Education Nerdaxe Products

NerdQaxe++ Revision 7: Full Breakdown, Benchmarks & VRM Redesign

NerdQaxe++ Rev 7 board showing TPS546D24A two-phase VRM, LilyGO LCD at 7262.5 GH/s, and Assembled in the USA badge

The NerdQaxe++ Revision 7 is the latest iteration of the open-source quad-chip Bitcoin solo miner. It delivers ~4.800 TH/s at ~69.5 watts (~14.48 J/TH) at stock settings. The Nerd*OS firmware on Rev 7 unlocks overclocking up to 1000 MHz, pushing hashrate to ~8.1 TH/s with an upgraded power supply.

The defining change in Revision 7 is a redesigned voltage regulator circuit. The three-phase 95472MC configuration used since the original NerdQaxe++ has been replaced with a two-phase TPS546D24A layout adapted from the Bitaxe 801 GT. The new topology runs half the current at twice the voltage, improving power delivery stability, thermal performance, and long-term component reliability.

This article covers the full engineering history of the NerdQaxe++ platform: what each revision changed, why, and how Rev 7’s power delivery redesign opens new performance territory.


NerdQaxe++ Revision History: How Each Generation Improved

Open-source hardware development is iterative. Each NerdQaxe++ revision addressed specific engineering constraints identified by the community and home miners who run these devices daily. Understanding the full revision history provides context for why Rev 7’s changes matter.

Rev 5 and Earlier: The Foundation

The original NerdQaxe++ established the quad-chip BM1370 platform at approximately 4.8 TH/s. It proved that four Bitmain S21 Pro ASICs could run on a single compact PCB for desktop Bitcoin mining. However, the power delivery system had several constraints that limited reliability under sustained operation:

  • Mechanical fuse on the 12V rail: A one-shot fuse rated at 8A would blow during current spikes & overclocking, in some cases requiring physical replacement to restore operation.
  • 8A barrel jack connector: Voltage sagged below 12V when clock speeds exceeded 700 MHz, reducing hashrate stability.
  • Thin copper traces: The power path between the buck converter and rail generated excess heat under load.
  • Temperature sensor placement: Sensors were positioned on the PCB surface rather than near the ASIC cores, reading approximately 10°C below actual chip temperatures. This caused firmware thermal protections to respond late.

These constraints were well-documented by the community. PlebBase’s April 2025 review measured Rev 5 drawing 86 to 88W at stock 600 MHz settings (the UI reported 77W; a smart meter read 86.7W), working out to approximately 18 J/TH. VRM temperatures exceeded 70°C and the power supply peaked at 97W before any overclocking was applied.

Rev 6 and 6.1: Power Path Overhaul

Revision 6 was a substantial redesign of the power delivery path. It addressed every major constraint from Rev 5:

  • XT30 power connector: Rated for 15A with minimal voltage drop, replacing the 8A barrel jack.
  • Fuse-free PCB design: Overcurrent protection moved to the power supply side. No more single-use fuses on the board.
  • 1oz copper on a 2mm FR-4 PCB: Wider, thicker traces reduced heat in the power path by approximately 21%.
  • Spring-mounted heatsink: Uniform pressure across all four BM1370 ASICs for consistent thermal contact.
  • Relocated temperature sensors: Positioned closer to ASIC cores for accurate readings, enabling proper firmware thermal management.

The results were significant. At stock settings (600 MHz / 1150mV), Rev 6.1 delivered 4.82 TH/s at 70.6W (14.66 J/TH) with ASIC temperatures at 48.1°C and VRM temperatures at 41.2°C. Overclocked to 800 MHz (the firmware ceiling at the time), it reached 6+ TH/s at 103W with VRMs at 54.4°C.

Rev 6.1 retained the original three-phase 95472MC voltage regulator configuration. This was a known design choice. The 95472MC performed within specification at stock settings and moderate overclocks. However, the open-source community identified that the three-phase topology had less thermal margin than alternative configurations when pushed to higher sustained frequencies, particularly in environments with limited airflow or elevated ambient temperatures.

Rev 7: Voltage Regulator Redesign

Revision 7 targets the voltage regulator circuit specifically. The manufacturing & development team applied the two-phase TPS546D24A configuration that had proven reliable in the Bitaxe 801 GT, adapting it for the NerdQaxe++ quad-chip layout.

The engineering rationale: the TPS546D24A two-phase design delivers equivalent power to four BM1370 ASICs using half the current at twice the voltage compared to the 95472MC three-phase system. Lower current per phase means less resistive heating in the VRM circuit, better voltage regulation under transient loads, and improved long-term reliability.

