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.
Please read the disclaimer before proceeding. We review and update guided solutions like this regularly. If you have suggestions or requests, please write support@kfocus.org.
kfocus-condaRun the app from App Menu > Kubuntu Focus > Anaconda GPU AI Suite Installer, or open a terminal and enter kfocus-conda (on 22.04 LTS) or kfocus-001-conda (on 20.04 LTS). 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 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 Conda script. Please compare to the official results. All tests are current as of May 2022.
| Results[1] | RTX 3080 Ti[2] | RTX 3070 Ti[2] | RTX 3060[2] | RTX 3080 Max-Q[3] | RTX 3070 Max-Q[3] | RTX 3060 Max-Q[3] | Titan XP W10[4] |
|---|---|---|---|---|---|---|---|
| Inference | 14,245 | 13,153 | 10,265 | 12,606 | 11,576 | 10,189 | 9,741 |
| Training | 16,034 | 14,782 | 11,806 | 13,109 | 12,090 | 10,609 | 10,375 |
| AI Score | 30,279 | 27,935 | 22,071 | 25,715 | 23,666 | 20,798 | 20,089 |
| Video RAM | 16 GB | 8 GB | 6 GB | 16 GB | 8 GB | 6 GB | 12 GB |
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.
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.
This is a partial revision history. See the git repository for all entries.
2021-11-07 3e32a343 Update table to current2021-10-10 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.