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 the Kubuntu Focus M1 and M2 models, as they have the required Nvidia hardware.

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

Install tools for machine learning and AI with the 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 Start Menu > Kubuntu Focus Tools > Anaconda GPU AI Suite Installer, or open a terminal and enter kfocus-conda. This installs an Anaconda suite that is professionally vetted, comprehensive, and GPU accelerated. You will be prompted to optionally install the following environments:

Anaconda (required)
(GPU) Jupyter Notebook
(GPU) MXNet
(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 AI suite 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. The best solution is to click on Switch to Intel GPU (Power Save Mode). This will free all VRAM by using the iGPU and system RAM for the display. Only the primary laptop display may be available, although some models support a second display with DisplayPort.

If you need to drive more displays, you can 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

You may run comprehensive TensorFlow benchmarks at the end of the Kfocus Conda script. Please compare to the official results. All tests are current as of June 2023.

Results^[1,2,3]	RTX 4090 M^[2]	RTX 4080 M	RTX 4070 M	RTX 3080 Ti M	RTX 4060 M	RTX 3070 Ti M	RTX 3060 M
Inference	26,263	21,077	13,343	14,245	11,843	13,153	10,265
Training	31,791	25,201	15,872	16,034	14,164	14,782	11,806
AI Score	58,054	46,277	29,215	30,279	26,007	27,935	22,071
Video RAM	16 GB	12 GB	8 GB	16 GB	8 GB	8 GB	6 GB

[1] See official comparison chart

[2] All 40-series from Kubuntu Focus M2^{GEN 5}

[3] All 30-series from Kubuntu Focus M2^{GEN 4}

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 an xorg.conf file. Doing so may disable your internal display from functioning with a GUI upon logout or restart, and you will not be able to sign-in. 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.

You should be able to sign-in normally.

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

A: The Focus M2 ships with OpenGL and Vulkan libraries installed. When you run the Kfocus Conda script, you may be prompted to download and install additional libraries such as CUDA and CuDNN. Please ensure you are connected to the Internet with good bandwidth, as these can be quite large.If you still receive these errors, you can cancel, run the FocusRx diagnostics shown below, 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.

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
2023-04-16 1bc15058 Update display advice concerning iGPU Intel
2022-06-11 b304575d Change spelling from MaxQ to Max-Q
2023-05-18 0e693861 Refine opening language
2023-05-18 1a4bd7dc Added M2 GEN 4 details in benchmark
2023-05-18 1a4bd7dc Move focusrx to troubleshooting
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.