Kubuntu Focus: AI Deep Learning Guided Solution

From 5 to 30 Times Faster

AI Deep Learning

Install and run Anaconda, TensorFlow, PyTorch, and an AI benchmark with a single command. Enjoy NVIDIA RTX, CUDA, and CuDNN-powered performance 5 to 30 times greater than popular Unix® laptops. Develop with the industry-standard PyCharm IDE using the expertly packaged JetBrains Toolbox and desktop integration. This solution only applies to Focus models with NVIDIA GPUs.

Run kfocus-conda
Maximize VRAM with Intel GPU
Benchmarks
Troubleshooting
Revisions
Disclaimer

Install tools for machine learning and AI with the Anaconda GPU AI Suite Installer.

Please read the disclaimer before proceeding. We review and update guided solutions regularly. If you have suggestions or requests, please write support@kfocus.org.

Run `kfocus-conda`

Run the app from either the curated app icon, Start Menu > Kubuntu Focus Tools > Anaconda GPU AI Suite Installer, or open a terminal and enter kfocus-conda. This installs a GPU-accelerated and professionally vetted Anaconda suite. It will ask you if you'd like to install the following environments:

Anaconda (required)
(GPU) Jupyter Notebook
(GPU) PyTorch
(GPU) TensorFlow
(GPU) Theano

Kfocus Conda will not overwrite or delete data unless you tell it to do so. You may run it again to add environments as needed. For example, you may install only TensorFlow at first and then add PyTorch later. Run the benchmark at the end to compare to the results shown in the Benchmarks section.

When you wish to build on these Conda environments, we recommend using Pycharm. You may install this with the JetBrains Toolbox installed on all Kubuntu Focus laptops.

Maximize VRAM with Intel GPU

You can change the GPU used for your displays using the GPU Widget found in the system tray. When you use Nvidia GPU (Performance Mode), every display pixel consumes additional VRAM. Total display VRAM may exceed 2 GB. Run watch nvidia-smi in a terminal to see VRAM use.

Sometimes you may want to use all VRAM for a pipeline so it can complete or run faster. In this case, you need both the Nvidia drivers installed and the display to use system ram instead. The best choice for this is to switch to Hybrid GPU (On-Demand Mode) as shown here. Remember to switch back when you are done.

Configure GPUs and free up VRAM for intensive tasks with Plasma Optimus.

Benchmarks

We tested using Stable Diffusion 1.5 and Qwen3-0-6b tests. The TSMC “4N” lithography and fifth-generation Tensor cores make the 50-series GPUs faster, cooler, and more efficient. This is the same test you can run at the end of the KFocus Conda (kfocus-conda) script.

Benchmark	RTX 5090 M	RTX 5080 M	RTX 5070 Ti M	RTX 4090 M	RTX 4080 M	RTX 4070 M
Stable Diffusion 1.5	143 Images	116 Images	92 Images	127 Images	118 Images	80 Images
Qwen03-0-6b	544 Gen	534 Gen	482 Gen	579 Gen	528 Gen	459 Gen
Video RAM	24 GB	16 GB	12 GB	16 GB	12 GB	8 GB

Troubleshooting

Q: Starting NVIDIA Visual Profiler results in an error. How can I fix this?

A: One must install Java 8 and then call nvvp with the correct path.

# Confirmed on 17 May 2022 sudo apt install openjdk-8-jdk nvvp -vm /usr/lib/jvm/java-8-openjdk-amd64/jre/bin/java

Q: I booted my Kubuntu Focus, and it never got past the splash screen. How do I fix this?

A: NVIDIA Settings and other applications may ask you to save the xorg.conf file. This action could disable your internal display after logout or restart, blocking your login. If this happens, we recommend you remove this file:

Press [CTRL][ALT][F3] simultaneously.
Sign-in with your username and password.
Type sudo /bin/rm /etc/X11/xorg.conf, type your password when prompted.
Type sudo systemctl restart sddm && exit.

After you have completed the above steps, you should be able to log in normally.

Q: I see some library errors when installing Anaconda. What can I do?

A: Focus systems with NVIDIA GPUs ship with OpenGL and Vulkan libraries installed. When you run the KFocus Conda script, it will prompt you to download and install additional libraries, such as CUDA and CuDNN. Ensure you have a fast internet connection, as the download can be quite large. If you still receive these errors, you can cancel, run the FocusRx System Repair tool, and then try again.

# Run the package-check script when prompted focusrx -f

Revisions

This is a partial revision history. See the git repository for all entries.

2025-11-15 106a1e68 Minor grammar fixes, ensure uniform page title
2025-11-15 106a1e68 Switch to the latest AI tests, hide the old AI bench
2025-04-19 274c0ff1 Update grammar and content per audit
2025-04-06 a2c572b2 Update language and add icon
2025-02-26 2eaebe3b Normalize app names, descriptions, and keywords; fix typos, and minor issues
2024-12-06 35123b93 Update reinstall of hints widget
2024-10-04 7b76e6ca Update deep scores, battery reset
2024-09-10 73fcff02 Finalize Lightbox and doc update
2024-09-05 14887652 Major Lightbox updates, fix some out-of-date documentation
2024-08-23 297b11df Fix image href; update to bold orange text
2024-08-19 57a2389d Add white theme
2024-07-28 ee7b4601 Update metadata content, collapsable sections, lightbox
2024-07-26 7814a97d Sync primary images between guided solutions slideshow and articles
2024-07-16 e9155e41 Merge master link fixes
2024-07-07 39d71fcd Add Git hash to revisions
2024-06-06 ad0fcf25 Update for Noble
2024-03-18 fb9b592d Add syntax highlighting for codeblocks
2023-11-18 e770b1c6 Add search and help bar
2023-10-18 fac730e3 Enhance VRAM usage advice
2023-06-19 a7af817b Update RTX 40-series benches, add image, revise VRAM text
2023-05-18 1a4bd7dc Move focusrx to troubleshooting
2023-05-18 1a4bd7dc Added M2 GEN 4 details in benchmark
2023-05-18 0e693861 Refine opening language
2023-04-16 1bc15058 Update display advice concerning iGPU Intel
2022-06-11 b304575d Change spelling from MaxQ to Max-Q
2022-01-16 3ae944e8 Fix HTML Entities; replace dist-upgrade with full-upgrade
2021-11-07 3e32a343 Update table to current
2021-10-10 5728326e Reformat to 2-column
2021-09-22 dc862884 Update link and headline colors
2021-08-23 681261b4 Review and update codeblocks
2021-08-19 b994dd79 Add keywords for hybrid graphics mode
2021-03-07 e3c1285d Revise conda launch guide, update styles
2020-09-28 5d089610 Update to M2 scores
2020-06-10 c4ed9299 Restructure layout
2020-05-25 581c6523 Expand benchmarks
2020-05-20 d00295d7 Add Library check
2020-05-18 8c4b5b82 Add GPU library fix
2020-05-15 0684cd5c Add benchmark
2020-02-18 a4d437c5 Add conda script
2020-01-31 70b4aa40 First publication

Disclaimer

We try hard to provide a useful solution validated by professionals. However, we cannot anticipate every situation, and therefore cannot guarantee this procedure will work for your needs. Always backup your data and test the solution to determine the correct procedure for you.

THIS SOLUTION IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOLUTION, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.