Support_
Docs.

We use a tiered delivery mechanism:
• M4 Standard Node: Usually automated within 10 minutes after order confirmation. 10 minutes after order confirmation.
• Non-standard/Custom: Involves additional hardware scheduling, usually completed within 3 hours for physical mounting and environment initialization. 3 hours for physical mounting and environment initialization.

M4 is suitable for light development and testing; M4 Pro features larger unified memory (up to 64GB) and higher memory bandwidth, making it the top choice for large-scale Xcode compilation and AI training.

Since MACGPU provides native bare-metal hardware, the specifications are fixed at delivery. If higher specs are required, you will need to order a new instance and migrate your data.

Hong Kong (HKG) is recommended for mainland China users; Singapore (SIN) for Southeast Asia; North American users can choose between USE and USW based on proximity.

Initial Password: To ensure privacy, the system does not distribute preset passwords. Your initial login credential is the password you set during MACGPU account registration. during MACGPU account registration .
Reset Password: If you forget your password, click 'Forgot Password' on the login page to reset it online after email authentication.

Yes. You have full, unrestricted root access to the physical node and can execute any sudo command, including disabling SIP or modifying kernel extensions.

Just like a local Mac, you can add multiple sub-accounts via the 'sysadminctl' command line or through the 'Users & Groups' settings in the VNC interface.

Yes. The web-based VNC is transmitted via an SSL-encrypted channel, ensuring your GUI operations are not hijacked or eavesdropped on over public networks.

Absolutely. As a native bare-metal machine, you can utilize 100% of the GPU compute power from the M4 series silicon. We fully support mainstream models like DeepSeek and Llama-3. DeepSeek and Llama-3 .

Supported via Thunderbolt 5 (TB5) physical-level parallelization for ultra-fast interconnects. If you have multiple instances in the same availability zone, you can build a symmetric private network of up to 80Gbps between nodes using TB5 links. Thunderbolt 5 (TB5) physical-level parallelization for ultra-fast interconnects. If you have multiple instances in the same availability zone, you can build a symmetric private network of up to 80Gbps between nodes using TB5 links.

MACGPU MACGPU provides genuine 100% dedicated physical bandwidth with unlimited traffic. 100% dedicated physical bandwidth with unlimited traffic.

Latency usually stems from local ISP routing. You can try switching the 'Network Acceleration Link' in the console.

We provide a 2-hour renewal grace period. If the lease is not renewed by then, the system will immediately execute physical erase commands.

We will replace your node with a brand-new one of the same specification free of charge and provide double the lease duration as compensation based on the downtime.

Due to the high cost of physical hardware deployment and environment wiping, the minimum rental period is currently 'Daily'.