nd th th
ENERGY EFFICIENT BITCOIN MINING TO MAXIMIZE THE
MINING PROFIT: USING DATA FROM 119 BITCOIN MINING
HARDWARE SETUPS
1
AMILA PATHIRANA, 2MALKA N. HALGAMUGE, 3ALI SYED
1,3
School of Computing and Mathematics, Charles Sturt University, Melbourne, Victoria 3000, Australia
2
Department of Electrical and Electronic Engineering, The University of Melbourne, Parkville, VIC 3010, Australia E-mail: [email protected], [email protected], [email protected]
Abstract - In recent years, cryptocurrency such as Bitcoin has attracted much global attention. Bitcoins are produced by mining as a fee for authenticating any transaction. This process requires a significant amount of computing power. In the early stage, mining started from personal computers and currently, specialized hardware developed to increase the mining speed by reducing power consumption. Therefore, this research aims to explore the impact of hardware efficiency from the early stages of Bitcoin mining hardware to current customized mining hardware for better Bitcoin mining process to maximize mining profit. In this study, we analyzed miner performance from basic to advance hardware and discuss how to identify a suitable miner to increase the profit by comparing their performance and price. We extracted data from 30 peer-reviewed scientific publications (2013–2018) describing 119 Bitcoin mining hardware setups to record the power consumption and hash rates for cryptocurrency mining. Hardware efficiencies were calculated using power consumption and hash rate, then compared with different miners for identification purposes. Our findings suggest that Nvidia GPUs are inefficient and more expensive than the ATI GPUs. Additionally, ASICs and FPGA miners are more efficient compared to CPU and GPU miners. The BitmainAntminer S9 is the most efficient miner for cryptocurrency mining. This study could be utilized to identify common hardware efficiencies to maximize Bitcoin mining profits. Moving to cloud/hosting solutions rather than spending money on upgrading mining hardware could be an exciting future research avenue.
Keywords - Cryptocurrency, Bitcoin, Mining, Efficiency, Hash rates, Hardware.
Graphical Abstract
I. INTRODUCTION
Cryptocurrency is a decentralized digital currency designed using cryptography to work as an exchange medium over the internet due to the rapid growth of information technology. Bitcoin is the first implementation of cryptocurrency, which was
introduced in late 2008 [1] by Satoshi Nakamoto. Since bitcoin deployment in 2009 [2], it has become the most successful digital currency despite there being around 900 [3] cryptocurrencies. Bitcoin can be acquired by "merchandising, providing services, exchanging with other currencies" [3] besides, mining is the practice used to harvest these virtual currencies.
nd th th As the first step of Bitcoin mining, users install free
software, which is known as a wallet (free client) to their computers to send and receive money using Bitcoin addresses. Participants (miners) who are joined to the Bitcoin network broadcast blocks (A block is a container of Bitcoin transaction [4]) into the public blockchain using a secure hashing algorithm (SHA-256) [5] after verifying transactions into blocks. For each block, miners get rewarded with 25 new Bitcoins and every 10 minutes [1] a new block is generated by one of the miners in the Bitcoin network. As this mining process creates hashes of transactions, the miner keeps their block in a loop as fast as possible thus consuming a lot of “processor cycles” [4] and therefore a large amount of electricity.
Initially, mining started on standard computers. As it gained popularity, miners attempted to increase their hash rates [6] by using different hardware. As a result, Bitcoin protocols made mining more difficult [7] over time. The general public who used their computing resources such as 1st and 2nd generation Central Processing Unit (CPU) and Graphic Processing Unit (GPU) miners became out of date as their hardware failed to keep up with difficulty to gain a reasonable profit. Additionally, selecting a mining pool also became complicated due to the introduction of various pools to the cryptocurrency network, which requests specific hardware. Hence, conducting more comprehensive research will be essential to identify the appropriate parameters such as energy consumption, hash rates and hardware prices when choosing the right tools for Bitcoin mining.
A miner called “Andrew Geyls” [4] and researched studies pointed out how many people could be losing money while mining Bitcoin. According to Lee and Kim [8], the main reason is, miners, fail to continue to invest in hardware as the network increases the mining difficulty every two weeks. As a result, over time, the hardware generates fewer hashes and consumes more electricity. Therefore, many papers compare the electricity consumption price with the Bitcoin value as it is a way of trading energy without focusing on inexpensive, efficient hardware. An estimation of hardware energy consumption is essential [9-11], as it consumes energy in different ways.
Development of mining-specific Field-programmable Gate Arrays (FPGAs) and Application Specific Integrated Circuits (ASICs) has made the use of consumer CPUs and GPUs miners outdated [8]. Nonetheless, graphics cards (GPUs) can be used to develop customized ASIC miners to achieve higher hash rate and compete with other devices. Therefore, this paper used 119 hardware devices to extract experimental data from 30 studies to analyze their efficiency and expenses by comparing each device to maximize the Bitcoin mining profit.
II. MATERIAL & METHODS
2.1. Collection of raw data
Bitcoin mining hardware data was collected such as hash rates, power consumption, price, release dates and the efficiency of different hardware types such as central processing unit (CPU), graphics processing unit (GPU), application-specific integrated circuits (ASIC) and field-programmable gate arrays (FPGA). The individual data for these elements is shown in Table 1 and Table S1. This investigation contains 30 studies that used 119 different hardware types for Bitcoin mining.
Device type The core elements of a computer. Ex. CPU, GPU. Hardware The collection of physical parts
of a computer system. Power
consumption
The electricity power used by a device.
Hash rate
The speed at which a computer is completing an operation in the
Bitcoin code. Efficiency Ability to perform a task with
minimum wastage. ASIC Application-specific integrated
circuit.
FPGA Field-programmable gate array.
Table 1: Parameters used in raw data collection (Table S1)
Since some authors [2, 5, 9] did not mention the efficiency of the hardware while measuring the hash rates and the power consumption, we calculated the efficiency using equation (1). However, it should be noted that a few authors [4, 6,] have calculated the efficiency by dividing the Hash rate by Power consumption.
