DePIN.as Node Operator Briefing: Multi-Mac Thunderbolt Racks & Exo Integration
Multiply your node earnings by clustering two or more Apple Silicon Macs into a single high-memory compute rack. Powered by the Exo open-source engine, Orchard v1.3.7 pools unified memory across Thunderbolt links to host high-parameter models like Qwen 3.8 27B.
Why Build a Thunderbolt Compute Rack?
Access Premium Revenue Tiers: Single 8 GB / 16 GB Macs are restricted to standard 3B models. Pooled 32 GB+ racks unlock high-yield jobs (qwen3_27b and qwen3-vl-32b) at elevated per-token payout rates.
Low-Latency Tensor Parallelism: Point-to-point Thunderbolt links (40–80 Gbps) enable ring-allreduce execution with near-zero inter-chip overhead.
Zero Upgrades Required: Combine existing hardware (e.g., an M4 Mac mini + M4 MacBook Air) into one enterprise-grade node without buying heavy server hardware.
Hardware & Software Requirements
| Component | Specification |
|---|---|
| Compute Nodes | 2× Apple Silicon Macs (M-Series) with ≥32 GB aggregate unified memory |
| Physical Interconnect | 1× Thunderbolt 4 or Thunderbolt 5 cable |
| Software Engine | Orchard Worker v1.3.7+ (with embedded Exo engine) |
| Network Link | Thunderbolt Bridge (macOS auto-assigns IP over Thunderbolt) |
Step-by-Step Operator Setup
- Physical Cable Connection: Plug the Thunderbolt cable directly into a Thunderbolt port on each Mac. Navigate to System Settings > Network to confirm the Thunderbolt Bridge status is Active.
- Launch Orchard v1.3.7: Open the Orchard application on both target machines.
- Enable Cluster Mode: In the Orchard settings panel on both nodes, toggle OrchardClusterMode to ON.
- Peer Discovery: Orchard automatically detects the Exo peer link over the Thunderbolt bridge, pools available unified memory, and registers the pair as a unified cluster.
- Verify Rack Status: Check your operator dashboard in the Fleet Observatory. Your setup will display a RACK ×2 badge with aggregate RAM available for premium dispatch.
Task Routing & Earnings Overview
Single-Mac Nodes: Automatically assigned standard batch workloads (llama-3.2-3b, qwen2.5-vl-3b).
Thunderbolt Racks (32 GB+): Prioritized for high-value reasoning, long-context extraction, and complex coding jobs (qwen3_27b, trellis-2.4b).
For technical support or cluster troubleshooting, join the DePIN.as operator channel at info@depin.as.
Love,
Martin