Best settings for RTX 4090 mining? (Hashrate Max)

Power Limit and Voltage Tuning

1. Set the power limit to 320W using MSI Afterburner or EVGA Precision X1 — this value balances thermal headroom and electrical efficiency for sustained mining workloads.

2. Reduce core voltage to 750mV while maintaining stable operation; lower voltage decreases heat generation without sacrificing hash rate on ETH-like algorithms.

3. Lock core clock at 1200 MHz to prevent dynamic scaling fluctuations that introduce hashrate variance during DAG epoch transitions.

4. Disable GPU boost entirely to eliminate clock drift caused by temperature or power throttling during long-duration mining sessions.

Memory Overclocking Strategy

1. Increase memory clock to +1400 MHz on GDDR6X modules — this range consistently delivers optimal bandwidth utilization for Ethash and KawPoW kernels.

2. Apply +80 mV to memory controller voltage only if instability occurs at higher memory clocks; avoid exceeding this margin to prevent VRAM degradation over time.

3. Test memory stability with T-Rex Miner’s built-in memory stress test before deploying full-time, ensuring zero stale shares under load.

4. Disable memory thermal throttling in BIOS-level GPU firmware where supported — some ASUS and Gigabyte cards allow disabling memory junction temperature limits via VBIOS modding.

Cooling and Ambient Optimization

1. Maintain ambient room temperature below 22°C — every 5°C rise above this threshold correlates with measurable hashrate loss due to thermal backoff in memory subsystems.

2. Replace default thermal pads on VRAM and VRMs with 8W/mK graphite-based pads to improve heat transfer from memory dies to heatsink fins.

3. Use open-air test benches during tuning phases to isolate airflow constraints; enclosed cases reduce effective memory cooling by up to 18% under continuous load.

4. Configure fan curve to sustain 75% RPM at 65°C GPU junction temp — this prevents sudden fan ramp-ups that disrupt acoustic consistency in residential environments.

Driver and Software Stack Configuration

1. Install NVIDIA Game Ready Driver version 535.98 — this release contains specific optimizations for compute workloads on AD102 architecture GPUs.

2. Disable hardware-accelerated GPU scheduling in Windows Settings → System → Display → Graphics Settings to reduce kernel-mode latency spikes affecting share submission timing.

3. Run miner as administrator with real-time process priority to minimize OS-level scheduling interference during DAG generation cycles.

4. Use T-Rex Miner v0.26.11 with --ethstratum=ethproxy parameter to ensure compatibility with latest pool protocol updates and reduce rejected share rates.

Frequently Asked Questions

Q: Does increasing memory clock beyond +1400 MHz yield additional hashrate on RTX 4090?Increasing memory clock past +1400 MHz rarely improves Ethash hashrate meaningfully and often introduces instability requiring higher voltage, which accelerates memory wear.

Q: Can I use Linux instead of Windows for higher RTX 4090 mining efficiency?Linux offers marginal gains in driver overhead but lacks mature support for AD102 memory tuning interfaces; most stable configurations are validated on Windows 11 22H2 with WSL2 disabled.

Q: Is it safe to undervolt the GPU core below 750mV?Core undervolting below 750mV risks computational errors in DAG verification, leading to increased stale shares and potential block rejection by pools — stability testing must include 48-hour continuous validation.

Q: Why does my RTX 4090 report lower hashrate on NBMiner compared to T-Rex?NBMiner uses different CUDA kernel launch parameters optimized for older architectures; its AD102 implementation lags behind T-Rex’s custom memory access patterns, resulting in ~3–5% lower observed Ethash throughput.