// blog · · by Nathan Baldwin
// Three Bitaxe Gammas ran efficiency, balanced, and aggressive profiles for seven days. Here's the real hashrate, power draw, and J/TH data for each.
I ran three BM1370 Gammas side by side for seven days, each locked to a different tuning profile, on the same pool and the same room air. The aggressive board put out 42% more hashrate than the efficiency board and burned 92% more power to do it. That trade is the whole story of Bitaxe tuning, and a week of logging makes it concrete instead of a forum argument.
Same hardware, same firmware (AxeOS 2.9.x), same 24°C ambient, same fan controller settings. The only variables were core voltage and frequency. Everything below is averaged from CSV logs pulled every 15 seconds, so these are steady-state numbers, not the peak your dashboard flashes for three seconds after a restart.
I picked settings that bracket the useful range for a Gamma rather than chasing a record. Each board got a name and a fixed voltage/frequency pair for the full week.
Efficiency (board A): 1050 mV, 425 MHz. The “runs cool and sips power” setup.
Balanced (board B): 1150 mV, 490 MHz. Close to Gamma stock, the setting most people leave untouched.
Aggressive (board C): 1250 mV, 575 MHz. Pushed until the hardware error rate started climbing but before it fell over.
No auto-tuning, no thermal throttling kicked in during the week. All three stayed inside their voltage regulator’s safe zone, which is the number that actually gates a Gamma. If you haven’t read why VR temp is the limit that matters, start with the VR temperature breakdown before you push voltage.
Here’s what each board actually did, averaged across the full run.
| Metric | Efficiency (A) | Balanced (B) | Aggressive (C) |
|---|---|---|---|
| Core voltage | 1050 mV | 1150 mV | 1250 mV |
| Frequency | 425 MHz | 490 MHz | 575 MHz |
| Avg hashrate | 1.02 TH/s | 1.21 TH/s | 1.45 TH/s |
| Power at the wall | 12.6 W | 17.1 W | 24.2 W |
| Efficiency | 12.4 J/TH | 14.1 J/TH | 16.7 J/TH |
| VR temp (avg) | 51°C | 58°C | 71°C |
| ASIC temp (avg) | 49°C | 55°C | 64°C |
| HW error rate | 0.04% | 0.11% | 0.68% |
The efficiency board is the clear winner on joules per terahash. It produced hashrate at 12.4 J/TH, about 26% more efficient than the aggressive board’s 16.7 J/TH. If your electricity is the thing you care about, that gap compounds every hour the miner is on.
The aggressive board wins on raw output and nothing else. It found more hashrate, but it paid for every extra terahash at a worse and worse exchange rate. That’s the shape of every ASIC power curve: efficiency degrades faster than hashrate improves once you’re past the knee.
Averages hide the interesting part. The aggressive board didn’t just run hotter, it ran noisier. Its 15-second hashrate samples had a standard deviation of 118 GH/s, against 41 GH/s for the efficiency board. That’s nearly 3x the swing.
Some of that is normal share-luck jitter that shows up on any miner. But a chunk of it on board C traced back to its hardware error rate. At 0.68% HW errors, the aggressive board was throwing away roughly one in 150 computed hashes as invalid, and those rejects show up as dips in the reported rate. The efficiency board’s 0.04% error rate is basically noise-free by comparison.
If your Gamma’s hashrate graph looks like a seismograph, the fix is usually not the pool and not the firmware. It’s voltage set too high for the silicon you happened to get. For what a healthy error rate looks like in the logs, I broke that down in the hardware error rate post.
Board C produced about 0.43 TH/s more than board A, sustained across the week. On a Gamma, at current network difficulty, that’s a rounding error in expected earnings whether you’re solo or pooled. You can sanity-check that yourself against live difficulty on mempool.space before you talk yourself into an overclock.
Meanwhile the extra 11.6 W that board C pulled, running 24/7, is real and it’s every single day. Over a year that’s roughly 100 kWh of additional draw per board for a hashrate bump that barely moves your odds. The efficiency profile loses a little expected value on hashrate and wins it back several times over on the power bill.
The balanced board is the honest default. It gave up 20% of the aggressive board’s output while running 21°C cooler at the VR and using 29% less power. For most people running a Gamma in a room they also sleep in, balanced is the profile that doesn’t make you think about it again.
Overclocking isn’t wrong, it’s just situational. If you’re solo mining and treating the whole thing as a lottery ticket, raw hashrate is the only variable that changes your odds, and efficiency is beside the point. Push it. Watch the VR temp, keep the HW error rate under about 1%, and accept the power bill as the cost of more tickets.
The same logic applies if power is free or flat-rate to you, or if you’re deliberately using the miner as a space heater in winter. In those cases the efficiency penalty isn’t a penalty at all.
Pick your profile based on which number you’re optimizing, not on which sounds fastest. If you pay for power and want the best return, run something near the efficiency profile: drop voltage to around 1050 mV and frequency to 425 MHz, then confirm your HW error rate stays under 0.1% over a full day. If you’re chasing a solo block, run aggressive and watch VR temp instead of ASIC temp. Either way, log for a full week before you decide anything. A single afternoon of numbers will lie to you, because thermals, share luck, and error rates all need days to average out.
Try it yourself: Bitaxe Baller is a free desktop app for Mac and Windows that surfaces these recommendations automatically across your fleet — live monitoring, tuning suggestions, pool config, all in a native window. Open source on GitHub.
