How to install CUDA on UBUNTU(Linux) | CUDA installation in 2024

everyone welcome back I just switched a new machine this machine already has a GPU which is capable with Cuda also and uh in this new environment which is Ubuntu 22.04 which is the latest version of Ubuntu and I'm gonna install my libraries my tools my packages from the beginning before uh installing before creating my new environment also I want to show you how to install this kind of things to new environment and I'm gonna start with Cuda of course let's start how to install Cuda and in the in the future videos I'm gonna also show you how to install other kind of packages libraries which I'm gonna use in my projects okay maybe we need to write and after that just enter to the first page and here the main page of Cuda toolkit and we need for installation guide Linux and after that we are gonna download ours kind of things to our mission from this website and after that uh before starting the most important one we need to check for the installation actions this is very important because most of the people directly servicing to do Cuda toolkit which is wrong you need to check first of all your GPU is supported by Cuda or not how to check that just copy this to your terminal and after that you are gonna see your GPU model and after that there is a list in here which is gpus included capable Cuda and here you need to find your GPU which is mine in here GeForce 3060 which is this one also similarly you need to check your GPU is in this list or not this is the most important case and there are other also installation uh actions we don't need this because we just I just create my own Ubuntu environment I I believe of course there's a GCC because I didn't do something dangerous of course it's gonna give me default some GCC so to be sure you can also follow these steps they are in your environment is established is already exist or not okay since we already showed our GPU is capable then we can switch to this page and in here if I'm not wrong you need to join this Nvidia developer or maybe you don't need I don't know and after that we need to choose this one because also with you will name a you are gonna see the type of the machine X6 x86 underscore 64 and after that we need to switch to Ubuntu 22.04 in your case if you have even 18 is gonna support this 20 maybe if I'm not wrong or you can uh go back to the other Cuda toolkit since this is the latest version I'm gonna continue with this one and Debian local we need to choose after that uh the the biggest one the biggest folder is this Debian folder which is I already downloaded to here in from properties we can see this is 3.3 gigabyte already and this is a huge folder uh maybe uh with this step you are gonna download it this can take little long time according to your internet speed okay let's start with this uh line and after that we can copy that to the terminal it's just gonna get the pin and after that we need to move that to uh desired folder let's finish let's wait it to finish it first of all and after done we need to move it and the next step is downloading this since I already download I don't need to do it again and the other one is just the packaging [Music] and when this is finished we are gonna copy also that one let's wait it to finish and the next step is copying just and the other one is just we need to update and and the latest uh step is just installation Cuda since I already did that before to spend to earn time so I just gonna it's just gonna finish quickly I believe yeah it's just done so the next one is we need to reboot our system to get the drivers properly and after that uh when we came back we are gonna continue with the post installation actions which is just setting some uh Library pass and kind of pets and after that we are gonna also try some um GitHub samples so we can be totally sure the installation is fine and the Cuda is working fine okay after rebooting see see you again hi again we just rebooted our system so now we need to uh check we already installed our Kudo or not how to check that let's open a terminal uh the people generally trying this Nvidia SMI which is the wrong command to check the Cuda version it's saying in here Kuda version 12.2 then the people are thinking okay I installed Coda properly but this is the wrong way this is not the comment we need to check Kuda installation uh result we need to use the command mvcc version which is this one but after rebooting the first system the first time they are gonna see this mvcc not found but can be installed with disconnect and if you use this one you are gonna go in a terrible way because this is gonna install this is gonna try to install akuda again which is the default one maybe 10.0 but you are trying to let's say 12.2 you are trying to install so there is gonna be conflict in the back end so this is a dangerous command to not do this one you already installed your Kudo you can check also via on this computer generally is uh installed to user and the local one and here you see already it's installed you just need to set some pads to your bash RC how to do that in here you need to also check this post installation actions which is already given in Coda documentation there is already too bad we need to set this one and this one just copy this one and go to home folder just Ctrl H and see the hidden files which is best RC in here just paste to here and the other one also we need to paste by the way this is for 64-bit this for 32-bit be careful also similarly for the pet generally is installed to local but if you install two different location or your or your version is different you need to update you need to modify accordingly okay also I don't need here after copying these patch you need to save by charcity and close and we can try mvcc version again but we are gonna see the same thing because this is the terminal we opened before before updating bash RC let's reopen and after that write the comment again so now I believe it's gonna show me the correct one which is 12.2 already installed but still I want to be sure more clearly I installed properly or not my Cuda is working my GPU is working or not how to do that in here there is a GitHub page which is including Cuda samples you need to clone this to your folder any folder I already downloaded to my home folder which is Coda samples I'm just entering to Performance let's say this this example let's try just open a terminal inside and make random make it's gonna give you an executable file which is large kernel parameter let's run it and see it's working fine or not yes test test is made some simple tests while on GPU and is working very fine so uh by this video we already learned how to install Cuda and we already installed properly in the future videos I'm gonna continue with the other kind of installations like opencv and kind of computer vision libraries maybe packages tools we are going to install buy one buy with this video at least we installed or find this Cuda see you in the next videos [Music]

Channel: Computer Vision Lab

Views: 15,911

Keywords: how to install cuda on ubuntu, ubuntu cuda installation, cuda installation in 2023, install cuda faster, ubuntu, computer vision, cuda, cuda-12.2, ubuntu-22.04, cuda capable gpus, cuda supported gpus, computer vision lab, cudnn, nvcuvid, ubuntu cuda, nvidia drivers cuda, gpu drivers, cuda toolkit, cuda toolkit installation on ubuntu, nvcc, nvidia-smi, cuda set-up, pytorch cuda, nvidia cuda, cuda geforce rtx 3060, linux

Id: 8i3BiWa5AZ4

Length: 8min 32sec (512 seconds)

Published: Mon Sep 04 2023