2.2. Data Processing
In our survey, power consumption data was recorded in watts (W) and hash rates in Giga hashes per second (Gh/s). Since few studies used different units to analyze the hash rates, we used Giga hash value (1Gh/s = 1000Mh/s) as shown in Table S1.
2.3. Efficiency Calculation
Attributes such as electricity cost and performance per dollar are not crucial for this evaluation since efficiency is the most essential element. Some authors [7, 10] calculated the efficiency of Bitcoin mining using W/Gh/s and equation 1.
That is, the mining efficiency attribute was calculated using power consumption and hash rates that are given by [6]. As such, the lowest value indicates the most efficient hardware:
power consumption (W)
mining efficiency = --- (1) hash rate (Gh/s)
nd th th
2.4. Data Analysis
Hardware specifications data presented in Table S1 was mainly analyzed according to their individual performance and prices. Due to the publication period and operating conditions, studies indicated differentprices and hash rates for the same hardware. Parameters such as release date and price data are not mentioned in every study. Therefore, those data were extracted from the manufacturer’s website or referenced web pages.
III. RESULTS
The main goal of this research is to identify the most efficient hardware device for Bitcoin mining without spending large amounts of money. To simplify the process, raw data were categorized according to device types. Subsequently, the data were branded accordingly for identification purposes. Four criteria were developed using the analyzed data to identify appropriate mining hardware. Figure 1 demonstrates the different processors' data discussed in the 30
studies. Specifically, Fig. 1(a) shows Intel processors, Fig. 1(b) shows Advanced Micro Devices (AMD) processors and Fig. 1(c) compares the two. Intel’s price comparison indicates a slight increase in the middle while the AMD Fig. 1(b) has a consistent price. Similarly, in terms of efficiency, Intel slightly increases in the middle and AMD is steady. Fig. 1(c) confirms that AMD processors are more expensive than Intel’s as the price comparison has a sudden peak in the middle. The efficiency comparison in Fig. 1(a), fluctuates from start to end.Therefore, Intel processors such as Atom N450, i7 990x, i7 2760QM, i7 3930k, and Xeon Phi 5100 can be identified as more efficient processors. According to Fig. 1(c), AMD processors are not efficient because they consume more power compared to the Intel series. Furthermore, Xeon Phi 5100 is more suitable than the rest, as it is inexpensive and consumes a reasonable amount of power to generate additional hashes. Therefore, AMD processors and rest of the Intel processors could not be identified as efficient mining hardware.
Fig.1. AMD and Intel manufacturer’s processors (first generation miners) were compared to their prices and to their efficiencies. (a) for Intel, (b) for AMD and (c) for both Intel and AMD. 30 Bitcoin mining studies (from 2013 to 2018) were used to identify the
efficiency and the price values. Since few papers provided power consumption and hash rate values, processor efficiency is calculated using the given equation. Efficiency = Power consumption(W) / Hash rate (Gh/s). Please note that price values are in
nd th th Graphics cards’ efficiencies and prices are illustrated
in Fig. 2. Figure 2(a) shows the results for ATI graphic cards. Fig. 2(b) illustrates the results for Nvidia cards and Fig. 2(c) compares the two. ATI cards’ price and efficiency fluctuate dramatically between 0 to 2000 W/Gh/s except for the Radeon RX Vega 56, which is noticeably inefficient compared to the rest except for a sharp increase at the end. The most efficient and affordable cards identified from the ATI (Fig. 2(a)) are Radeon 5850, 5970 and 6990. As can be seen from Fig. 2(b), Nvidia’s efficiency increases slightly from the beginning and gradually decreases up to the Tesla before growing significantly at the end. The price comparison has a similar behaviour for the Nvidia GTX 570 and has a sudden increase near Tesla s2070 and sudden drop at the end. According to Fig. 2(b), the most efficient Nvidia card is the Tesla s2070, however, it is also the most expensive. However, Fig. 2(c) clearly illustrates that the ATI cards are more efficient and cheaper than the Nvidia cards. Both brands consume a similar amount of power except the ATI Radeon 5870x6 (crossfire) and Nvidia Tesla s2070. Therefore, the cards’ efficiency mainly depends on hash rates. As a result, investing in efficient ATI cards will help to gain more profits. Asics miners’ results are illustrated in Fig. 3.
Due to the rapid growth of cryptocurrency mining, there are several hardware manufactures which makes a comparison between them difficult. For instance, the efficiency comparisons have slight variations in every figure, and the price lines fluctuate from start to end. According to Fig. 3(a) BitmainAntminer s2 is the most expensive miner out of the Antminer series. It is not as efficient as the Antminer S9, which is the latest and most reasonably priced Bitmain product. From Fig. 3(b) the Avalon 6 miner is the most efficient miner out of Canaan series and costs under $1000. Sapphire miners, according to Fig. 3(c), are also inexpensive and cost under $100.00 However, the Sapphire is more efficient compared to the Blade. According to Fig. 3(d), KnC Neptune is almost three times more expensive than the Knc Jupiter. Although it is efficient and powerful, it has a higher hash rate and can solve complex algorithms. Fig. 3(e) contains more ASIC miners developed by different manufacturers. The Monarch BPU 600C is the most efficient and reasonably priced miner in the group. Comparing all the ASIC miners from Fig. 3(f), most of them are powerful, and they have their advantages and disadvantages. Though it is clear that BitmainAntminer S9 is the most efficient and affordable miner compare to the rest.
