Comprehensive reverse-engineered documentation for NVIDIA DGX systems.
This documentation is generated through forensic analysis of factory recovery images. It is not official NVIDIA documentation.
| Platform | Status | Source |
|---|---|---|
| DGX Spark | Active | FastOS 1.105.17 recovery image |
| Blackwell | Planned | - |
docs/
├── spark/
│ ├── README.md # Platform overview, quick reference
│ │
│ ├── hardware/
│ │ ├── README.md # Hardware architecture overview
│ │ ├── soc-th500.md # TH500 Thor SoC details
│ │ ├── gpu-gb10.md # GB10 Blackwell GPU (1CTA/sm_120)
│ │ ├── memory-architecture.md # LPDDR5X, unified memory, NVLink-C2C
│ │ ├── ec-mec-n1x.md # Microchip MEC N1X embedded controller
│ │ ├── tpm-nuvoton.md # TPM 2.0 implementation
│ │ ├── usb-pd-ccg8.md # Infineon CCG8DF USB-PD controller
│ │ ├── networking.md # ConnectX-7, WiFi 6E, Bluetooth 5.3
│ │ ├── storage.md # NVMe, eMMC boot
│ │ └── device-tree.md # DTB analysis, hardware topology
│ │
│ ├── firmware/
│ │ ├── README.md # Firmware chain overview
│ │ ├── boot-sequence.md # Power-on to Linux timeline
│ │ ├── socfw-cap.md # SoC firmware capsule format
│ │ ├── ec-firmware.md # EC firmware (Zephyr RTOS)
│ │ ├── ec-fused-vs-nofuse.md # Fuse state variants
│ │ ├── tpm-firmware.md # TPM capsule analysis
│ │ ├── usbpd-firmware.md # USB-PD controller firmware
│ │ ├── mlnx-connectx7.md # Mellanox NIC firmware
│ │ ├── capsule-format.md # UEFI capsule structure, signatures
│ │ ├── signing-chain.md # X.509 certificate hierarchy
│ │ └── update-mechanism.md # fwupd integration, rollback
│ │
│ ├── boot/
│ │ ├── README.md # Boot architecture overview
│ │ ├── uefi.md # UEFI firmware, variables
│ │ ├── secureboot.md # Secure Boot chain, MOK
│ │ ├── shim.md # SHIM 15.8 analysis
│ │ ├── grub.md # GRUB config, kernel params
│ │ ├── initrd.md # initramfs contents, hooks
│ │ ├── kernel-cmdline.md # Boot parameters
│ │ └── recovery-mode.md # FastOS recovery boot
│ │
│ ├── kernel/
│ │ ├── README.md # Kernel overview
│ │ ├── config.md # Kernel config analysis
│ │ ├── tegra241-cmdqv.md # Tegra CMDQV driver
│ │ ├── nvidia-modules.md # nvidia.ko, nvidia-uvm.ko, etc
│ │ ├── driver-580.md # Open driver 580.95.05
│ │ ├── device-drivers.md # Platform-specific drivers
│ │ └── patches.md # Ubuntu/NVIDIA kernel patches
│ │
│ ├── os/
│ │ ├── README.md # OS layer overview
│ │ ├── ubuntu-base.md # Ubuntu 24.04.3 Noble
│ │ ├── package-manifest.md # All 2,298 packages
│ │ ├── apt-repositories.md # DGX repo, NVIDIA repo, Ubuntu Pro
│ │ ├── systemd-services.md # Service dependency graph
│ │ ├── users-groups.md # System accounts, permissions
│ │ ├── filesystem-layout.md # Partition scheme, mount points
│ │ └── update-policy.md # APT pinning, unattended upgrades
│ │
│ ├── nvidia-stack/
│ │ ├── README.md # NVIDIA software stack overview
│ │ ├── cuda-13.md # CUDA 13.0 toolkit
│ │ ├── driver-architecture.md # Open vs proprietary, GSP
│ │ ├── nvml.md # NVIDIA Management Library
│ │ ├── persistenced.md # nvidia-persistenced
│ │ ├── container-toolkit.md # nvidia-container-toolkit, CDI
│ │ ├── fabric-manager.md # NVLink/NVSwitch (if applicable)
│ │ └── dcgm.md # Data Center GPU Manager
│ │
│ ├── services/
│ │ ├── README.md # DGX services overview
│ │ │
│ │ ├── oobe/
│ │ │ ├── README.md # OOBE system overview
│ │ │ ├── architecture.md # Component diagram
│ │ │ ├── state-machine.md # OOBE flow states
│ │ │ ├── oobe-service.md # Main Go service (port 80)
│ │ │ ├── oobe-admin.md # Admin service (D-Bus)
│ │ │ ├── oobe-desktop.md # Electron kiosk app
│ │ │ ├── hotspot.md # WiFi AP setup
│ │ │ ├── hostname-generation.md # Hostname algorithm
│ │ │ ├── bluetooth-pairing.md # HID device pairing
│ │ │ ├── network-config.md # Network provisioning
│ │ │ ├── user-creation.md # Initial user setup
│ │ │ ├── eula-flow.md # EULA acceptance
│ │ │ ├── first-run.md # Post-OOBE first login
│ │ │ └── dbus-interface.md # com.nvidia.dgx.oobe.admin1
│ │ │
│ │ ├── dashboard/
