developers to manage GPU devices, allocate memory on the GPU, launch kernels (parallel functions executed on the GPU), and synchronize between the CPU and GPU.


GPU Architecture: CUDA takes advantage of the massively parallel architecture of modern NVIDIA GPUs, which consist of thousands of cores capable of executing computations simultaneously. CUDA enables developers to exploit this parallelism to accelerate a wide range of tasks, including scientific simulations, data analytics, image processing, and deep learning.


NVCC and Nvidia-SMI

Let’s clarify two important command-line tools in the CUDA ecosystem: nvcc, the NVIDIA CUDA Compiler, and nvidia-smi, the NVIDIA System Management Interface. nvcc serves as the compiler for CUDA, allowing developers to compile CUDA-accelerated applications. On the other hand, nvidia-smi is a command-line utility provided by NVIDIA for monitoring and managing NVIDIA GPU devices. Both nvcc and nvidia-smi are closely tied to specific versions of CUDA. CUDA itself offers both a runtime API and a driver API. The CUDA version reported by nvcc corresponds to the runtime API, while the version displayed in nvidia-smi corresponds to the CUDA version associated with the driver API. It’s important to note that if you install nvcc and the driver separately, or for other reasons, the CUDA version reported by nvcc and nvidia-smi may differ. This discrepancy can occur due to various factors, necessitating careful attention to ensure compatibility between CUDA components.


Typically, the driver API is installed alongside the GPU driver installer, ensuring that nvidia-smi is available for use. Conversely, the runtime API and nvcc are bundled with the CUDA Toolkit Installer. Notably, the CUDA Toolkit Installer does not have to be awareness of the GPU driver API. Consequently, even without a GPU, one can still install the CUDA Toolkit, providing a software environment conducive to coding with CUDA parallel computing, albeit without hardware specifics. This setup allows users to install multiple CUDA versions on a single machine, allowing them the flexibility to select the preferred version. However, it’s crucial to note that the driver version of CUDA is backward compatible with earlier versions of CUDA supported by nvcc. So in general, the CUDA version of driver API is higher or equal to the CUDA version of runtime API.


The compatibility of CUDA version and GPU version can be found from table 3 in https://docs.nvidia.com/deploy/cuda-compatibility/index.html


Install different CUDA versions

Here are all the CUDA versions for installation:


Let us use CUDA Toolkit 12.0 as an example:

Very Important for the last option of Installer Type: runfile (local)

If you chose other options like deb, it may reinstall the old driver, and uninstall your newer GPU driver. But runfile will give you an option during the installation to skip updating the GPU driver, so you may keep your newer drivers. This is very important for case you have already installed the GPU driver separately. 


Install GPU Drivers


sudo apt search nvidia-driver

sudo apt install nvidia-driver-510

sudo reboot




Multiple CUDA Version Switching

To begin with, you need to set up the CUDA environment variables for the actual version in use. Open the .bashrc file (vim ~/.bashrc) and add the following statements:




export LD_LIBRARY_PATH=”$LD_LIBRARY_PATH:/usr/local/cuda/lib64″



This indicates that when CUDA is required, the system will search in the /usr/local/cuda directory. However, the CUDA installations typically include version numbers, such as cuda-11.0. So, what should we do? Here comes the need to create symbolic links. The command for creating symbolic links is as follows:



sudo ln -s /usr/local/cuda-11.0/ /usr/local/cuda



After this is done, a cuda file will appear in the /usr/local/ directory, which points to the cuda-11.0 folder. Accessing this file is equivalent to accessing cuda-11.0. This can be seen in the figure below:


At this point, running nvcc –version will display:



nvcc: NVIDIA (R) Cuda compiler driver

Copyright (c) 2005-2022 NVIDIA Corporation

Built on Sun_Jan__9_22:14:01_CDT_2022

Cuda compilation tools, release 11.0, V11.0.218



For instance, if you need to set up a deep learning environment with Python 3.9.8 + TensorFlow 2.7.0 + CUDA 11.0, follow these steps:


First, create a Python environment with Python 3.9.8 using Anaconda:


conda create -n myenv python=3.9.8

conda activate myenv



Then, install TensorFlow 2.7.0 using pip:


pip install tensorflow==2.7.0



That’s it! Since this Python 3.9.8 + TensorFlow 2.7.0 + CUDA 11.0 environment generally meets the requirements in the code, it is certainly compatible. We just need to ensure that the CUDA version matches the version required by the author.


Solve the driver and CUDA version problems

As we already know the relationship between Nvidia driver and CUDA, we may already know how to solve the above-mentioned problems. 


If you do not want to bother to search over the internet, you can simply remove all Nvidia drivers and CUDA versions, and reinstall them by following the previous steps. Here is one way to get rid of all previous Nividia-related packages.

“` bash

sudo apt-get remove –purge ‘^nvidia-.*’

sudo apt-get remove –purge ‘^libnvidia-.*’

sudo apt-get remove –purge ‘^cuda-.*’

Then run:
sudo apt-get install linux-headers-$(uname -r)


If you plan to upgrade your GPU for advanced AI computing, check with us. You can also sell your used GPU to us for top cash!