How to build a 6-GPU mining rig from scratch? (Hardware guide)

GPU Selection and Compatibility Considerations

1. Choose GPUs with high hash rate per watt for Ethereum Classic or Ravencoin, such as the AMD RX 6700 XT or NVIDIA RTX 3060 v2 — both offer strong performance while maintaining reasonable power draw.

2. Avoid GPUs with known firmware lockouts or driver-level restrictions imposed by manufacturers for mining use cases.

3. Ensure all six GPUs share the same memory configuration and BIOS version to minimize instability during simultaneous operation.

4. Verify PCIe lane allocation support on the motherboard — some chipsets throttle bandwidth when more than four slots are populated.

5. Confirm physical clearance between cards; dual-slot coolers may require spacing brackets or riser cable offsets to prevent thermal stacking.

Power Supply and Electrical Infrastructure

1. A fully loaded 6-GPU rig typically draws between 1200W and 1800W under load, demanding a PSU rated at minimum 2000W with 80 PLUS Platinum certification.

2. Use a single high-capacity PSU rather than multiple units to avoid ground loop complications and inconsistent rail regulation.

3. Install dedicated 240V/30A circuits with NEMA 6-30 or 14-30 outlets in the mining location — standard 120V residential outlets cannot safely sustain continuous draw above 1200W.

4. Integrate industrial-grade PDU units with individual outlet monitoring to track per-GPU consumption and detect early signs of voltage sag or overheating.

5. Route PCIe power cables using flat ribbon designs to reduce airflow obstruction inside the chassis and improve cable management aesthetics.

Motherboard and Expansion Architecture

1. Select motherboards with at least six PCIe x1 slots and chipset-level support for bifurcation — ASRock H110 Pro BTC+, B450 BTC+, and X399 Taichi BTC+ remain widely used models.

2. Enable Above 4G Decoding and Resizable BAR in BIOS to ensure full GPU memory addressing and optimal kernel scheduling across devices.

3. Disable unused onboard peripherals including audio controllers, LAN chips, and USB 3.x root hubs to reduce background interrupt load and system latency.

4. Flash latest stable BIOS versions before installation — many BTC-series boards received critical fixes for multi-GPU enumeration stability after 2022.

5. Mount the motherboard on non-conductive standoffs and verify no metal contact occurs between PCB traces and chassis surfaces to prevent short circuits.

Cooling and Physical Rig Assembly

1. Build inside an open-air test bench or modified server chassis with front-to-back airflow paths — enclosed ATX cases induce thermal throttling even with six case fans installed.

2. Orient GPUs vertically using PCIe riser brackets angled at 25–30 degrees to maximize laminar airflow across heatsinks without recirculating hot exhaust.

3. Use thermal pads rated for ≥12W/mK conductivity on VRAM and VRM components — stock pads degrade rapidly under sustained mining loads.

4. Install ambient temperature sensors near each GPU intake and log readings via IPMI or external microcontroller to correlate fan speed adjustments with local thermal gradients.

5. Apply high-viscosity thermal paste (e.g., Thermal Grizzly Kryonaut) only to GPU dies — avoid spreading over memory chips unless using specialized memory-specific compounds.

Firmware and Driver Optimization

1. Flash custom VBIOS images that unlock undervolted memory timings and disable HDMI audio engines to reduce idle power leakage.

2. Deploy Linux-based mining OSes like HiveOS or RaveOS which include pre-tuned kernel modules for AMDGPU-PRO and NVIDIA drivers optimized for persistent compute workloads.

3. Disable Windows Fast Startup and hibernation features if running on desktop OS — these interfere with PCIe device reinitialization cycles during reboot sequences.

4. Configure GPU clocks using --mclock and --pl parameters in lolMiner or T-Rex binaries to maintain stable memory bandwidth without exceeding 95°C junction temperatures.

5. Monitor PCIe link width negotiation with lspci -vv -s [device_id] | grep Width to confirm all devices operate at x1 gen3 instead of falling back to gen2 due to signal integrity issues.

Frequently Asked Questions

Q: Can I mix AMD and NVIDIA GPUs in the same rig?A: No. Mixed-vendor setups cause driver conflicts, inconsistent memory mapping, and unpredictable crash behavior during DAG epoch transitions.

Q: Is PCIe 4.0 necessary for mining performance?A: No. Mining workloads are memory-bound and compute-intensive but not bandwidth-sensitive — PCIe 3.0 x1 provides sufficient throughput for all current algorithms.

Q: Why do some rigs fail to detect more than four GPUs?A: This usually results from insufficient PCIe lane count, incorrect BIOS settings for ARI/ACS, or faulty riser cables introducing signal noise above acceptable thresholds.

Q: Do I need a CPU cooler beyond stock?A: Yes. Even low-TDP CPUs like the Athlon 3000G generate heat under constant PCIe polling load — a 92mm tower cooler prevents thermal throttling of the PCIe root complex.