Connecting Your GPU to GPUnex

Overview

GPUnex allows you to earn revenue by contributing your GPU hardware to a global compute marketplace. AI companies, researchers, and developers rent your GPU resources for machine learning training, inference, rendering, and other compute-intensive workloads. You maintain full ownership of your hardware while the GPUnex platform handles matchmaking, billing, and payments.

As a provider, you earn weekly or monthly payouts in USDC (on Solana) or via bank transfer. The platform takes care of customer acquisition, usage tracking, and payment processing — your role is to keep your hardware online and available.

This guide walks you through every step from initial application to receiving your first payout.

System Requirements

Before applying, make sure your hardware meets the minimum specifications. These requirements ensure a consistent, high-quality experience for customers renting GPU compute on the platform.

Supported NVIDIA GPU models include:

• NVIDIA H100 (80 GB)
• NVIDIA A100 (40 GB / 80 GB)
• NVIDIA L40S (48 GB)
• NVIDIA L4 (24 GB)
• NVIDIA RTX 4090 (24 GB)
• NVIDIA RTX A6000 (48 GB)

If your GPU model is not listed here, you can still submit an application. The GPUnex team evaluates new hardware on a case-by-case basis and regularly expands the supported model list.

Applying as a Provider

The provider application process allows GPUnex to review your hardware, location, and availability before onboarding you to the marketplace.

1. Log in to your GPUnex account and navigate to the Roles page from the main navigation menu.

2. Click “Apply to Provide GPUs”. This opens the provider application form.

3. Fill out your contact and organization details:

   • Full Name — Your legal name as the primary point of contact.
   • Email Address — A reliable email where the team can reach you regarding your application and ongoing operations.
   • Company Name (optional) — If you are applying on behalf of a business or datacenter, enter the company name here.

4. Add your GPU entries. For each GPU or group of identical GPUs you want to provide, fill in the following details:

   • GPU Model — Select the NVIDIA model from the dropdown list (e.g., H100, A100 80GB, L40S).
   • Quantity — The number of GPUs of this model you are offering.
   • System RAM — Total RAM available on the host machine for this GPU group.
   • Internet Speed — Your measured upload and download speed in Gbps.
   • Availability — Whether the hardware is available 24/7 or on a specific schedule.
   • Location — The city, country, or datacenter region where the hardware is physically located.

   You can add multiple GPU entries if you have different models or setups across different machines.

5. Provide additional infrastructure details:

   • UPS Status — Whether your hardware is protected by an uninterruptible power supply. UPS protection is strongly recommended and required for Enterprise tier.
   • Commitment Period — How long you plan to keep the hardware available on the platform (e.g., 3 months, 6 months, 12 months). Longer commitments improve your marketplace visibility.
   • Maintenance Notes — Any scheduled maintenance windows or known limitations. For example, “Planned 2-hour maintenance window on the first Sunday of each month.”

6. Submit your application. Click the Submit button. Your application will be reviewed by the GPUnex team. You will receive an email notification once your application is approved or if additional information is needed. Review typically takes 1 to 3 business days.

Installing the Docker Agent

Once your provider application is approved, you will receive an email with instructions to install the GPUnex Docker agent on your machine. The installation process takes approximately 5 minutes.

What the Agent Does

The GPUnex agent is a lightweight Docker container that runs on your host machine. It performs the following functions:

• Auto-detects your GPU hardware and reports the model, VRAM, driver version, and CUDA version to the platform.
• Manages workload isolation so that customer jobs run in fully sandboxed containers with access only to the allocated GPU resources. The agent does not have access to your host filesystem, personal data, or any system resources beyond the GPUs you have designated for the platform.
• Handles communication between your machine and the GPUnex orchestration layer, including job scheduling, health checks, and telemetry reporting.

Installation Steps

1. Ensure Docker is installed and running on your host machine. If Docker is not yet installed, follow the official Docker installation guide for your Linux distribution.

2. Ensure the NVIDIA Container Toolkit is installed. This is required so that Docker containers can access your NVIDIA GPUs. Verify by running:

   nvidia-smi
   docker run --rm --gpus all nvidia/cuda:12.4.0-base-ubuntu22.04 nvidia-smi

   Both commands should display your GPU information without errors.

3. Run the GPUnex agent installer. Execute the following command in your terminal:

   curl -fsSL https://get.gpunex.com/agent | bash

   The installer will pull the latest agent image, configure it for your system, and start the agent as a background service. During installation, you will be prompted to enter your GPUnex provider credentials (the API key or token provided in your approval email).

4. Verify the agent is running. After installation, confirm the agent is active:

   docker ps | grep gpunex-agent

   You should see the gpunex-agent container listed with a status of “Up”.

5. Check your provider dashboard. Log in to GPUnex and navigate to the Provider Dashboard. Your machine and GPUs should appear as “Online” within a few minutes of the agent starting.

