How do I log in to my GPU server?

Connect, check GPUs, and run your first workload.

  1. Connect to the bastion using the SSH config emailed at provisioning.
  2. From the bastion, SSH into your assigned node (hostname format gpu-N-IDCODE.apetops-us.com).
  3. Run nvidia-smi to confirm all GPUs are visible and driver version matches.
  4. Activate the pre-installed conda environment: source /opt/apetops/activate.

Common first-run checks:

  • nvidia-smi topo -m — verify NVLink topology.
  • ib_write_bw — quick InfiniBand smoke test (HPC nodes only).
  • df -h /mnt/shared — confirm your shared parallel filesystem mount.

Troubleshooting tip: if GPUs are not visible, reload the NVIDIA modules with sudo systemctl restart nvidia-persistenced and try again.

More in Technical Support

GPU Resource Calculator

Use our interactive calculator to estimate VRAM and GPU count for your workload.

VPN Access — How to Connect

Step-by-step VPN setup for accessing your ApeTops GPU servers.

What is a bastion host?

Using our hardened jump boxes to reach compute nodes securely.