The Nerd*OS firmware for Rev 7 unlocks overclocking up to 1000 MHz (up from the 800 MHz firmware ceiling on Rev 6.1). This was only possible because the new VRM provides sufficient power headroom and thermal stability at the higher frequencies and voltages required.

Side-by-side comparison of NerdQaxe++ Rev 7 and Bitaxe 801 GT boards showing the shared TPS546D24A two-phase voltage regulator design


TPS546D24A Thermal Limits: How Much Headroom Does Rev 7 Have?

According to the official Texas Instruments TPS546D24A datasheet (SLUSDN0A), the chip is rated for an operating junction temperature (TJ) of -40°C to 150°C. The absolute maximum storage temperature is -55°C to 150°C. Each device is a 40-A integrated buck converter in a 7mm × 5mm package.

This is significant context for the benchmark data. At the maximum 1000 MHz overclock, Rev 7’s VRM temperature reads 70°C to 74°C. The TPS546D24A reports its internal die temperature directly through PMBus telemetry; there is no separate external sensor. That measured die temperature leaves a substantial margin below the 150°C junction temperature rating, meaning the TPS546D24A is operating well within its specified safe range at full overclock.

TPS546D24A Thermal Specification Rating
Operating Junction Temp (TJ) -40°C to 150°C
Storage Temp (Tstg) -55°C to 150°C
Rated Current Output 40 A per device
Rev 7 VRM Temp at 1000 MHz (measured) 70°C to 74°C
Margin to Max Junction Temp ~76°C to 80°C below rating

For comparison, Bitmain typically caps the BM1370 ASICs at around 75°C PCB temperature, and Rev 7 holds them between 60°C and 67°C across the entire benchmark range. Both the VRM and the ASICs operate with comfortable thermal margin, which is exactly what the two-phase TPS546D24A design was selected to provide.


NerdQaxe++ Revision 7 Overclocking Benchmarks (Factory Configuration)

Solo Satoshi tested the NerdQaxe++ Revision 7 across nine frequency tiers from stock (600 MHz) to the firmware maximum (1000 MHz) using factory components including a Noctua NF-A8 rear fan oriented to blow cool air toward the board. All benchmarks were conducted at 73°F ambient temperature with automatic fan control targeting 60°C ASIC temperature.

Frequency Voltage Hashrate Power Efficiency ASIC Temp VRM Temp Fan
600 MHz (Stock) 1150mV 4.800 TH/s 69.5 W 14.48 J/TH 60.0°C 57.0°C 37%
650 MHz 1170mV 5.300 TH/s 76.3 W 14.40 J/TH 60.0°C 57.0°C 38%
700 MHz 1180mV 5.700 TH/s 82.0 W 14.39 J/TH 60.0°C 58.0°C 43%
750 MHz 1200mV 6.100 TH/s 89.0 W 15.00 J/TH 60.0°C 59.0°C 51%
800 MHz 1240mV 6.500 TH/s 94.0 W 15.46 J/TH 60.0°C 59.0°C 59%
850 MHz† 1250mV 6.900 TH/s 108.0 W 15.65 J/TH 60.0°C 61.0°C 86%
900 MHz† 1280mV 7.300 TH/s 117.0 W 16.03 J/TH 61.0°C 62.0°C 100%
950 MHz† 1300mV 7.700 TH/s 127.0 W 16.49 J/TH 63.0°C 66.0°C 100%
1000 MHz† 1330mV 8.100 TH/s 140.0 W 17.28 J/TH 66.0°C 70.0°C 100%

Test station: 73°F ambient, Thermalright AXP90-X36 Black stock heatsink, Thermal Grizzly Kryonaut Extreme Thermal Paste, factory front fan, Noctua NF-A8 rear fan (oriented to blow cool air toward the board), auto fan targeting 60°C. Mean Well LRS-350-12 PSU used for all tiers (the included 124W PSU is rated for 99.2W continuous at 80% derating).

†Requires upgraded PSU. The included power supply cannot sustain these power levels. A Mean Well LRS-350-12 or equivalent is recommended for operation above 800 MHz.

*Overclocking Disclaimer: Overclocking is performed at the end user’s own risk and is subject to the silicon lottery. Overclocking may void warranty coverage if damage occurs due to excessive voltage, frequency, or temperatures. Solo Satoshi is not responsible for damage resulting from user-modified settings. Always monitor temperatures and ensure adequate cooling before adjusting settings beyond factory defaults. All data provided in this benchmark table is subject to ±15%.