Automated Benchmark

After the agent is installed and connected, the platform automatically initiates a benchmark sequence on your GPUs. This process is fully automated and requires no action on your part.

What the Benchmark Tests

• Compute throughput — Measures floating-point operations per second (FLOPS) across FP16, FP32, and tensor core operations.
• Memory bandwidth — Tests GPU memory read and write speeds to ensure VRAM is performing to specification.
• Network latency and throughput — Verifies that your network connection meets the minimum requirements for data transfer between the platform and your machine.
• Storage I/O — Checks SSD read and write speeds to confirm the storage subsystem can handle dataset loading without bottlenecking GPU utilization.

Benchmark Results

Results are displayed in your Provider Dashboard under the Benchmarks tab. Each GPU receives a performance score that is visible to potential customers browsing the marketplace. If any GPU falls below the minimum quality thresholds, it will be automatically delisted from the marketplace until the issue is resolved. You will receive a notification if this occurs, along with guidance on what needs attention.

Benchmarks are re-run periodically to ensure ongoing quality.

Setting Your Pricing

As a provider, you have full control over how much you charge for your GPU resources.

1. Navigate to the Provider Dashboard and open the Pricing tab.

2. Set your hourly rate for each GPU model you are offering. The rate is denominated in USDC per GPU per hour. You can adjust this rate at any time.

3. Set your availability schedule. If your hardware is not available 24/7, define the hours and days when your GPUs should be listed on the marketplace. Outside of these windows, your GPUs will automatically be hidden from customers.

4. Review platform pricing recommendations. The GPUnex platform provides market-based pricing suggestions based on current supply and demand for your GPU model in your region. You are not required to follow these recommendations, but pricing competitively increases the likelihood that your GPUs will be rented.

5. Understand the visibility impact. Higher uptime and competitive pricing result in better placement in marketplace search results. Providers with consistent availability and fair pricing receive more rental requests.

Enterprise vs Community Tier

GPUnex supports two provider tiers to accommodate a range of infrastructure setups.

Enterprise Tier is designed for professional datacenter operators and hosting companies. It comes with stricter requirements but offers premium marketplace placement and access to enterprise customer contracts.

Community Tier welcomes individual providers, small businesses, and home or office setups. Requirements are more flexible, and providers can contribute hardware on their own schedule. Community providers still benefit from full marketplace visibility and competitive earnings.

Your tier is determined during the application review process based on the infrastructure details you provide.

Earnings

Providers retain the majority of rental revenue generated by their hardware. The GPUnex platform applies a competitive platform fee that covers payment processing, customer support, and infrastructure costs.

Payout Schedule and Methods

• Weekly payouts — Available for providers who have earned a minimum threshold. Payouts are processed every Monday.
• Monthly payouts — Processed on the first business day of each month.
• USDC on Solana — Payouts are sent directly to your connected Solana wallet address.
• Bank transfer — Available in supported regions. Configure your bank details in the Provider Dashboard under the Payouts tab.

Tracking Your Earnings

Your Provider Dashboard includes a real-time earnings tracker that shows:

• Total revenue earned for the current period (week or month).
• Revenue per GPU broken down by individual GPU and model.
• Historical earnings with filterable date ranges.
• Pending payouts and payout history with transaction details.

What Affects Your Revenue

Your earnings depend on several factors:

• GPU model — Higher-end models like H100 and A100 command higher hourly rates.
• Availability — GPUs that are online and available more consistently generate more rental hours.
• Market demand — Demand for GPU compute fluctuates based on AI industry activity. During peak training seasons, rates and utilization tend to increase.
• Pricing strategy — Competitive pricing relative to other providers with the same GPU model increases your utilization rate.

Reliability Score

Every provider on GPUnex receives a reliability score that directly impacts marketplace visibility and the volume of rental requests you receive.

How the Score Is Calculated

The reliability score is based on three key metrics:

1. Uptime tracking — The percentage of time your GPUs are online and available compared to your stated availability schedule. Unplanned downtime reduces your score.

2. Job completion rate — The percentage of customer jobs that run to completion without interruption caused by provider-side issues (hardware failures, network outages, or unexpected shutdowns).

3. Benchmark consistency — Whether your GPUs continue to perform at the level measured during initial and periodic benchmarks. Degraded performance (for example, due to thermal throttling) can lower your score.

Why It Matters

• Higher score = higher marketplace visibility. Providers with strong reliability scores appear higher in search results when customers browse available GPUs.
• Higher score = more jobs. Customers and automated job schedulers prefer providers with a proven track record of reliability.
• Score recovery. If your score drops due to a temporary issue, it will recover over time as you demonstrate consistent uptime and successful job completions. The score uses a rolling window so recent performance carries more weight than older events.

You can view your current reliability score and its component metrics in the Provider Dashboard under the Performance tab.