Fig.2. ATI and Nvidia graphics card (second-generation miners) comparisons to their prices and to their efficiencies. (a) for ATI, (b) for Nvidia and (c) for both GPUs. Graphics cards' efficiencies and prices were used from 30 papers. Please note, selected papers provided the power consumption and the hash rates. Therefore, efficiency is calculated using the following equation. Efficie ncy = Power consumption(W) / Hash rate(Gh/s). Plots indicate that price values are in solid blue lines, and efficiency values are in orange
nd th th Four different type of FPGA miners’ efficiencies and
prices are demonstrated in Fig. 4 using 30 peer review papers. All miners’ results were analyzed from criteria a to c for each brand and Fig. 4(d) compares all the FPGA miners. Significantly, the ASIC and FPGA miners have similar efficiency values. However, price comparisons have sharp differences. For example, in Fig. 4(a), the Xilinx Spartan 6 is the most efficient but also the costliest compared to the Xilinx Spartan 6-150 miner. Similarly, the Butterfly Labs Mini Rig is more
efficient than the Butterfly labs Single but again more expensive. The price comparison in Fig. 4(b) exhibits a rapid increase between Butterfly labs Single and Mini Rig miners. Fig. 4(c) illustrates two different studies about the DigilentNexys 2 500k. Both studies delivered similar efficiency values; nevertheless, there is a significant price difference. To conclude, after analyzing all the FPGA miners in Fig. 4(d), Xilinx Spartan 6 could be identified as the most affordable and efficient FPGA miner for Bitcoin mining.
Fig.3. ASIC miner hardware compared to their prices and their efficiencies. (a) for Bitmain, (b) for Canaan, (c) for block, (d) for KnC, (e) for rest of the miners and (f) for all the ASIC miners. Values were obtained from 30 different studies. Certain pape rs provided the power consumption and the hash rates. Hence, efficiency is calculated using the Efficiency = Power consumption(W) /
nd th th
Table 2: Most efficient and affordable hardware devices identified by individual manufacturer
Fig.4. FPGA miners compared to their prices and their efficiencies. (a) for Xilinx, (b) for BFL, (c) for Digilent and (d) for all the FPGA miners. Plot data were acquired from 30 studies (from 2013 to 2018). Please note, several papers provided the power
consumption and the hash rates. Therefore, efficiency is calculated using the following equation. Efficiency = Power consumption(W) / Hash rate (Gh/s). Price values are in solid blue lines, and efficiency values are in or ange dotted lines.
nd th th
Fig.5. Intel, AMD, ATI, Nvidia, Bitmain, Canaan, Block erupter, KnC, Monarch, Xilinx, Digilent, Butterfly Labs manufacturers’ most efficient miner’s efficiencies and prices documented while analyzing our study.
IV. RESULTS
In this study, we examined a variety of hardware performance aspects from the early stage of Bitcoin mining to current customized mining. The choice of hardware is an essential factor to consider when mining Bitcoin due to the volatility of the Bitcoin price and the rising mining difficulty. New hardware costs a large amount of the mining income and miners must always buy the hardware with the lowest cost per Gh/s [12] to increase their chances of recovering the hardware cost as quickly as possible in order to gain profit.
Energy usage, hash rate and cost in each hardware were measured to identify the most suitable mining device. Through a combination of energy consumption and hash rate, we identified the efficiency of each hardware and compared with the cost to make sure the hardware must be able to earn more money than it uses. However, different authors used different methods to calculate efficiency. For instance, Krishnan et al. [15] calculated the hardware efficiency dividing the hash rates by power consumption; according to this equation, a higher efficiency value is better. The rest of the papers used in our work calculated the efficiency using the same values, although with a different technique. Therefore, (eq (1): Efficiency = Power consumption (W) / Hash rate (Gh/s)) was used to extract the hardware efficiencies obtained in Table S1.
A significant part of a miner's total costs come from hardware purchases. In order to start mining, the miner must initially buy the mining hardware. Mining hardware can be costly, with some FPGA miners such as Butterfly Labs Mini Rigs reaching up to tens
of thousands of dollars. Therefore, it is essential that the mining hardware eventually pays for it.
To summarize, the identified hardware device data from each criterion was collected in Table 2 to develop Fig. 5 to identify the best mining hardware. KnC Neptune and Butterfly Labs Mini Rigs are high-priced compared to the other miners. CPU and GPU miners are not efficient enough while they are less expensive than the rest. According to Fig. 5, BitmainAntminer S9, Canaan Avalon 6, Block Erupter Sapphire, Monarch BPU 600 C are the most efficient and affordable miners. Considering the power consumption and computing power, which known as the hash rate gathered from Table 2, BitmainAntminer S9 is the appropriate hardware that currently exists for Bitcoin mining.
V. CONCLUSION
In this study, we performed a survey of the data obtained from 30 peer-reviewed articles from 2013-2018 describing 119 different hardware observations carried out in the scientific literature, which discuss the Bitcoin mining hardware performance from the deployment of Bitcoin cryptocurrency to currently available hardware in the market. After analyzing the scientific data, we have identified that BitmainAntminer S9 is currently the most excellent miner for cryptocurrency mining. Our survey also indicates that first-generation mining hardware (CPUs) can be ignored as a result of inefficiency and mining difficulty. Additionally, our results show ATI GPUs are efficient and inexpensive compared to Nvidia GPUs. Therefore, ATI GPUs can be used to assemble custom ASIC mining hardware such as
nd th th Mining Rigs. Moreover, this research could be useful
as a reference for many researchers and miners to identify the common hardware efficiencies to maximize their mining profits in cryptocurrency.
AUTHOR CONTRIBUTION
A.P. and M.N.H. conceived the study idea and developed the analysis plan. A.P. analyzed the data and wrote the initial paper. M.N.H. helped to prepare the figures and tables and finalizing the manuscript. All authors read the manuscript.
REFERENCES
[1] Y. Liu, X. Chen, L. Zhang, C. Tang and H. Kang, "An Intelligent Strategy to Gain Profit for Bitcoin Mining Pools", Int. Symposium on Computational Intelligence and Design, vol. 10, p. 427, 2017.
[2] M. Bedford Taylor, "The Evolution of Bitcoin Hardware", Computer, vol. 50, no. 9, pp. 60-61, 2017. [3] M. Salimitari, M. Chatterjee, M. Yuksel and E. Pasiliao,
"Profit Maximization for Bitcoin Pool Mining: A Prospect Theoretic Approach", International Conference on Collaboration and Internet Computing, vol. 3, p. 271, 2017.
[4] O. Ala-Peijari, "Bitcoin the Virtual Currency: Energy Efficient Mining of Bitcoins", pp. 58-76, 2014.
[5] A. Kampl, "Bitcoin 2-Phase Immersion Cooling and the Implications for High Performance Computing", Electronics Cooling Magazine, p. 1, 2014.
