In October 2020, we have already conducted a detailed test of miners for mining Ethereum ETH. Then the winners were t-rex and gminer for Nvidia video cards, as well as TeamRedMiner and lolMiner for AMD video cards.
Three months passed, so we decided to repeat the experiment. However, this time we approached the testing process even more thoroughly.
Why compare ETH mining software?
The answer is obvious. All miners (we mean people) want to know which miner (we mean a program) is the best, namely the most reliable and with the highest hashrate. Moreover, the hash rate is important precisely on the pool, and not in the program window. Yes, this is important, because the pools pay you for the balls that they received from the miner, and how fabulous hashrate is drawn in the farm console is absolutely not important.
The so-called “Reported hashrate” does not affect profits, remember this.
Holivars on the topic “which miner is better” are conducted regularly on all fronts: on YouTube, Reddit, in Telegram chats, on the forums. “T-rex is the fastest, Phoenix overestimates the hash rate, Gminer will beat them all, lolMiner is the most adequate, nothing beats TeamRedMiner, NBMiner can only be understood by the Chinese” – these and many other phrases we hear every day.
But what’s wrong with our old test, and why conduct a new one? There are several reasons.
- new versions of programs were released.
- the old test was not accurate enough, and we figured out how to make the research more accurate.
- Ethereum’s price is breaking all possible records, so you need to squeeze the most out of your hardware.
- Experimental conditions
We took the very best ETH mining software and ran a test to determine the best of the best. How is this possible? Very simple. The more the share the miner sends to the pool, the more reward he will receive. It doesn’t matter which reward system the pool uses: PPLNS, PPLNT, PPS or PPS +.
More ball on the pool = more money. This is the law.
It was absolutely not important for us what the miner’s commission was. We compared only those balls that reach the pool and are credited to the account of our miner. How many minutes the program mined to the developer’s wallet and how it did it is absolutely unimportant. The test in each program was carried out for three hours, which means that in three hours the miners managed to remember their commission several times.
For the experiment, we chose two configured rigs: one of 9 Nvidia 1660Ti cards, the other of 4 AMD 5700 cards. These rigs successfully mine ether on the 2Miners mining pool around the clock, so we did not change any settings and did not change them “for each miner “. Rigi worked exactly three hours on each program. The countdown always started from the beginning of the program start.
We have created a special Ethereum pool with a low difficulty level.
More details about balls are described in the article “What are balls in mining. How the complexity of decisions affects the mining of cryptocurrency. By the way, the ball is a word that is simply taken from the English language, that is, “share” or share. The miners did not come up with Russian terminology, which is why they began to use the English “share”. If you think that we are talking about balls / balls – as in billiards for example – then this is not so.
Currently, the Ethereum pool uses the 8.72G ball. On the test pool, we made the difficulty 64M, that is, 136 times less – in order to get a lot of balls and not carry out the test for months. We consider this to be the optimal value for obtaining the maximum number of shares, but at the same time so that the miner can cope with the work, and the processor is not overloaded and slows down the process.
How is the hashrate determined on the pool and in the miner?
We agreed that the number of balls on the pool at the end of the test is the most important parameter. In addition, the table with the results shows the hash rates:
real, that is, the hash rate that turned out to be on the pool. It should be noted that here we are talking about the average hash rate for the entire experiment time of three hours.
hashrate in the miner or what you see in the program when it is running. It was taken from the miner’s log at the very end of the work, that is, the last hash value was taken before the miner was turned off.
The hashrate in the miner is determined quite simply, the program iterates over the block solutions – millions of solutions per second. Be sure to read “What is mining? Good luck in mining. ” The program counts how many solutions it has already gone through: good, bad – it doesn’t matter. With each new decision, +1 is added to the counter.
Let’s say you see 200 MH / s in the miner window, which means that the program, using your video cards, goes through 200 million solutions per second.
The hash rate on the pool is a calculated value. Each ball sent to the pool has a difficulty of at least 64M, because the difficulty of the ball is set by the pool. Let’s say the miner sent 1200 shares in ten minutes. This means that the miner sends an average of 1200/10/60 = 2 shares per second. We multiply 2 balls by the complexity of the ball 64M and have a hash rate of 128 MH / s.
Remember that in mining and cryptocurrencies, everything can be understood on the fingers and counted. The blockchain remembers everything. If you don’t understand something, read the Cryptocurrency Bible.
Testing miners for NVIDIA video cards
The mining farm where the test took place:
- MSI Z270 A-pro motherboard
- Celeron G3930 processor
- RAM 12 GB
- SSD 120 GB
- 2 power supplies DeepCool KCAS-1000M 1000 W
- 9 Nvidia MSI Gaming X GTX 1660 Ti 6 GB Graphics Cards
- Overclocking settings: core (core) -650, memory (mem) 2290, consumption limit (PL) 70
- Nvidia driver 460.39
- Linux operating system
If you are still afraid of the word Linux, it’s in vain. Linux-based RaveOS is now completely free for 2Miners pool users, can be installed in a few clicks, and your life will become much easier. Forget Windows like a bad dream. Recommended.
