Install a comprehensive Python and GPU powered deep learning suite. The NVIDIA RTX tensor cores can accelerate deep learning tasks by up to 30x (3,000%) versus other popular developer laptops. This solution only applies to the Kubuntu Focus M1 and M2. The Focus XE does not have the required Nvidia hardware.
kfocus-001-condaPlease read the disclaimer before proceeding. This solution is updated regularly. If you have suggestions or requests, please write support@kfocus.org.
The Focus XE ships with OpenGL and Vulkan libraries and usually just works for gaming. The Focus M2 ships with CUDA, TensorRT, cuDNN, OpenGL, and Vulkan libraries installed. This enables sophisticated compute and gaming. However, adding repositories or packages can alter and break the libraries. You may check and reset them as shown in the code block.
kfocus-001-condaRun the app from App Menu > Kubuntu Focus > Anaconda GPU AI Suite Installer, or open a terminal and enter kfocus-001-conda. This installs an Anaconda suite that is professionally vetted, comprehensive, and GPU accelerated. You will be prompted to optionally install the following environments:
(GPU) Jupyter Notebook(GPU) MXNet(GPU) PyTorch(GPU) TensorFlow(GPU) Theanokfocus-001-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.
When running Deep Learning workloads, you may notice that the software reports less available RAM than the GPU actually has. Part of this is because of driver overhead, but also because the GPU uses some memory for your display. One can instead use the Intel GPU for the display and use the NVIDIA GPU for only Deep Learning or other computing tasks. To do this, use the GPU widget in the tool tray or NVIDIA Settings > Prime Profiles to select NVIDIA On-Demand and reboot. This mode, also called Hybrid Graphics Mode, does not significantly improve battery life, so we recommend using NVIDIA (Performance mode) or Intel (Power Saving Mode) as it suits your needs at other times.
You may run comprehensive TensorFlow benchmarks at the end of the kfocus-001-conda script. Please compare to the official results. All tests are current as of March 2021.
| RTX 2080 Super[1] | RTX 2070[1] | RTX 2060[1] | Titan X[2] | Titan XP W10[2] | Titan XP Ubuntu[2] | |
|---|---|---|---|---|---|---|
| Inference | 9,626 | 8,881 | 7,785 | 7,443 | 9,741 | 11,948 |
| Training | 9,767 | 9,295 | 8,671 | 7,908 | 10,375 | 12,922 |
| AI Score | 19,393 | 18,176 | 16,878 | 15,351 | 20,089 | 24,870 |
[1] Kubuntu Focus 2.2Kg, 20mm thick
[2] Unconstrained Desktop GPU
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.
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:
Ctrl-Alt-F2 simultaneously.sudo /bin/rm /etc/X11/xorg.conf, type your password when prompted.sudo systemctl restart sddm && exit.You should be able to sign-in normally.
This is a partial revision history. See the git repository for all entries.
2021-10-10 5728326e Reformat to 2-column2021-09-22 dc862884 Update link and headline colors2021-08-23 681261b4 Review and update codeblocks2021-08-19 b994dd79 Add keywords for hybrid graphics mode2021-03-07 e3c1285d Revise conda launch guide, update styles2020-09-28 5d089610 Update to M2 scores2020-06-10 c4ed9299 Restructure layout2020-05-25 581c6523 Expand benchmarks2020-05-20 d00295d7 Add Library check2020-05-18 8c4b5b82 Add GPU library fix2020-05-15 0684cd5c Add benchmark2020-02-18 a4d437c5 Add conda script2020-01-31 70b4aa40 First publicationWe 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.