[6] K. O'dwyer and D. Malone, "Bitcoin Mining and its Energy FootprintBitcoin Mining and its Energy Footprint", pp. 3-4, 2014.
[7] L. Cocco and M. Marchesi, "Modeling and Simulation of the Economics of Mining in the Bitcoin Market", PLOS ONE, vol. 11, no. 10, p. 7, 2016.
[8] J. Lee and Y. Kim, "An Analysis of Energy and Hardware Impacts on the Bitcoin Mining Network", ACM Transactions on Economics and Computation, vol. 5, no., pp. 1-10. [9] M. N. Halgamuge, M. Zukerman, R. Kotagiri, and H. Vu,
"An Estimation of Sensor Energy Consumption", Progress In Electromagnetics Research B, Volume 12, 259-295, Jan. 2009
[10] M. N. Halgamuge, S. M. Guru, and A. Jennings, "Centralized Strategies for Cluster Formation in Sensor Networks", Classification and Clustering for Knowledge Discovery, Springer-Verlag, ISBN: 3-540-26073-0, Chapter 20, pp. 315-334, Aug 2005
[11] M. N. Halgamuge, "Efficient Battery Management for Sensor Lifetime", Proceedings of IEEE 21st International Conference on Advanced Information Networking and Applications (AINA’07), Volume 1, pp 56-61, Niagara Falls, Canada, 21-23 May 2007.
Energy Efficient Bitcoin Mining to Maximize the Mining Profit: Using Data from 119 Bitcoin Mining Hardware Setups
SUPPLEMENTARY MATERIALS
Table S1: Raw hardware data collected from 30 peer review papers from 2013 to 2018
Study Cryptocurrency Device type Hardware Price USD Released date Power consumption
(W)
Hash rate
Gh/s Efficiency W/Gh/s Taylor Bitcoin CPU Intel (Core i7 990x) $339.98^ Feb-11 130[2] 0.033 3939.39+ (2017) [1] Bitcoin GPU Nvidia (GTX 570) $153.29^ Dec-10 219[3] 0.155 1412.9+ Bitcoin GPU ATI (Radeon 7970) $450.00 Jan-12 300[4] 0.675 444.44+ Bitcoin ASIC Bitmain (Antminer S9) $2,100.00 14/7/2016* 1323 13500 0.098+ Ala-Peijari Bitcoin GPU Mac Mini - 1/1/2005* 22 0.0065 3384.61+ (2014) [5] Bitcoin CPU Intel (Atom D510) $63.00# 1/12/2009* 37 0.00076 48684.21+ Bitcoin CPU Intel (Core i7 920) $348.95# 1/11/2008* 184 0.0038 48421.05+ Kampl Bitcoin CPU Intel (Xeon E5530 2.40GHz) $65.45^ 2009 Q4 80 0.00714 11204.48+ (2014) [6] Bitcoin GPU ATI (Radeon HD 5830) $49.97^ 2010 Q1 192 0.245 783.67+ Bitcoin GPU ATI (Radeon HD 6990) $500.00^ 2011 Q1 450 0.8 562.5+ Bitcoin FPGA Xilinx (Spartan-6) - 2012 Q1 10 0.245 40.81+ Bitcoin FPGA Xilinx (Spartan-6 (4)) - 2012 Q2 40 0.85 47.05+ Bitcoin ASIC ASIC 28nm (single) $450.00^ 2014 Q1 500 500 1+ O'Dwyer and Malone Bitcoin CPU Intel (Core i7 950) $398.95# 1/6/2009* 150 0.0189 7936.5 (2014) [7] Bitcoin CPU Intel (Atom N450) $64.00# 1/1/2010* 6.5 0.0016 4062.5 Bitcoin CELL Sony PlayStation 3 $169.85# 1/11/2006* 60 0.021 2857.14 Bitcoin GPU ATI (Radeon 4850) $108.00^ 1/6/2008* 110 0.101 1089.1 Bitcoin GPU ATI (Radeon 5770) $194.99^ 1/10/2012* 108 0.2145 503.49 Cocco and Marchesi Bitcoin GPU ATI (Radeon 5830) $99.00! 1/2/2009* 302[9] 0.509+ 593.22 (2016) [8] Bitcoin GPU ATI (Radeon 5850) $324.00! 1/9/2009* 314[9] 0.787+ 398.94 Bitcoin GPU ATI (Radeon 5870) $211.00! 1/9/2009* 338[9] 0.722+ 467.66 Bitcoin GPU ATI (Radeon 5970) $700.00! 1/11/2009* 465[9] 1.186+ 392 Bitcoin GPU ATI (Radeon 6870) $190.00! 1/10/2010* 317[9] 0.629+ 503.33 Bitcoin GPU ATI (Radeon 6950) $275.00! 1/12/2010* 225[10] 0.45+ 500 Bitcoin GPU ATI (Radeon 6990) $480.00! 1/3/2011* 375[11] 1.139+ 328.95 Lee and Kim Bitcoin ASIC Canaan (Avalon 1) $1,299.00 Jan-13 400 66 6.06 [12] Bitcoin ASIC Block Erupter (Sapphire) - Apr-13 2.55 0.336 7.59
Bitcoin ASIC Jalapeno - Jun-13 30 5 6
Bitcoin ASIC KnC (Jupiter) $4,995.00 Aug-13 600 400 1.5 Bitcoin ASIC Bitmain (Antminer S1) $299.00 Nov-13 200 100 2 Bitcoin ASIC Canaan (Avalon 2) $200.00 Jan-14 340 105 3.23 Bitcoin ASIC Bitmain (Antminer S2) $3,899.00 Apr-14 1100 1000 1.1 Bitcoin ASIC Canaan (Avalon 3) $480.00 Apr-14 405 325 1.25 Bitcoin ASIC Canaan (Avalon 3 2U) $1,800.00 Apr-14 1037 890 1.17 Bitcoin ASIC Bitmain (Antminer S3) $450.00 Jun-14 366 478 0.77 Bitcoin ASIC Prospero X-1 $415.00 Jul-14 75 100 0.75 Langland and Skordal Bitcoin CPU Intel (Core i7 990x) $339.98^ 1/2/2011* 130[2] 0.033 3939.39+
Energy Efficient Bitcoin Mining to Maximize the Mining Profit: Using Data from 119 Bitcoin Mining Hardware Setups
