If this does not work, then just remove the folder manually. The uninstaller may not work if you have done ssh -X. Uninstallįor uninstalling CST, you have to run the specified uninstaller in: /opt/cst/CST_STUDIO_SUITE_2018/Uninstall\ CST_STUDIO_SUITE_2018/Uninstall\ CST_STUDIO_SUITE_2018 and also for GPU usage: watch -n 0.5 nvidia-smiĬST GUI and related services such as update seem to have problems running whenever IPv6 is deactivated. You can see the current calculation jobs: /opt/cst/CST_STUDIO_SUITE_2018/LinuxAMD64/CSTDCMainControl_AMD64 You can perform a system check: /opt/cst/CST_STUDIO_SUITE_XXXX/cst_system_checkĭisable error correction code for optimal performance: sudo nvidia-smi -i -e 0 You can check if CST sees your GPUs: /opt/cst/CST_STUDIO_SUITE_XXXX/LinuxAMD64/CSTHardwareInfo_AMD64 You can use the to stop/start/restart the DC-services, e.g.: /etc/init.d/lsb-cst-maincontroller startĬheck number of available GPU cards: nvidia-smi -L You can easily update the installation using service packs: sudo /opt/cst/CST_Studio_Suite_2021/update_with_supfile CST_S2_v2021_LIN_SP5_.supĪfter you are finished, you should unmount /mnt, but actually a reboot is much better. In my case it was the dummy frame buffer: sudo apt-get install -y xvfb If you see a missing library it could be intalled. Then run the check script: /opt/cst/CST_Studio_Suite_2021/cst_system_check The installation completed successfully without any warnings or errors. After successful installation you have: *** Installation successful *** It is recommended to use an alternative user who has the permissions to run and stop the computing node. The components can be installed by comma separated values, like 1,2,4. Many of the reported missing libraries are get actually installed by the GUI. You need the DC Controller and Solver Server, but installing the GUI is of no harm. It will ask you which component to install. You can just download the ZIP version for the Linu圆4 from 3DS website according to your license. Now comes the coda toolkit sudo apt install nvidia-cuda-toolkit If all goes well, you can run: nvidia-smi You can reboot here once, because if there is an old version in the memory and if you run nvidia-smi you will get: Failed to initialize NVML: Driver/library version mismatch
Nvidia version 470 nvidia + cuda installationĬheck if you see your GPUs: lspci | grep -i nvidiaįirst try to remove old versions sudo apt-get purge *nvidia* Installing 3DS CST Microwave Studio Suite Distributed Computing Node on Ubunntu
0 kommentar(er)
