B++ Logo

Mining Hardware Evolution

Bitcoin mining hardware has undergone a remarkable evolution, from hobbyists running CPUs on laptops to industrial-scale operations with purpose-built silicon. This progression represents one of the fastest hardware development cycles in computing history, driven by the economic incentives of proof-of-work mining.

The Four Eras of Mining Hardware

Era 1: CPU Mining (2009-2010)

The Beginning

When Satoshi launched Bitcoin, mining was done on ordinary computer CPUs:

  • Hardware: Intel/AMD desktop processors
  • Hashrate: 1-20 MH/s
  • Power: 50-150W
  • Efficiency: ~0.1 MH/J
  • Who mined: Cypherpunks, early adopters, Satoshi

Why CPUs Work

SHA-256 (Bitcoin's hash function) is computationally simple:

  • Bitwise operations (AND, OR, XOR)
  • 32-bit additions
  • No complex branching
  • Highly parallelizable

CPUs can do this, but they're general-purpose. Most of their transistors do things irrelevant to hashing.

The End of CPU Mining

By late 2010, GPU mining emerged. CPU miners quickly became unprofitable and disappeared.

Era 2: GPU Mining (2010-2013)

The GPU Advantage

Graphics cards excel at parallel computation:

  • Hardware: AMD Radeon, NVIDIA GeForce
  • Hashrate: 10-800 MH/s
  • Power: 100-300W
  • Efficiency: ~2-4 MH/J
  • Improvement: 10-50x over CPUs

Why GPUs Excel

CPU: 4-8 cores, each very fast, complex
     Good at: varied tasks, branching logic
     
GPU: 1000+ cores, each slower, simpler
     Good at: same operation on many data points
     Perfect for: trying billions of nonces

The GPU Mining Era

  • AMD cards preferred (better for SHA-256)
  • Mining farms with racks of graphics cards
  • Gaming GPU shortages began
  • Some miners used OpenCL/CUDA optimizations

Popular GPU Mining Cards

CardHashratePowerEra
ATI Radeon 5870400 MH/s200W2010
AMD Radeon 7970700 MH/s250W2012
AMD R9 290X900 MH/s300W2013

Era 3: FPGA Mining (2011-2013)

Field Programmable Gate Arrays

FPGAs are chips that can be reconfigured for specific tasks:

  • Hardware: Xilinx, Altera FPGAs
  • Hashrate: 100-800 MH/s
  • Power: 20-80W
  • Efficiency: ~10-20 MH/J
  • Improvement: 5-10x efficiency over GPUs

FPGA Advantages

  • Much more power efficient than GPUs
  • Could be reprogrammed if algorithm changed
  • Lower heat generation

FPGA Disadvantages

  • Expensive development
  • Limited availability
  • Required technical expertise
  • Quickly obsoleted by ASICs

Short-Lived Era

FPGAs were a transitional technology. The efficiency gains made ASICs inevitable and economically viable.

Era 4: ASIC Mining (2013-Present)

Application-Specific Integrated Circuits

ASICs are chips designed to do one thing only: SHA-256 hashing.

  • Hardware: Custom silicon from Bitmain, MicroBT, Canaan, etc.
  • Hashrate: 1 TH/s → 250+ TH/s (2013 → 2024)
  • Power: 500-3500W per unit
  • Efficiency: 100 J/TH → 15 J/TH (improving constantly)
  • Improvement: 1000x+ over GPUs

Why ASICs Dominate

GPU: General-purpose silicon
     - 30% doing hashing
     - 70% doing other stuff

ASIC: Purpose-built silicon
     - 100% doing hashing
     - Nothing wasted

Every transistor in an ASIC is dedicated to SHA-256. No graphics processing, no floating point, no cache hierarchy, just hashing.


ASIC Evolution

Generation Timeline

EraExampleHashrateEfficiencyYear
Gen 1Avalon 166 GH/s9,000 J/TH2013
Gen 2Antminer S1180 GH/s2,000 J/TH2013
Gen 3Antminer S51.15 TH/s510 J/TH2014
Gen 4Antminer S74.7 TH/s250 J/TH2015
Gen 5Antminer S914 TH/s100 J/TH2016
Gen 6Antminer S1756 TH/s45 J/TH2019
Gen 7Antminer S1995 TH/s34 J/TH2020
Gen 8Antminer S19 XP140 TH/s21 J/TH2022
Gen 9Antminer S21200 TH/s17.5 J/TH2023
Gen 10Antminer S21 XP270 TH/s13.5 J/TH2024

Process Node Shrinks

ASIC efficiency improves primarily through semiconductor process improvements:

2013: 110nm, 55nm
2014: 28nm
2016: 16nm
2018: 7nm
2020: 5nm
2023: 3nm (emerging)

Smaller transistors = less power per hash = better efficiency.

