You will also need to download some pretty huge packages along the way.Īlso, make sure that you have installed the latest Nvidia driver after downloading the official driver from the official website of Nvidia. So make sure you're properly connected to the internet and have consistent electricity. This whole process can take a lot of time. Step 1: Make Sure You Have Solid Internet and Electricity I will be using Windows 11 Pro (Version 22H2) and WSL2 (of course!). Monitor: MSI Optix G24 Gaming Curved 75Hz.Motherboard: Gigabyte B450M Aorus Elite.GPU: Zotac Nvidia GeForce RTX 3050 8GB GDDR6.Processor: Ryzen 5 3500X 6 Core and 6 Threads.If you are also interested in the current specification that I used for this task, here you go: So, here we go! 🎉 My Computer Specificationįor this guide, I have used my Desktop workstation. I am writing the same handbook here as well. Still, after researching a lot, I have found out that WSL2 should work just fine.Īfter trying for more than a few days, I have successfully managed to set up everything necessary and can use my graphics card's CUDA cores in my Windows machine! An interesting thing is, in this process, you do not need to download or use Microsoft Visual Studio 2022 and download huge 30/35GB files just to install the recommended compilers and so on.īecause of this, I wrote a complete handbook on my GitHub (here's the repo: CUDA-WSL2-Ubuntu, and here's the website: /CUDA-WSL2-Ubuntu). There was some gap in the latest version of PyTorch with the Windows 11 kernel in CUDA. I followed a lot of videos but couldn't implement it after trying many times, unfortunately. And yes, Linux is the best for server-related stuff).Īlso, if you own the latest Nvidia GPU, then you're probably already familiar with the hassle regarding the graphics driver and so on.įor all these reasons, I was thinking about trying something different: utilizing the new Windows 11 operating system to use the CUDA cores from my Graphics Card. But, I do not like Linux as a desktop operating system (do not get offended, as it is my personal preference.

If you can afford a good Nvidia Graphics Card (with a decent amount of CUDA cores) then you can easily use your graphics card for this type of intensive work.Ī lot of developers use Linux for this. But you might wonder if the free version is adequate. If you are learning machine learning / deep learning, you may be using the free Google Colab.

Deep Work for windows instal