Skip to main content
Microsoft
separator
https://catalogartifact.azureedge.net/publicartifacts/numericalai.cem-app-61185645-0854-4f39-b9d6-7286b75036db/image7_logo300x300216x216.png

CEM – GPU Electromagnetic Simulations with Palace

by Empirisch Tech GmbH

Run Palace electromagnetic simulations on NVIDIA A100/H100 GPUs in your own Azure subscription.

CEM turns Palace, the open-source 3D finite element solver for computational electromagnetics, into a ready-to-use web application on a GPU virtual machine in your own Azure subscription. Your engineers set up, check, run and evaluate electromagnetic simulations in the browser, without building a cluster, compiling solvers or moving data out of your tenant.

Who it is for

RF and microwave engineers, antenna designers, signal-integrity teams and superconducting quantum-hardware groups, in industry and research, who need full-wave 3D simulations: resonant modes and quality factors, S-parameters, time-domain pulses, capacitance and inductance matrices. IT teams get one self-contained deployment that they control.

Simulate

  • Start from 18 ready-to-run Palace examples in seven categories (antennas, waveguides, resonators, transmission lines, superconducting quantum hardware and more), or upload your own case: a Palace configuration with its mesh.
  • Every case is checked before it runs: the configuration against Palace's schema and rules, the mesh and its boundary attributes, and a dry run of Palace itself.
  • Choose 1, 2, 4 or 8 GPUs per job: one MPI rank per GPU, with NVIDIA cuDSS solving across the GPUs. Follow the Palace log while the job runs, cancel it, and download all results as a zip file.
  • View results in the browser: animated field views rendered and quality-checked automatically, and results dashboards for eigenmode and quantum-hardware cases (frequencies, quality factors, energy participation).
  • Optional AI assistance

    Connect your own model (Anthropic, Google Gemini, Azure OpenAI or another OpenAI-compatible API) for an AI review of your configuration, advice when a case fails its checks, suggested job names, and an analysis with a corrected configuration when a run fails. Without a key these features are off, and simulations never depend on them.

    Administer

    • Deploys from the Azure portal as a managed application: one VM with a data disk, a virtual network and an HTTPS endpoint.
    • User accounts with administrator and user roles, and limits for GPUs per job and run time.
    • Keep results on your own Azure Files share or on the VM's data disk.
    • Reachable only from the IP range you choose, over HTTPS; keys are stored encrypted on the VM. The application sends no data to the publisher.
    • On the 8-GPU sizes, NVIDIA Fabric Manager is installed on demand from NVIDIA after an administrator accepts NVIDIA's license.
    • Supported VM sizes

      • NC A100 v4: 1, 2 or 4 NVIDIA A100 80 GB GPUs
      • ND A100 v4 series: 8 NVIDIA A100 GPUs (40 GB or 80 GB) with NVSwitch
      • NC H100 v5: 1 or 2 NVIDIA H100 NVL GPUs
      • ND H100 v5: 8 NVIDIA H100 GPUs with NVSwitch
      • Pricing and requirements

        An hourly software fee per VM hour, shown before you deploy, plus the Azure resources you use (VM, disks, public IP address, optional storage). Stop (deallocate) the VM when it is idle: software and compute charges pause, disks remain billed. You need quota for the chosen GPU size in your region. One simulation runs at a time and uses the GPUs you select.

        CEM is proprietary software, licensed per VM hour through Microsoft Marketplace; it is not free to use. Palace is open-source software under the Apache License 2.0 (github.com/awslabs/palace) and keeps its license. NVIDIA, A100 and H100 are trademarks of NVIDIA Corporation.


        At a glance

        https://catalogartifact.azureedge.net/publicartifacts/numericalai.cem-app-61185645-0854-4f39-b9d6-7286b75036db/trailer_4015271936317105514_trailer.png
        /staticstorage/20261029.1/assets/videoOverlay_62a424ca921ff733.png
        https://catalogartifact.azureedge.net/publicartifacts/numericalai.cem-app-61185645-0854-4f39-b9d6-7286b75036db/image1_01cemfieldviewgpu.png
        https://catalogartifact.azureedge.net/publicartifacts/numericalai.cem-app-61185645-0854-4f39-b9d6-7286b75036db/image0_02cemcasechecksaireview.png
        https://catalogartifact.azureedge.net/publicartifacts/numericalai.cem-app-61185645-0854-4f39-b9d6-7286b75036db/image6_03cempalaceexampletemplates.png
        https://catalogartifact.azureedge.net/publicartifacts/numericalai.cem-app-61185645-0854-4f39-b9d6-7286b75036db/image2_04cemresultsdashboardmodes.png
        https://catalogartifact.azureedge.net/publicartifacts/numericalai.cem-app-61185645-0854-4f39-b9d6-7286b75036db/image5_05cemadminazurefilesstorage.png
        English (United States)
        Your Privacy Choices Opt-Out Icon Your Privacy Choices
        Consumer Health Privacy Sitemap Contact Us Privacy & Cookies Terms of Use About our ads Manage cookies