(2015) [13] Bitcoin FPGA Xilinx (Spartan 6) $1995.00# 1/8/2009* 10 0.22 45.45+ Bitcoin ASIC Goldstrike 1 - 1/4/2015* 500 504 0.99+ McCook Bitcoin ASIC Cointerra TerraMiner IV $15750.00* - 2200 2000 1.1 (2014) [14] Bitcoin ASIC KnC (Neptune) $12995.00* 2014* 2200 3000 0.733 Bitcoin ASIC Bitfury BF3500 $366.96* - 2800 3500 0.8 Bitcoin ASIC Bitmain (AntMiner S3) $450.00* 1/6/2014* 366 478 0.766 Bitcoin ASIC BitMain (Antminer S2) $3899.00* 1/4/2014* 1100 1000 1.1 Bitcoin ASIC Spondoolies (SP30 Yukon) $699.00* 2015* 2500 6000 0.417 Krishnan et al. Bitcoin CPU Intel (Core i7 950) $398.95# 1/6/2009* 150 0.0189 7936.5 (2015) [15] Bitcoin CPU Intel (Atom N450) $64.00# 1/1/2010* 6.5 0.0016 4062.5 Bitcoin CELL Sony PlayStation 3 $169.85# 1/11/2006* 60 0.021 2857.14 Bitcoin GPU ATI (Radeon 4850) $108.00^ 1/6/2008* 110 0.101 1089.1 Bitcoin GPU ATI (Radeon 5770) $194.99^ 1/10/2012* 108 0.2145 503.49 Bitcoin FPGA Digilent Nexys 2 500K $374.95^ - 5 0.005 1000 Bitcoin ASIC Monarch BPU 600 C $4,680.00 - 350 600 0.58 Bitcoin ASIC Block Erupter (Sapphire) $66.99^ - 2.55 0.333 7.65 Chen Bitcoin GPU ATI (Radeon HD 7970) $350.00 1/1/2012* 300[4] 0.6038 496.85+ [16] Bitcoin GPU ATI (Radeon HD 7950) $299.99 1/2/2012* 300[4] 0.517 580.27+ Bitcoin GPU ATI (Radeon HD 6970) $169.99 1/10/2010* 550[17] 0.3899 1410.61+ Bitcoin GPU Nvidia (GTX 770) $370.00 1/5/2013* 230[18] 0.123 1869.91+ Bitcoin GPU Nvidia (GTX 670) $289.00 1/3/2012* 170[19] 0.112 1517.85+ Bitcoin GPU Nvidia (GTX 660 Ti) $189.00 1/3/2012* 150[20] 0.096 1562.5+ Taylor Bitcoin CPU Intel (Core i7 990x) $339.98^ Feb-11 130[2] 0.033 3939.39+ (2013) [21] Bitcoin GPU Nvidia (GTX 570) $153.29^ Dec-10 219[3] 0.155 1412.9+ Bitcoin GPU ATI (Radeon 7970) $450.00 Jan-12 300[4] 0.675 444.44+ Rodrigues Bitcoin ASIC Bitmain (Antminer S7) $619.00 2015-2016 1175+ 4700 0.25 [22] Bitcoin ASIC Canaan (Avalon 6) $750.95 2015-2016 1015+ 3500 0.29 Bitcoin ASIC SP20 Jackson $248.99 2015-2016 845+ 1300 0.65 Croft
(2016) [23]
Bitcoin ASIC Bitmain (Antminer S7) $720.00 2015 1200 5000 0.24+ King
(2017) [24]
Bitcoin ASIC Bitmain (Antminer S3) $460.00 2014 330 440 0.75+ Biryukov and Pustogarov
(2014) [25]
Bitcoin CPU Intel (Core i7-2760QM) $378.00 Q4 2011* 45[26] 0.0096 4687.5+ Jeannot
(2014) [27]
Bitcoin ASIC Bitmain (Antminer U2) $16.00 2014* 2 2 1+ Kagalwalla Bitcoin FPGA Xilinx (Spartan6-150) $100.00* 2012 Q1* 15 0.21 71.42+ [28] Bitcoin FPGA Butterfly Labs (BFL Single) $274.00* - 40 0.415 96.38+ Bitcoin ASIC Canaan (Avalon) $499.95* 2012* 2.8 0.28 10.00+ Romano and Schmid Bitcoin CPU Intel (Core i7 3930K) $594.00* Q4 2011* 130 0.066 1969.69+ (2017) [29] Bitcoin CPU AMD (4x Opteron 6174) $3768.94* 1/9/2007* 320 0.115 2782.60+ Bitcoin CPU Intel (Xeon Phi 5100) $399.95# 1/11/2012* 225 0.14 1607.14+ Bitcoin GPU Nvidia (Tesla S2070) $1200.00# 1/11/2009* 900 0.749 1201.60+ Bitcoin GPU ATI (Radeon 5870x6) $1266.00! 1/9/2009* 1200 2.568 467.28+ Bitcoin FPGA Butterfly Labs (Mini Rig) $22484* - 1250 25.2 49.60+ Bitcoin ASIC Bitmain (Antminer S9) $1227.00* 2017* 1375 14000 0.098+
Energy Efficient Bitcoin Mining to Maximize the Mining Profit: Using Data from 119 Bitcoin Mining Hardware Setups
Bitcoin ASIC Spondooliestech (SP35) $2795.00# 1/1/2015* 3650 5500 0.663+ D’Herdt Bitcoin CPU Intel (Core i7 3930k) $594.00* Q4 2011* 130[29] 0.066 1969.69+ (2015) [30] Bitcoin GPU ATI (Radeon HD5770) $80.00* 1/10/2012* 108[15] 0.23 469.56+
Tedeschi (2017) [31]
Bitcoin ASIC Bitmain (Antminer S9) $3,000.00 2017* 1375 14000 0.098+ Heid
[32]
Bitcoin ASIC ATI 5850 Quad Crossfire $649.00 1/9/2009* 305[33] 1.384 220.37+ Forte et al. Bitcoin CPU Intel (Core i7 3930K) $594.00* Q4 2011* 130 0.066 1969.69+ (2015) [34] Bitcoin CPU AMD (4x Opteron 6174) $3768.94* 1/9/2007* 320 0.115 2782.6+ Bitcoin CPU Intel (Xeon Phi 5100) $399.95# 1/11/2012* 225 0.14 1607.14+ Bitcoin GPU Nvidia (Tesla s2070) $1200.00# 1/11/2009* 900 0.749 1201.6+ Bitcoin GPU ATI (Radeon 5870x6) $1266.00! 1/9/2009* 1200 2.568 467.28+ Bitcoin FPGA Butterfly Labs (Mini Rig) $22484.00* - 1250 25.2 49.6+ Bitcoin ASIC Bitmain (Antminer S5) $449.00# 1/12/2014* 590 1155 0.51+ Bitcoin ASIC Spondooliestech (SP35) $2795.00# 1/1/2015* 3650 5500 0.663+ Dini
