Choosing a GPU for Gonka: Hardware Recommendations

Gonka requires an NVIDIA GPU with CUDA support and a minimum of 40GB VRAM per MLNode. This is a strict hardware limitation: the Qwen3-235B model with MoE architecture (22 billion active parameters out of 235 billion total) requires significant video memory to load weights and perform inference. AMD and Intel GPUs are not supported — Gonka uses NVIDIA's CUDA stack, including cuBLAS for matrix operations and cuDNN for neural network layers.

Full list of requirements:

• GPU: NVIDIA with CUDA, minimum 40GB VRAM. Cards with less memory (RTX 4090 with 24GB, RTX 3090 with 24GB) are not suitable — they physically cannot load model shards.
• CPU: AVX instruction set support is mandatory — without it, inferenced will not start (SIGILL at launch).
• RAM: 64GB+ RAM is recommended for comfortable operation, loading weights, and serving request queues.
• Disk: NVMe SSD with sufficient space — the full Qwen3-235B weight set occupies ~640GB, and fast loading from NVMe is critical for node cold start time.
• Internet: Minimum 100 Mbps stable connection — the node receives requests from Transfer Agents and sends results to clients in real-time.
• Uptime: 24/7 — skipping epochs reduces rewards, and prolonged downtime can lead to exclusion from the task pool.

Let's examine each recommended card in detail:

NVIDIA H100 80GB — the flagship of the current generation and the optimal choice for Gonka. TDP (thermal design power) 700 W, cost ~$25—35K per card. Supports FP8 inference, which speeds up request processing without quality loss. NVLink allows combining multiple H100s into a cluster with high-speed communication between cards. For a full Qwen3-235B cluster, 8 H100 cards are required (8 x 80GB = 640GB total VRAM). This is the most common configuration in the Gonka network.

NVIDIA H200 141GB — the next generation with almost double the memory. Increased VRAM allows processing more requests simultaneously (larger batch size), which increases GNK revenue per unit of time. HBM3e memory bandwidth is higher than H100 — faster weight loading, faster inference. For a Qwen3-235B cluster, 5 H200 cards are sufficient instead of 8 H100s, simplifying infrastructure.

NVIDIA A100 40/80GB — the previous generation, but still supported by the network. Price ~$10—15K per card — significantly cheaper than H100. Performance is lower: no FP8, slower HBM2e. The 40GB version is the minimally acceptable for Gonka, the 80GB version is preferred. A100 is suitable for entering the network with lower initial investments.

What is not suitable: consumer cards RTX 4090 (24GB), RTX 3090 (24GB), RTX 4080 (16GB) — insufficient VRAM for working with Qwen3-235B. Even the most powerful consumer card does not meet the minimum threshold of 40GB. For a full cluster (640GB VRAM), 8 H100 cards, 5 H200 cards, or 8—16 A100 80GB cards will be required, depending on the configuration.

The MLNode in the Gonka network is a server with a GPU that performs AI inference. Setting up a node involves several stages, each critical for stable operation and maximizing GNK earnings.

Software: The main component is the inferenced CLI, which manages model loading, request processing, and communication with the blockchain. Inferenced runs inside a Docker container, simplifying deployment and updates. A full Qwen3-235B configuration requires 640GB of total VRAM — for example, 8 H100 cards with 80GB each. Model weights (~640GB) are loaded from an NVMe SSD when the node starts.

Registration: After installation, the node is registered on-chain — it creates a record in the Gonka blockchain indicating its address, supported models, and characteristics (VRAM, bandwidth, location). From this point, Transfer Agents begin directing AI requests from users to the node.

Network operation: Each incoming request — a prompt from a user — is processed by the GPU via the neural network. The result is sent back through the Transfer Agent to the client. The Sprint consensus takes into account each completed computation when forming a block, and the reward is distributed proportionally to the amount of work. The node can publish updated characteristics in real-time — if the load increases, Transfer Agents will redirect some requests to less loaded nodes. Detailed setup instructions can be found in the mining guide.

If you don't have your own equipment, there are three ways to get GPU access for Gonka, each with a different balance of cost, complexity, and control:

Pools (from $100): Gonka.Top, GonkaPool.ai, Hashiro, Mingles CloudMine — operators rent GPUs, set up nodes, and monitor uptime. You receive GNK proportional to your contribution, without dealing with technical details. Ideal for beginners and passive investors.

Dedicated servers (from $12,000/month): Gonka.Top offers not only pools but also dedicated servers with full maintenance. You get a ready node — the operator handles inferenced setup, 24/7 monitoring, updates, and troubleshooting. Mining goes directly to your wallet — all GNK income is yours, minus a fixed rental fee.

Bare-metal rental: Spheron provides bare-metal servers with H100/H200 that you configure yourself. This path is for technical users familiar with Linux, Docker, and CLI. Maximum control, but also maximum responsibility for setup, uptime, and updates. A detailed comparison of all providers is on the “Get GNK” page.