RightNow AI logoRightNow AI<Preview>

Remote GPU Execution

Run CUDA code on remote GPUs through SSH

Quick Start

Remote GPU Connection Dialog

Connect to Remote GPU

  1. Connect to Remote GPU: Press Ctrl+Shift+G or click "Remote GPU: Connect" in command palette
  2. Choose Connection Type:
    • Quick Connect: Paste your SSH command
    • RunPod/Vast.ai/Lambda: Select your provider
    • Manual Setup: Enter server details
  3. Enter Connection Info: Provide your SSH details (Host/IP address, Username, SSH key or password)
  4. Start Using Remote GPU: Once connected, your code runs on the remote GPU automatically

Supported Cloud Providers

RunPod

  • Copy SSH command from RunPod dashboard
  • Paste in Quick Connect
  • RightNow AI handles the rest

Popular for on-demand GPU rentals with RTX 4090, A100, and H100 availability

Vast.ai

  • Get instance SSH details from Vast.ai console
  • Use Quick Connect or Manual Setup
  • Supports custom ports

Cost-effective marketplace for peer-to-peer GPU rentals

Lambda Labs

  • Use instance IP address
  • Username: ubuntu
  • Add your SSH key during instance creation

Enterprise-grade GPU cloud with A100 and H100 instances

Paperspace

  • Get SSH endpoint from Paperspace console
  • Username: paperspace
  • Connect via Manual Setup

Managed GPU instances with pre-configured CUDA environments

How to Use

Open Remote Terminal

SSH Terminal Remote

Press Ctrl+Shift+T to open SSH terminal to your remote GPU.

  • Run commands directly on remote server
  • Monitor GPU status with nvidia-smi
  • Install dependencies and configure environment
  • Debug compilation and runtime issues

Run CUDA Code

  1. Open your .cu file
  2. Click Build button in the editor
  3. Code automatically compiles and runs on remote GPU
  4. Profiling results display in real-time

Switch GPUs

Click GPU name in status bar to switch between local and remote GPUs:

  • Select different remote connections
  • Switch back to local GPU
  • Compare performance across different hardware

Connection Status

Remote GPU Status Bar

Status Bar Indicators

Look at the status bar for connection status:

  • Connected: Shows remote GPU name and connection status
  • Disconnected: Shows local GPU or "No GPU detected"
  • Connecting: Shows progress indicator
  • Error: Displays connection error with troubleshooting link

Advanced Configuration

SSH Key Authentication

Recommended for security:

  1. Generate SSH key pair: ssh-keygen -t ed25519
  2. Copy public key to remote server: ssh-copy-id user@remote-host
  3. Configure RightNow AI to use private key path
  4. Test connection with SSH key authentication

Custom Port Configuration

For non-standard SSH ports:

  • Use Manual Setup option
  • Enter custom port number (default: 22)
  • Common alternative ports: 2222, 2200, 22000
  • Verify firewall allows connection on specified port

Saved Connections

Save frequently used connections for quick access:

  • Give connections memorable names
  • Store multiple cloud provider configurations
  • Quick reconnect after disconnection
  • Manage saved connections in Settings

Tips and Troubleshooting

Best Practices

  • Use SSH keys instead of passwords for better security
  • Save connections for quick access to frequently used servers
  • Check GPU availability with nvidia-smi in terminal
  • Monitor GPU usage in status bar during execution
  • Keep remote CUDA Toolkit version aligned with local code

Troubleshooting

  • Can't connect? Check firewall settings and SSH server status
  • GPU not detected? Run nvidia-smi in remote terminal to verify GPU access
  • Slow connection? Check network speed and latency to remote server
  • Permission denied? Verify SSH key permissions and user access rights
  • CUDA not found? Ensure CUDA Toolkit is installed on remote server

Need more help? Check the Troubleshooting Guide or visit our Discord community.