(2014) [35] Bitcoin ASIC CoinTerraMiner IV $6,000.00 - 1200 1600 0.75+ Ho Bitcoin CPU Intel (Core i7 920) $500.00 1/11/2008* 200 0.025 8000+ (2013) [36] Bitcoin CPU Intel (Core i7 2600k) $400.00 1/1/2011* 400 0.025 16000+ Bitcoin GPU ATI (Radeon 6990) $900.00 1/3/2011* 600 0.9 666.66+ Bitcoin GPU ATI (Radeon 7750) $90.00 1/2/2012* 80 0.2 400+ Bitcoin ASIC Block Erupter (Sapphire) $80.00 - 4 0.5 8+ Bitcoin ASIC Block Erupter (Blade) $700.00 - 95 10.5 9.04+ Meiners Bitcoin GPU ATI (Radeon HD 7990) $999.00 1/4/2013* 555 1.2 462.5+ (2013) [37] Bitcoin GPU ATI (Radeon HD 6990) $699.00 1/3/2011* 375 0.75 500+ Bitcoin GPU ATI (Radeon HD 5970) $599.00 1/11/2009* 294 0.75 392+
Bitcoin ASIC BitForce SC $274.00 - 30 5 6+
Solberg (2017) [38]
Bitcoin ASIC Bitmain (Antminer S9) $1227.00* 2017* 1370 14000 0.097+ Malone Bitcoin CPU Intel (Core i7 950) $350.00 1/6/2009* 150 0.0189 7936.5 (2018) [39] Bitcoin CPU Intel (Atom N450) $169.00 1/1/2010* 6.5 0.0016 4062.5 Bitcoin CELL Sony PlayStation 3 $296.00 1/11/2006* 60 0.021 2857.14 Bitcoin GPU ATI (Radeon 4850) $45.00 1/6/2008* 110 0.101 1089.1 Bitcoin GPU ATI (Radeon 5770) $80.00 1/10/2012* 108 0.2145 503.49 Bitcoin FPGA Digilent Nexys 2 500K $189.00 - 5 0.005 1000 Bitcoin ASIC Monarch BPU 600 C $2,196.00 - 350 600 0.58 Bitcoin ASIC Bitmain (Antminer S9) $2,400.00 - 1400 14000 0.1 Jamali et al.
[40] Bitcoin ASIC Bitmain (Antminer S9) $3,000.00 2017* 1340* 14000 0.095+ Bonneau Bitcoin ASIC Bitmain (Antminer S9) $2,000.00 2017* 1370 14000 0.097+ [41] Ethereum GPU ATI (Radeon Rx Vega 56) $550.00 2017* 210[42] 0.036 5833.33+ Nijhuis Bitcoin CPU Intel (Core i7 960) $305.00* 1/10/2009* 130[44] 0.012 12833.33+ [43] Bitcoin GPU Nvidia (GTX 580) $399.00! 1/11/2010* 244[45] 0.06 4066.66
Bitcoin ASIC Bitmain (Antminer S9 & Antminer
Energy Efficient Bitcoin Mining to Maximize the Mining Profit: Using Data from 119 Bitcoin Mining Hardware Setups
+Efficiency was calculated using the following equation (Efficiency = Power consumption (W) / Hash rate (Gh/s)). #Current hardware price extracted from Amazon (this can be varied according to the time).
! Current hardware price extracted from Futuremark.com (this can be varied according to the time). ^refurbished hardware prices from Amazon (this can be varied according to the time).
*Released date and price found on the manufacturer’s website. - values are not available
4. DISCUSSION
Electricity cost is the second principal concern [12] when running the Bitcoin hardware. Although our results indicated that, Bitcoin miners such as KnC (Neptune), Bitfury BF3500, Spondoolies (SP30 Yukon), and
Spondooliestech (SP35) have a massive amount of computational power, they consumed over 2000 watts of
electricity. As a result, the amounts of Bitcoins that can be mined from each hardware cost a vast amount of electricity cost to make profits.
Finally, hardware costs mainly depended on the release date. Due to the growing mining difficulty [7], miners' trend to upgrade their hardware frequently; hence, original and second-hand hardware prices reduced rapidly. Refurbished and second-hand Bitcoin mining hardware is a good starting point for miners who are in the initial stage and miners who are planning to expand their computing power. On the other hand, release dates of hardware also can be used as a reference point to identify the generation of miners as the mining hardware is swiftly outdated.
The energy used in Bitcoin mining is becoming a big concern, especially with the increasing global hash rate. Therefore, the operating cost is (the electricity costs of running the hardware) analyzed by many papers used in this study. This cost will often determine the profitability of the mining hardware, as the hardware must be able to earn more money than it consumes. Consequently, the cost of energy is a limiting factor; newer hardware will have to have a higher hash rate and lower energy consumption.
