From 5 to 30 Times Faster

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.

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.

Run kfocus-conda

Run 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:

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.

M2 GEN 5 17.3"

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. But So please remember to switch back when you are done!

System Tray GPU Widget

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 MRTX 4070 MRTX 3080 Ti MRTX 4060 MRTX 3070 Ti MRTX 3060 M
Inference23,600*21,07713,34314,24511,84313,15310,265
Training28,220*25,20115,87216,03414,16414,78211,806
AI Score51,830*46,27729,21530,27926,00727,93522,071
Video RAM16 GB12 GB8 GB16 GB8 GB8 GB6 GB

* Estimate

[1] See official comparison chart

[2] All 40-series from Kubuntu Focus M2GEN 5

[3] All 30-series from Kubuntu Focus M2GEN 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 Kubuntu 20.04 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:

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.

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.

HAVE QUESTIONS? Call 844-536-2871 or write
TellMeMore@kfocus.org | GET FOCUS GEAR

KFocus Logo