If you don’t know how to overclock video cards, read our article “What is mining? The Bible of a beginner miner and the basics of overclocking video cards on different algorithms ”and then go to our Telegram chat on“ hardware ”, which already numbers more than 2 thousand people.
Programs that participated in the experiment:
- GMiner 2.43
- T-Rex 0.19.9
- Ethminer 0.19.0
- lolMiner 1.20
- Phoenix Miner 5.5c
- NBMiner 36.1
Unfortunately, GMiner did not run on our AMD farm. Perhaps the cause was inappropriate overclocking parameters. We have sent a bug report to the program developers. They promised to sort out the problem as soon as possible, and we hope that in the next test we will see GMiner in the list of AMD miners as well.
Best Ethereum miner for Nvidia GPUs. Test results
The first place was taken by Gminer with a minimal margin. Right behind him is the T-rex. We would like to note that these miners became the absolute winners of the last test. This time they overtook everyone again, only now Gminer became the first.
Do miners show the hashrate honestly? We think it’s fair. 285 MH / s in the miner window and 282 MH / s in the pool, a few percent difference caused by the miner’s commission.
The miners’ developer commission is charged as follows: for a short time, the program mines to the developer’s wallet. Usually, once an hour, a little less than one minute, the miner mines for the developer.
The third place belongs to the only open-source miner Ethminer. As you can see, the miner has no commission, so the hash rate on the pool and in the miner coincided almost perfectly, the difference was 0.01 percent. However, as you can see, you shouldn’t think that a free open-source miner will bring you more money because it doesn’t feed the developers. If you choose Gminer or T-rex, your reward will be higher.
The fourth was the newcomer of the lolMiner test. It is worth noting that lolMiner released its Nvidia miner just a few months ago, and has already surpassed the most popular Phoenix miner in performance.
At the bottom of the rating are Phoenix and NBMiner. Strangely, both the hashrate in the miner window turned out to be much overestimated than in reality. As for Phoenix, we have already determined during past tests that you cannot focus on its hash rate, you can only watch the so-called effective speed. In this experiment, the Effective Speed was 276.86 MH / s, which is very similar to the truth and is confirmed by the number of balls on the bullet. Phoenix has three hash rate displays: speed, Average speed (5 min) and Effective speed.
In our opinion, you can only focus on Effective speed! Be careful.
An example from the miner’s log:
- Eth speed: 284.577 MH / s
- Eth: Average speed (5 min): 284.557 MH / s
- Eth: Effective speed: 276.86 MH / s; at pool: 276.83 MH / s
Unfortunately, during the experiment, the hashrate in the NBMiner program turned out to be higher than in the pool. Surprisingly, the hashrate in the program was higher than that of all other miners! We checked the log, there were no rejected balls – stale or reject.
We conclude: Phoenix and NBMiner overestimate the hash rate, but it will not bring you any profit.
Testing miners for AMD video cards
The mining farm where the test took place:
MSI Z270 A-pro motherboard
Celeron G3930 processor
RAM 4 GB
SSD 120 GB
Cougar CMX 1200W 1200 Watt Power Supply
4 AMD Power Color RX 5700 Reference graphics cards, Micron memory, flashed into XT
Overclocking settings: core 1300, memory 930
AMD driver 20.40
Linux operating system
Programs that participated in the experiment:
Team Red Miner 0.8.0
Phoenix Miner 5.5c
The best Ethereum miner for AMD graphics cards. Test results
TeamRedMiner took the first place once again. This time he broke much more from the competitors, his hash rate was almost 5 percent higher than the second place. It is curious that the TRM hash rate on the pool and in the miner turned out to be almost identical. This may indicate, for example, that TeamRedMiner does not take into account the commission when calculating the hash rate.
The second and third places were taken by Phoenix and lolMiner with a tiny difference. Note that Phoenix, as usual, showed too high a hash rate, but Effective Speed was super accurate: 214.58 MH / s.
The fourth place belongs to NBMiner, and the gap from TeamRedMiner was almost 10 percent.
Test bias and future tests
This time our test was five times more accurate than the last time because we analyzed five times the ball. However, this is not enough to conduct a highly accurate test. The developers of TeamRedMiner laid out their thoughts that for comparing miners, the results should be considered as a Poisson distribution with all the resulting errors.
In our experiment with 50 thousand balls, we can talk about an error of more than 1 percent. If you look at the results of the experiment, it becomes clear that 1 percent can seriously change the place in the “leaderboard”.
To talk about a really accurate experiment, each miner must send 10 million shares, which means that the test of each program must last for a month – an unaffordable luxury. Miner developers will already release updates by this time, in general, the test will be down the drain. However, in the future, we plan to conduct new tests with new versions of miners and with higher accuracy.
We would like to emphasize the fact that if the developers of miners honestly consider the hash rate and did not make mistakes in the code, then when choosing a miner, you can safely focus on the hash rate that he writes.
Please use this research primarily as a guide, not as a common truth. Do not think that Gminer and TeamRedMiner are always the best miners, period, and NBMiner is the worst. The results may be different for your cards, overclocking and farm configuration.