When maximizing the Bitcoin mining profit, Solo and Pool Mining are also important [36]. The concept of solo mining is where miners communicate all mining-related information directly with the Bitcoin network. When a block is solved the hash puzzle and confirmed through the network, the reward is issued to the miner who mined the block. The advantages of solo mining are miners get a high reward for solving a block and can stay active for as long as the network remains active. The drawbacks of solo mining are unpredictable income and outdated mining practices.
Pooled mining is a concept of mining with a group of other miners as a team. The pool operator governs how the rewards are divided among the members and the fees that are included in each solved block. Currently, hash rates over mining pools outnumber solo hash rates 3 to 1 [36]. The benefits of pooled mining are stability and high-income rate. Therefore, pooled mining considers as a better concept than solo mining.
Since pool servers (mining pool) depend on a network, they do have downtime caused by natural or artificial actions or connectivity issues. Additionally, as pools must be optimized and maintained by operators, charged a cost for each block discovery. Consequently, profits over an extended period of time are lower than solo mining [36]. Therefore, pool-specific parameters must be considered when finding the most profitable pool.
Hardware heat transfer is also a significant challenge in maximizing the Bitcoin mining profit from the beginning. Keeping the hardware (ASIC/FPGA chips) cool [6] can result in higher hash rates. Manufactures used additional cooling systems such as fans, water cooling systems and heatsink to reduce the heat, which increases electricity usage. Therefore, setting up the Bitcoin mining hardware in a low-temperature location will be an advantage. Additionally, natural energy sources such as solar panels can be used to reduce energy cost. However, it may increase the capital investment for Bitcoin mining; nonetheless, it will be beneficial in the long run for individual miners and industrial mining farms. Conduit hydropower is another alternative considered by mining farms to utilize excess electricity. This is a method of using “mechanical energy of water as part of the water delivery system through man-made conduits to generate “green” electricity” [23] for Bitcoin mining and only suitable for large-scale Bitcoin miners.
Several assumptions were made by the researchers [4, 6, 23] when measuring the hash rates and power consumptions to calculate the efficiency. Firstly, according to Bitcoin hardware studies we assumed, miners were operated in ideal conditions (temperatures) which help to generate maximum hashes. Although certain mining
Energy Efficient Bitcoin Mining to Maximize the Mining Profit: Using Data from 119 Bitcoin Mining Hardware Setups
hardware needs additional cooling systems [1] to operate, their energy consumption values were not added to the total hardware power consumption.
According to our analysis, first-generation (CPU) and second-generation (GPU) miners need additional power supplies and hardware to operate. For instance, CPU or GPU cannot operate stand-alone without a power supply or a motherboard. However, all the additional hardware power usages were not considered when measuring the values in this research.
Comparing the hardware manufactured dates to the efficiencies, it is clear that Bitcoin mining hardware manufacturing companies are focusing on releasing powerful hardware with lower energy consumption. However, the hardware development process expected to slow down due to the technological barriers. Nonetheless, Koomey has described that mining hardware also follows the Koomey law [12] which is computer efficiency doubled every 1.5 years. Additionally, due to the competition between companies' miner prices expected to increase in the future. Therefore, significant companies are working on cloud/hosting solutions as the owners can delete the data anytime from the cloud and confirm the deletion [48] no matter how malignly the cloud server performs. BitFury has started offering the service to the customers [12] and many companies yet to begin due to the benefits consumer and the hardware companies could get.
Considering all the hardware efficiencies, it is clear that the first and the second-generation miners are not powerful enough to generate profits as they consume more than they generate. Therefore, mining beginners should use at least an ASIC or a FPGA miner and focus on hash rates to gain more profit.
REFERENCE
[1] M. Bedford Taylor, "The Evolution of Bitcoin Hardware", Computer, vol. 50, no. 9, pp. 60-61, 2017.
[2] "Intel® Core™ i7-990X Processor Extreme Edition (12M Cache, 3.46 GHz, 6.40 GT/s Intel® QPI) Product Specifications", Intel® ARK
(Product Specs), 2018. [Online]. Available:
https://ark.intel.com/products/52585/Intel-Core-i7-990X-Processor-Extreme-Edition-12M-Cache-3_46-GHz-6_40-GTs-Intel-QPI. [Accessed: 21- Mar- 2018].
[3] "GeForce GTX 570 | Specifications | GeForce", Geforce.com, 2018. [Online]. Available: https://www.geforce.com/hardware/desktop-gpus/geforce-gtx-570/specifications. [Accessed: 21- Mar- 2018].
[4] "Compare Products", Products.amd.com, 2018. [Online]. Available: https://products.amd.com/en-us/compare?prod1=20&type1=Desktop%20Graphics&prod2=21&type2=Desktop%20Graphics&prod3=&type3=&prod4=&type4=&prod5 =&type5=. [Accessed: 20- Mar- 2018].
[5] O. Ala-Peijari, "Bitcoin The Virtual Currency: Energy Efficient Mining of Bitcoins", pp. 58-76, 2014.
[6] A. Kampl, "Bitcoin 2-Phase Immersion Cooling and the Implications for High Performance Computing", Electronics Cooling Magazine, p. 1, 2014.
[7] K. O'dwyer and D. Malone, "Bitcoin Mining and its Energy FootprintBitcoin Mining and its Energy Footprint", pp. 3-4, 2014.
[8] L. Cocco and M. Marchesi, "Modeling and Simulation of the Economics of Mining in the Bitcoin Market", PLOS ONE, vol. 11, no. 10, p. 7, 2016.
[9] S. Walton, "AMD Radeon HD 6870 Review", TechSpot, 2018. [Online]. Available: https://www.techspot.com/review/325-amd-radeon-6870/page10.html. [Accessed: 15- Apr- 2018].
[10] S. Walton, "AMD Radeon HD 6950 Review", TechSpot, 2018. [Online]. Available: https://www.techspot.com/review/349-amd-radeon-6950/page12.html. [Accessed: 15- Apr- 2018].
[11] "AMD Radeon HD 6990 4GB Review | bit-tech.net", Bit-tech.net, 2018. [Online]. Available: https://www.bit-tech.net/reviews/tech/amd-radeon-hd-6990-review/3/. [Accessed: 15- Apr- 2018].