│ │ │ ├── README.md # Dashboard overview
│ │ │ ├── architecture.md # Go backend + React frontend
│ │ │ ├── api-reference.md # REST API endpoints
│ │ │ ├── authentication.md # Login, sessions
│ │ │ ├── gpu-telemetry.md # GPU metrics API
│ │ │ ├── system-info.md # System info API
│ │ │ ├── update-management.md # Software update API
│ │ │ ├── jupyterlab.md # JupyterLab integration
│ │ │ └── frontend.md # React app structure
│ │ │
│ │ ├── telemetry/
│ │ │ ├── README.md # Telemetry system overview
│ │ │ ├── consent-model.md # User/device consent
│ │ │ ├── sol-service.md # Sign of Life service
│ │ │ ├── endpoints.md # Phone-home URLs
│ │ │ ├── data-collected.md # What gets transmitted
│ │ │ ├── localized-config.md # Config API protocol
│ │ │ └── opt-out.md # How to disable
│ │ │
│ │ └── system/
│ │ ├── README.md # System services
│ │ ├── nv-scripts.md # /usr/local/sbin/nv_scripts/
│ │ ├── first-boot.md # First boot services
│ │ ├── power-management.md # Suspend/hibernate
│ │ └── platform-detection.md # get_platform_info.bash
│ │
│ ├── networking/
│ │ ├── README.md # Network architecture
│ │ ├── network-manager.md # NetworkManager config
│ │ ├── avahi-mdns.md # mDNS/Bonjour
│ │ ├── wifi.md # WiFi 6E configuration
│ │ ├── bluetooth.md # Bluetooth stack
│ │ └── connectx7.md # Mellanox NIC config
│ │
│ ├── security/
│ │ ├── README.md # Security model
│ │ ├── secure-boot-chain.md # Full chain analysis
│ │ ├── certificates.md # X.509 certs, NVIDIA CA
│ │ ├── tpm-integration.md # Measured boot
│ │ ├── apparmor.md # AppArmor profiles
│ │ ├── firewall.md # nftables rules
│ │ └── ubuntu-pro.md # ESM, Livepatch
│ │
│ ├── recovery/
│ │ ├── README.md # Recovery system overview
│ │ ├── fastos.md # FastOS installer
│ │ ├── usb-image-format.md # USB recovery image structure
│ │ ├── partition-scheme.md # GPT layout
│ │ ├── installer-flow.md # Installation sequence
│ │ └── factory-reset.md # Reset procedure
│ │
│ └── internals/
│ ├── README.md # Internal NVIDIA details
│ ├── codenames.md # Marlin, Wagyu, Jade, Yukon
│ ├── gitlab-paths.md # Internal GitLab structure
│ ├── artifactory.md # urm.nvidia.com paths
│ ├── confluence.md # Internal wiki references
│ ├── build-system.md # How FastOS is built
│ └── version-strings.md # All version identifiers
│
├── blackwell/ # Future: B200/B300/GB200
│ └── README.md # Placeholder
│
├── common/
│ ├── README.md # Cross-platform concepts
│ ├── capsule-format.md # UEFI capsule spec
│ ├── nvlink.md # NVLink protocol
│ └── gpu-architectures.md # Hopper to Blackwell evolution
│
└── glossary/
├── README.md # Terminology guide
├── codenames.md # Product to codename mappings
├── gpu-arch.md # 1CTA/2CTA, sm_xxx mappings
├── internal-urls.md # NVIDIA internal infrastructure
└── package-prefixes.md # dgx-*, nvidia-*, nv-* naming
- Create directory structure
- docs/spark/README.md - Platform overview
- docs/spark/hardware/* - Hardware documentation
- docs/spark/firmware/* - Firmware documentation
- docs/spark/boot/* - Boot chain documentation
- docs/spark/kernel/* - Kernel documentation
- docs/spark/os/* - OS layer documentation
- docs/spark/nvidia-stack/* - NVIDIA stack documentation
- docs/spark/services/oobe/* - OOBE documentation
- docs/spark/services/dashboard/* - Dashboard documentation
- docs/spark/services/telemetry/* - Telemetry documentation
- docs/spark/services/system/* - System services documentation
- docs/spark/networking/* - Network documentation
- docs/spark/security/* - Security documentation
- docs/spark/recovery/* - Recovery documentation
- docs/spark/internals/* - Internal NVIDIA documentation
- docs/glossary/* - Terminology glossary
- docs/common/* - Cross-platform documentation
- FastOS 1.105.17 recovery image (
dgx-spark-recovery-image-1.105.17.tar.gz) - Extracted partition mounted at
/mnt/fastos - artifact.json software BOM (2,298 packages)
- Firmware capsules (socfw.cap, ec_*.cap, tpm.cap, usbpd.cap)
- Kernel 6.14.0-1013-nvidia
- NVIDIA driver 580.95.05 (open)