Key Observations from Benchmark Data

  • Efficiency sweet spot: 700 MHz at 14.39 J/TH delivers the best efficiency in the entire table. Miners optimizing for cost per terahash at low electricity rates should target this tier.
  • Auto fan control holds 60°C flat from 600 to 850 MHz. The fan scales from 37% to 86% across these tiers without the ASIC exceeding its target. Thermal management only breaks the 60°C target at 950 MHz (63°C) and 1000 MHz (66°C).
  • VRM stays under 60°C through 800 MHz on factory cooling. This is the same frequency that was Rev 6.1’s firmware ceiling, running significantly cooler on Rev 7’s TPS546D24A.
  • The PSU becomes the bottleneck at 850 MHz. At 108W draw, you exceed the included PSU’s 99.2W continuous rating. Every tier above 800 MHz requires a Mean Well LRS-350-12 or equivalent.
  • Fan hits 100% at 900 MHz. This is the stock cooling ceiling. Above this tier, ASIC and VRM temperatures rise because the fan has no additional speed to give. Upgrading from the stock Thermalright AXP90-X36 to the AXP90-X47 full copper heatsink and premium thermal paste can extend the thermal envelope further.

Sustained 1000 MHz Stability Test

A single benchmark reading shows peak performance. Sustained operation shows reliability. To validate the maximum overclock, Solo Satoshi ran a NerdQaxe++ Revision 7 at the full 1000 MHz firmware ceiling overnight using the stock Thermalright AXP90-X36 Black heatsink, Thermal Grizzly Kryonaut Extreme thermal paste, a Noctua NF-A8 rear fan oriented to blow cool air toward the board, and a Mean Well LRS-350-12 power supply.

The unit sustained 8.14 TH/s for 17 hours and 42 minutes with zero crashes or thermal throttling. The upgrades over the stock configuration were the thermal paste, and the higher-rated PSU required to feed 142W.

NerdQaxe++ Rev 7 Nerd*OS chart showing 8.14 TH/s sustained over 17h 42m uptime at 1000 MHz with stable ASIC and VRM temperatures

Metric Sustained Reading (1000 MHz)
Hashrate (actual) 8.14 TH/s (8,140 GH/s)
Hashrate (expected) 8.16 TH/s (8,160 GH/s)
Uptime 17h 42m
Frequency 1,000 MHz
ASIC Voltage 1.32 V
Input Current 11.91 A
Input Voltage 12.0 V
Power Usage 142.4 W
Efficiency 17.49 J/TH
ASIC Temperature 67.4°C
VRM Temperature (int/ext) 74.0°C / 74.0°C
ASIC Fan 2,738 RPM (100%)
VRM Fan 2,183 RPM (100%)

The actual hashrate of 8.14 TH/s landed within 0.25% of the 8.16 TH/s expected value, demonstrating that the TPS546D24A two-phase VRM holds clocks stable even at the maximum frequency. VRM temperature stabilized at 74.0°C, well within the TPS546D24A operating range, while the ASIC held at 67.4°C. Both fans ran at 100% to maintain these temperatures, confirming that 1000 MHz on the stock heatsink is the practical ceiling. As shown in the next section, upgrading to the AXP90-X47 full copper heatsink measurably lowers these temperatures and gives the fans headroom to spare.

This is the kind of sustained, stable performance that the previous 95472MC three-phase VRM could not reliably deliver at frequencies this high.


Copper Heatsink Upgrade: AXP90-X47 vs Stock at 1000 MHz

To quantify the benefit of the copper heatsink upgrade, Solo Satoshi ran the same 1000 MHz overclock with the Thermalright AXP90-X47 full copper heatsink in place of the stock AXP90-X36 Black. All other components and conditions were identical: 73°F ambient, Kryonaut Extreme paste, Noctua NF-A8 rear fan, and Mean Well LRS-350-12 PSU.