[12] J. Lee and Y. Kim, "An Analysis of Energy and Hardware Impacts on the Bitcoin Mining Network", ACM Transactions on Economics
and Computation, vol. 5, no., pp. 1-10.
[13] T. Langland and K. Skordal, "Mining Bitcoins using a Heterogeneous Computer Architecture", pp. 11-29, 2015.
[14] H. McCook, "An Order-of-Magnitude Estimate of the Relative Sustainability of the Bitcoin Network", A critical assessment of the Bitcoin
mining industry, gold production industry, the legacy banking system, and the production of physical currency, vol. 2, p. 25, 2014.
[15] H. Krishnan, S. Saketh and V. Tej, "Cryptocurrency Mining – Transition to Cloud", International Journal of Advanced Computer Science
and Applications, vol. 6, no. 9, p. 117, 2015.
Energy Efficient Bitcoin Mining to Maximize the Mining Profit: Using Data from 119 Bitcoin Mining Hardware Setups
[17] "AMD Radeon™ HD 6970", Products.amd.com, 2018. [Online]. Available: https://products.amd.com/en-us/search/Desktop-Graphics/AMD-Radeon™-HD-6000-Series/AMD-Radeon™-HD-6900/AMD-Radeon™-HD-6970/13. [Accessed: 20- Mar- 2018]. [18] "GeForce GTX 770 | Specifications | GeForce", Geforce.com, 2018. [Online]. Available: https://www.geforce.com/hardware/desktop-gpus/geforce-gtx-770/specifications. [Accessed: 20- Mar- 2018].
[19] "GeForce GTX 670 | Specifications | GeForce", Geforce.com, 2018. [Online]. Available: https://www.geforce.com/hardware/desktop-gpus/geforce-gtx-670/specifications. [Accessed: 20- Mar- 2018].
[20] "GeForce GTX 660 Ti | Specifications | GeForce", Geforce.com, 2018. [Online]. Available: https://www.geforce.com/hardware/desktop-gpus/geforce-gtx-660ti/specifications. [Accessed: 20- Mar- 2018].
[21] M. Taylor, "Bitcoin and The Age of Bespoke Silicon", pp. 6-8, 2013. [22] H. Rodrigues, "Crypto-Currency Mining", Computer Architechtures, p. 3.
[23] N. Croft, "Using excess conduit hydro-generation power for Bitcoin mining", INCOSE SA Systems Engineering Conference, vol. 12, pp. 44-46, 2016.
[24] M. King, "Interview with a Bitcoin Miner", pp. 5-6, 2017.
[25] A. Biryukov and I. Pustogarov, "Proof-of-Work as Anonymous Micropayment: Rewarding a Tor Relay?", p. 6, 2014.
[26] "Intel® Core™ i7-2760QM Processor (6M Cache, up to 3.50 GHz) Product Specifications", Intel® ARK (Product Specs), 2018. [Online]. Available: https://ark.intel.com/products/53474/Intel-Core-i7-2760QM-Processor-6M-Cache-up-to-3_50-GHz. [Accessed: 29- Mar- 2018]. [27] F. Jeannot, "Antminer U2 V1.2 ASIC Bitcoins Hash BM1380 test", pp. 1-2, 2014.
[28] A. Kagalwalla, "Brief Tutorial on Bitcoins", p. 9.
[29] D. Romano and G. Schmid, "Beyond Bitcoin: A Critical Look at Blockchain-Based Systems", Cryptography, vol. 1, no. 2, p. 27, 2017. [30] J. D'Herdt, "Detecting Crypto Currency Mining in Corporate Environments", p. 3, 2015.
[31] E. Tedeschi, "Trading Network Performance for Cash in the Bitcoin Blockchain", pp. 48-49, 2017. [32] A. Heid, "Analysis of the Cryptocurrency Marketplace", p. 11.
[33] s. (function() { var site_match = new RegExp('tomshardware|optimizelypreview|cache|google|yahoo|bing|archive', "Power Consumption And Noise - Radeon HD 5850: Knocking Down GTX 295 In CrossFire", Tom's Hardware, 2018. [Online]. Available: http://www.tomshardware.com/reviews/radeon-hd-5850,2433-13.html. [Accessed: 02- Apr- 2018].
[34] P. Forte, D. Romano and G. Schmid, "A critical look at blockchain-based systems", Beyond Bitcoin – Part I:, p. 27, 2015. [35] M. Dini, "Use of FPGAs in Cryptocurrencies", p. 19, 2014.
[36] C. Ho, "Adaptation of All-Programmable SoC to Hardware Bitcoin Miners and Mining Servers", p. 27, 2013. [37] J. Meiners, "Bitcoin Mining-Hardware", p. 6, 2013.
[38] E. Solberg, "Tokenizing Cloud Mining", Cryptotech, p. 6, 2017. [39] D. Malone, "Bitcoin", p. 10, 2018.
[40] R. Jamali, S. Li and R. Pantoja, "cryptocurrency digital asset class of the future – Bitcoin vs ethereum?", the economist kraken Bitcoin
exchange, p. 7.
[41] J. Bonneau, "Hostile blockchain takeovers", p. 4.
[42] Radeon™ RX Vega⁵⁶", Radeon Gaming, 2018. [Online]. Available: https://gaming.radeon.com/en/product/vega/radeon-rx-vega-56/. [Accessed: 02- Apr- 2018].
[43] J. Nijhuis, "Effect of IoT botnets on Cryptocurrency", p. 3.
[44] "Intel® Core™ i7-960 Processor (8M Cache, 3.20 GHz, 4.80 GT/s Intel® QPI) Product Specifications", Intel® ARK (Product Specs), 2018. [Online]. Available: https://ark.intel.com/products/37151/Intel-Core-i7-960-Processor-8M-Cache-3_20-GHz-4_80-GTs-Intel-QPI. [Accessed: 02- Apr- 2018].
[45] "GeForce GTX 580 | Specifications | GeForce", Geforce.com, 2018. [Online]. Available: https://www.geforce.com/hardware/desktop-gpus/geforce-gtx-580/specifications. [Accessed: 02- Apr- 2018].