Major Manufacturers

Bitmain (China)

  • Market leader, Antminer series
  • Founded 2013 by Jihan Wu and Micree Zhan
  • Controversial: BCH support, internal conflicts
  • Products: S9, S17, S19, S21 series

MicroBT (China)

  • Strong competitor, Whatsminer series
  • Founded 2016 by former Bitmain engineer
  • Known for reliability
  • Products: M20, M30, M50, M60 series

Canaan (China)

  • First ASIC manufacturer (Avalon)
  • Publicly traded (NASDAQ: CAN)
  • Products: Avalon series

Intel (USA)

  • Entered market 2022 with Blockscale
  • Focused on efficiency
  • Exited market 2024

Bitfury (Netherlands/USA)

  • Vertically integrated (makes and uses chips)
  • Known for immersion cooling
  • Products: Clarke, Bitfury B8

Efficiency Metrics

Joules per Terahash (J/TH)

The key efficiency metric:

J/TH = Watts ÷ (Terahashes per second)

Example: 3000W machine doing 100 TH/s
Efficiency = 3000 ÷ 100 = 30 J/TH

Lower is better. Modern machines: 15-25 J/TH.

Hashrate per Dollar

Consider total cost of ownership:

Machine cost: $5,000
Hashrate: 100 TH/s
Lifespan: 3 years

Cost per TH/s/year: $5,000 ÷ 100 ÷ 3 = $16.67

Break-Even Analysis

Revenue per TH/day: ~$0.08 (varies with difficulty and price)
Electricity cost: $0.05/kWh
Machine efficiency: 25 J/TH

Power per TH/day: 25 J/s × 86,400 s = 2.16 MJ = 0.6 kWh
Electricity per TH/day: 0.6 × $0.05 = $0.03

Profit per TH/day: $0.08 - $0.03 = $0.05

Industrial Mining Operations

Scale

Modern mining farms:

  • Hashrate: 1-50 EH/s (exahashes per second)
  • Power: 50-500 MW
  • Machines: 10,000-100,000+ ASICs
  • Investment: $100M-$1B+

Infrastructure Requirements

Power

  • Cheap electricity is critical ($0.02-0.05/kWh ideal)
  • Substations, transformers, distribution
  • Often: stranded energy, renewables, flared gas

Cooling

  • ASICs generate enormous heat
  • Air cooling: fans, ducting, outdoor air
  • Immersion cooling: machines submerged in dielectric fluid
  • Target: 15-25°C ambient

Networking

  • Low latency to pools
  • Redundant connections
  • Monitoring systems

Security

  • Physical security (machines are valuable)
  • Cybersecurity (prevent hashrate theft)
  • Fire suppression

Geographic Distribution

Mining gravitates toward cheap power:

  • United States: Texas (wind), Georgia (nuclear), Wyoming
  • Canada: Quebec (hydro), Alberta
  • Kazakhstan: Coal power (declining due to regulations)
  • Russia: Siberia (hydro, cold climate)
  • Nordic countries: Hydro, geothermal, cold
  • Middle East: UAE, Oman (cheap natural gas)

Home Mining

Is It Viable?

For most people in most places: marginally, or no.

Challenges:

  • Electricity costs ($0.10-0.30/kWh residential)
  • Noise (70-80 dB, like a vacuum cleaner)
  • Heat (3kW space heater per machine)
  • Space and ventilation

Where It Works:

  • Cheap or free electricity
  • Cold climates (use heat)
  • Off-grid (solar, hydro)
  • Learning/hobby purposes

Home Mining Options

Full ASIC

  • Antminer S9 (old, cheap, inefficient)
  • Small new units (Antminer S19 nano)
  • Noise and heat issues

USB/Low-power

  • FutureBit Apollo
  • Nerdminer (ESP32)
  • Not profitable, but educational

Heating Integration

  • Heatbit, Hestiia
  • ASIC mining as home heating
  • Heat is a feature, not waste

The ASIC Trap

No Escape

Once ASICs exist, there's no going back:

  1. ASICs are 10,000x more efficient than GPUs
  2. GPU miners can't compete
  3. ASIC investment creates lock-in
  4. Algorithm changes would destroy investment

ASIC Resistance (Other Coins)

Some cryptocurrencies tried to resist ASICs:

  • Memory-hard algorithms: Ethereum (Ethash), Monero (RandomX)
  • Frequent algorithm changes: Monero
  • ASIC-resistant designs: Often just delays ASICs

Bitcoin's position: ASICs are a feature, not a bug. They represent committed capital that can't be repurposed.


Efficiency Limits

Physical limits are approaching:

  • 3nm/2nm processes: Near atomic scale
  • Thermodynamic limits: Minimum energy per computation
  • Diminishing returns: Each generation improves less

Immersion Cooling

Submerging ASICs in dielectric fluid:

  • Better cooling: Removes heat more efficiently
  • Overclocking: Run chips faster
  • Longevity: Less thermal stress
  • Density: More machines per space

Stranded Energy

Mining as flexible load:

  • Flared gas: Capture otherwise wasted energy
  • Curtailed renewables: Use excess wind/solar
  • Grid balancing: Ramp up/down based on demand
  • Remote locations: Monetize energy that can't reach grid

Vertical Integration

Large miners building their own chips:

  • Reduce reliance on Bitmain/MicroBT
  • Custom optimizations
  • Supply chain control
  • Examples: Bitfury, Intel (briefly), Block/Square (planned)

Summary

The evolution of mining hardware:

EraTechnologyEfficiency GainTimeline
1CPUBaseline2009-2010
2GPU10-50x2010-2013
3FPGA5-10x2011-2013
4ASIC1000x+2013-present

Key takeaways:

  • ASICs dominate and will continue to
  • Efficiency is everything: J/TH determines profitability
  • Industrial scale is increasingly dominant
  • Hardware is expensive and depreciates quickly
  • Location matters: Cheap power wins


Resources