Metric (1000 MHz) Stock AXP90-X36 Copper AXP90-X47 Improvement
Hashrate 8.14 TH/s 8.13 TH/s Equivalent
Power Draw 142.4 W 137.6 W −4.8 W
Efficiency 17.49 J/TH 16.93 J/TH −0.56 J/TH
ASIC Temperature 67.4°C 61.4°C −6.0°C
VRM Temperature 74.0°C 72.0°C −2.0°C
ASIC Fan Speed 100% (2,738 RPM) 99% (2,682 RPM) Headroom restored
VRM Fan Speed 100% (2,183 RPM) 99% (2,190 RPM) Headroom restored

The copper heatsink dropped ASIC temperature by 6°C and VRM temperature by 2°C at the same 1000 MHz overclock. More notably, power draw fell by 4.8W while holding equivalent hashrate. This is a real thermal-electrical relationship: cooler ASICs have lower electrical resistance, so the chips draw less power to produce the same hashrate. The result is a measurable efficiency gain from 17.49 J/TH to 16.93 J/TH purely from better cooling.

The fan speed difference is the practical takeaway for 24/7 operation. On the stock heatsink, both fans run at 100% to hold 1000 MHz, leaving no thermal headroom. With the copper heatsink, the fans settle at 99% with cooler temperatures, meaning the cooling system is no longer maxed out. For miners planning to run at the 1000 MHz ceiling continuously, the AXP90-X47 copper upgrade provides the thermal margin that makes sustained max-frequency operation comfortable rather than borderline.


NerdQaxe++ Rev 7 vs Rev 6.1 vs Rev 5: Three-Way Comparison

Metric Rev 5 (Stock 600 MHz) Rev 6.1 (Stock 600 MHz) Rev 6.1 (OC 800 MHz) Rev 7 (Stock 600 MHz) Rev 7 (OC 1000 MHz)
Hashrate ~4.8 TH/s 4.82 TH/s 6+ TH/s 4.800 TH/s 8.100 TH/s
Power Draw ~86.7 W* 70.6 W 103 W 69.5 W 140.0 W
Efficiency ~18 J/TH 14.66 J/TH 15.65 J/TH 14.48 J/TH 17.28 J/TH
ASIC Temp 70°C+ 48.1°C 57.2°C 60.0°C** 66.0°C**
VRM Temp 70°C+ 41.2°C 54.4°C 57.0°C** 70.0°C**
VRM Configuration 95472MC (3-phase) 95472MC (3-phase) TPS546D24A (2-phase)
Max Firmware OC Limited by hardware 800 MHz (firmware locked) 1000 MHz (firmware unlocked)
Power Connector 8A barrel jack XT30 (15A) XT30 (15A)
Fuse One-shot mechanical None (fuse-free) None (fuse-free)
Firmware AxeOS AxeOS / Nerd*OS Nerd*OS

*Rev 5 power draw per PlebBase’s April 2025 review (smart meter measurement; on-screen UI reported 77W).

**Rev 7 temperatures reflect automatic fan control targeting 60°C. Tested at 73°F ambient with factory components.


Assembled in the USA

Every NerdQaxe++ Revision 7 is 100% assembled, tested, and firmware flashed in the USA. Solo Satoshi ships same-day from Houston, Texas on orders placed before 12pm CST.

Solo Satoshi is the first to release the NerdQaxe++ Revision 7, and every benchmark, photograph, and test result in this article comes from our own units tested in-house. If you see these images or performance figures republished elsewhere, or sellers advertising Rev 7 pre-sales without original testing data, that content originated here.


Why Solo Miners Choose the NerdQaxe++ Revision 7

  • Open Source Firmware: No encrypted black box. Firmware that’s 100% transparent.
  • Decentralization: Your desk adds hashrate outside large industrial pools.
  • Low power draw: 69.5 watts at stock fits standard outlets and small solar setups.
  • Quiet operation: Fan runs at 37% at stock settings. Near-silent for home and office use.
  • Proven VRM: TPS546D24A two-phase design from the Bitaxe GT.
  • US support: Solo Satoshi provides direct human support from a team that uses these miners daily.

NerdQaxe++ Shipping, Warranty, and Support

Solo Satoshi ships every NerdQaxe++ same-day from Houston, Texas on orders placed before 12pm CST. Each unit carries a 90-day warranty covering manufacturing defects. Global delivery via FedEx and UPS.


Start Solo Mining with the NerdQaxe++ Rev 7

A redesigned power delivery platform proven in the Bitaxe GT. Firmware unlocked to 1000 MHz. 8.1 TH/s on factory cooling. The NerdQaxe++ Revision 7 is the most complete iteration of the open-source quad-chip Bitcoin solo miner to date.

Buy the NerdQaxe++ Revision 7 Shop NerdQaxe++ at Solo Satoshi →

Sign up for our VIP subscriber list and receive bi-weekly newsletters on the most up to date information and best deals on Bitcoin home miners!

To top