Hardware Requirements
The minimum and recommended hardware to run a ParalonCloud provider node — GPU, VRAM, RAM, disk, and internet.
These are the specs your machine needs to run a provider node and host rentals. You don't need a data-center rig — a single modern consumer GPU is enough to get started.
At a glance
| Component | Minimum | Recommended |
|---|---|---|
| GPU | NVIDIA, 8 GB VRAM | 24 GB+ (RTX 4090 / data-center) |
| GPU driver | Supports CUDA 12.4 | Latest stable driver |
| System RAM | 16 GB | 32 GB+ |
| Free disk | 50 GB (SSD) | 100 GB+ SSD/NVMe |
| Internet | 100 Mbps | Higher upload, wired |
GPU & VRAM
- An NVIDIA GPU is required — consumer cards (RTX 3090, 4090, 5090) through data-center GPUs (A100, H100, H200) are all supported.
- Minimum 8 GB of VRAM. More VRAM is better on two fronts: it lets your node host larger models and bigger rentals, and your incentivized-testnet points scale directly with VRAM (1 GB = 1 point per minute — see Rewards & Earnings).
- Your GPU driver must support CUDA 12.4 or newer. Most up-to-date NVIDIA drivers do.
System RAM
- Minimum 16 GB. The agent and each rental container run alongside the GPU workload, so the host needs headroom beyond what the GPU uses.
- 32 GB+ is recommended if you want to host larger or multiple sessions comfortably.
Disk
- Minimum 50 GB of free space, on an SSD. This covers the agent, the rental images that are pre-pulled (the Jupyter environment is several GB), and the renter's working data during a session.
- 100 GB+ on SSD/NVMe is recommended — faster disk means faster image pulls and a snappier experience for renters.
Internet
- Minimum 100 Mbps. Upload bandwidth matters most — renters pull images and data through your connection, so your upload speed is what they feel.
- A stable, wired connection is strongly preferred over Wi-Fi. The node connects out through a secure tunnel, so no port forwarding or public IP is required.
Verify your GPU
Once you have Docker and the NVIDIA Container Toolkit installed (see Prerequisites), confirm your GPU is visible to containers with a CUDA 12.4 image:
docker run --rm --gpus all nvidia/cuda:12.4.0-base-ubuntu22.04 nvidia-smi
If your GPU shows up in the output, your hardware is ready. Next, check the Prerequisites for the software you need, then Add a Node.