Skip to content

Bitcoin accepted at checkout  |  Ships from Laval, QC, Canada  |  Expert support since 2016

ASIC Overclock Headroom by Cooling (2026): Air vs Immersion vs Hydro

A Bitcoin miner runs out of cooling before it runs out of chip — so better heat removal is, quite literally, more hashrate. But how much more? This dataset puts real numbers on it per model: the sustainable hashrate, wall watts, efficiency and percent uplift you get from air overclocking, immersion, and factory hydro versus the air-stock baseline. Every figure is graded A/B/C for source strength, because we do not blur a manufacturer datasheet with a forum overclock.

Quick answer

A Bitcoin ASIC is thermally throttled long before it is electrically maxed out — so better cooling is free hashrate, up to a point. This dataset shows, per model, how much more a miner sustains on air overclock, immersion, and factory hydro versus its air-stock baseline: the hashrate, the wall-watt cost, the efficiency and the percent uplift. Because heat is removed from every surface at once, immersion and hydro let the boards run more power before the chips hit their junction-temperature limit.

The honest headroom picture: the manufacturers already answered the question with factory hydro SKUs. Bitmain rates the S21 at 335 TH hydro vs 200 TH air (+68%), the S21 XP at 473 TH vs 270 TH (+75%), and MicroBT the M63S at 390 TH hydro; cooler junctions even buy a small efficiency gain (S21 XP 13.5 to 12.0 J/TH). But read the caveat: a hydro SKU uses a DENSER hashboard with more chips, so that uplift is added hardware plus cooling, not cooling alone unlocking your existing air unit. For an air unit you already own, field immersion/air overclocks report roughly 8 to 30 percent more hashrate on advanced-OC firmware — but every one of those figures here is grade C (single-source or extrapolated): not one met the two-independent-source bar, so none is graded B, and none is a manufacturer spec or a D-Central measurement. If you want real, warrantied headroom, buy the factory hydro variant; overclocking an air miner in immersion is a gamble with no OEM backing. Free CSV/JSON under CC BY 4.0.

Download CSV Download JSON REST API →

Read the grades — this dataset has NO first-party bench data. Every row is graded for source strength: A = Strong (a manufacturer datasheet — factory hydro variants and rated air specs, 15 rows); B = Moderate (field/community overclocks corroborated by 2+ independent sources, 0 rows); C = Limited (single-source or estimated, 5 rows). A manufacturer spec and a forum overclock are never blurred. B/C figures are reported community/field experience, never a D-Central-measured rate. Two caveats even on the A rows: a factory hydro SKU uses a denser hashboard with more chips, so part of its +67% to +127% uplift is added hardware, not cooling alone unlocking existing silicon; and the junction-temperature explanations are D-Central engineering interpretation layered on the OEM numbers, not manufacturer statements. Overclocking voids warranties and can destroy hardware.
ModelCoolingHashrate
TH/s
Wall
W
Eff.
J/TH
UpliftFirmwareGrade
Antminer S19 Pro
BM1398
Air (stock)1103,25029.50%stockA
Antminer S19 Pro Hyd (184TH)
BM1398
Hydro (factory)1845,42829.5+67.3%factoryA
Antminer S19 XP
BM1366
Air (stock)1413,01021.30%stockA
Antminer S19 XP Hyd (257TH)
BM1366
Hydro (factory)2575,30420.6+82.3%factoryA
Antminer S19j Pro
BM1362
Air (stock)1043,06829.50%stockA
Immersion (OC)1364,00029.4+30.8%advanced-OC firmware (BraiinsOS autotune)C
Antminer S19k Pro
BM1366
Air (stock)1202,760230%stockA
Air (max OC)1303,35025.8+8.3%advanced-OC firmware (VNish/BraiinsOS)C
Antminer S21
BM1368
Air (stock)2003,50017.50%stockA
Antminer S21 (air unit in immersion)
BM1368
Immersion (OC)2354,20017.9+17.5%advanced-OC firmware (VNish immersion profile)C
Antminer S21 (air unit, air-OC)
BM1368
Air (max OC)2153,85017.9+7.5%advanced-OC firmware (VNish/LuxOS/BraiinsOS)C
Antminer S21 Hyd (335TH)
BM1368
Hydro (factory)3355,36016+67.5%factoryA
Antminer S21 XP
BM1370
Air (stock)2703,64513.50%stockA
Antminer S21 XP Hyd (473TH)
BM1370
Hydro (factory)4735,67612+75.2%factoryA
Whatsminer M30S++
MicroBT ASIC (Samsung 8nm)
Air (stock)1123,472310%stockA
Immersion (OC)1384,50032.6+23.2%advanced-OC firmware (VNish immersion profile)C
Whatsminer M50S (126TH)
MicroBT 5nm (proprietary SHA-256)
Air (stock)1263,276260%stockA
Whatsminer M53S Hydro (260TH)
MicroBT 5nm (proprietary SHA-256)
Hydro (factory)2606,76026+106.3%factoryA
Whatsminer M60 (172TH)
MicroBT 5nm (proprietary SHA-256)
Air (stock)1723,42219.90%stockA
Whatsminer M63S Hydro (390TH)
MicroBT 5nm (proprietary SHA-256)
Hydro (factory)3907,21518.5+126.7%factoryA

Air-stock baselines cross-reference the power-profiles database; for the cooling methods themselves see the cooling comparison and immersion fluids, and for the temperature limits behind the headroom the operating-temperature dataset. Ran your own numbers? Add them to the community bench measurements. D-Central designs immersion and hydro cooling systems — talk to us before you push a fleet.

The air-stock baselines line up with the power-profiles database; the cooling methods themselves are compared in air vs immersion vs hydro, with the fluids in immersion cooling fluids and the temperature ceilings in the operating-temperature dataset. Ran your own immersion numbers? Add them to the community bench measurements. Part of D-Central’s data hub — and our immersion/hydro engineering practice.