Contents
Why compare ETH mining software?
Clearly, any miner would like to know which mining software is the best, the most reliable, and has the highest hash rate. It’s important to look at the hashrate reported by the pool rather than by mining software. Pools pay you for shares received from mining software. Don’t be fooled by the hash rate displayed in your rig console as it’s completely irrelevant.
So-called reported hashrate has nothing to do with your profit.
We regularly see debates about the best mining software on Youtube, Reddit, Telegram chats, and on forums. “T-rex is the fastest, Phoenix overstates its hashrate, Gminer is the best of all, lolMiner is the most reasonable option, nothing compares to TeamRedMiner, you can’t figure out NBMiner if you’re not Chinese” – that’s what we hear all the time.
What is wrong with our old experiment and why are we holding the new one?
- There are new software versions available.
- The old experiment lacked precision and we figured out how to fix it.
- Ethereum price is at its record high and you’ve got to make the most of your mining hardware.
Test Conditions
We tested the best ETH mining software to see which is the best of the best. How do you detect the best? Simple. The more shares mining software sends to the pool, the higher the reward it gets. It doesn’t matter whether a pool uses PPLNS, PPLNT, PPS, or PPS+ as a reward system.
More shares mean more money.
The developer’s fee is completely irrelevant to us. We counted only those shares that reached the pool and appeared in the miner balance. It doesn’t matter how long mining software is mining to the developer’s wallet. The experiment lasted 3 hours, so a developer’s fee was mined multiple times.
We chose two configured rigs for the experiment: one with 9 Nvidia 1660Ti graphics cards, and the other with 4 AMD 5700 graphics cards. They mine Ether in the 2Miners ETH pool 24/7, we didn’t change settings for any specific piece of mining software. Rigs were operative for 3 hours for each program. The experiment starts when mining software launches.
We created an Ethereum pool with a low share difficulty.
You can read more about shares and share difficulty in this post: What is Share and the Share Difficulty When You Are Mining at the Pool.
Current share difficulty in the Ethereum pool is 8.72G. We set it at 64M for the test pool, which is 136x lower so that we get a lot of shares and don’t have to keep the experiment running for months. It is an optimal value to maximize the number of shares and make sure that mining software handles the load and the processor doesn’t slow down testing.
3 hours in this test pool equal 408.75 hours (17 days) in a real pool.
How Pools and Mining Software Calculate Hashrate
We agreed that the number of shares in the pool at the end of the experiment is the most important parameter. Plus, the final table features two types of hashrate:
- Real (from the pool). It’s an average hashrate obtained during the 3-hour experiment.
- Hashrate from mining software, that is what you see in the program when it’s running.
We took it from the miner log at the very end of the experiment. It was the last hash rate value before closing the software.
Hashrate in mining software is pretty straightforward. The program goes through millions of block solutions per second. Make sure to read: What Is Mining? Mining Luck. The software calculates how many solutions it already went through, and it doesn’t matter whether they are good or bad. Every new solution adds +1 to the count.
Say you see 200 MH/s in the miner window. It means that your mining software uses your graphics cards to go through 200 million solutions per second.
On the contrary, hashrate in the pool needs to be calculated. Each share sent to the pool has a difficulty no less than 64M (share difficulty is set by the pool). Say mining software sends 1200 shares in 10 minutes. So the average number of shares it sends per second is 1200 / 10 / 60 = 2. We multiply 2 shares by a share difficulty of 64M and get a hashrate of 128 MH/s.
In cryptocurrency mining everything can be explained and calculated. Blockchain remembers everything.
Testing NVIDIA Mining Software
Mining rig used for testing:
- Motherboard: MSI Z270 A-pro
- Processor: Celeron G3930
- RAM: 12 GB
- SSD: 120 GB
- 2 PSUs: DeepCool KCAS-1000M 1000W
- 9 GPUs: Nvidia MSI Gaming X GTX 1660 Ti 6 GB
- Overclocking: core -650, mem 2290, PL 70
- Driver: Nvidia 460.39
- OS: Linux
If you don’t know how to overclock GPUs, read: What Is Mining? Beginning Miner’s Bible & GPU Overclocking on Different Algorithms. Then go to our Telegram chat about mining.
Mining software featured in the experiment:
Best Ethereum Mining Software for Nvidia. Test results
Gminer came first. T-rex came a close second. It’s worth mentioning that the two became winners in the previous experiment as well. The only difference was that Gminer came second at that time.
How accurate is mining software hashrate? We think it’s quite accurate. 285 MH/s in the miner window vs 282 MH/s in the pool – the difference is caused by the developer’s fee.
Mining software developer’s fee is charged this way: for a short period of time, the software is mining to the developer’s wallet. Usually less than a minute every hour.
The only open-source mining software Ethminer came third. It doesn’t charge a developer’s fee, so hashrates in the program and in the pool are almost identical with a 0.01% difference. As you can see, if mining software is free, it doesn’t mean that you get more profit. If you choose Gminer or T-rex, you gain more.
LolMiner came fourth. LolMiner released its Nvidia mining software just a few months ago, but it already surpassed the most popular mining software Phoenix in terms of efficiency.
Phoenix and NBMiner came last. Both have much higher hashrates in the miner window than in reality. As to Phoenix, we already established last time that we should look not at its Eth speed, but at its Eth: Effective speed. This time Effective Speed was 276.86 MH/s which seems accurate and corresponds to the number of shares in the pool. Phoenix shows hashrate in 3 ways: speed, average speed (5 min), effective speed. We think you should only look at effective speed.
Example from the miner 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
Sadly, NBMiner hashrate in the miner window turned out to be higher than in the pool. Hashrate reported by NBMiner was higher than that of other mining programs. We checked the log and didn’t see any rejected or stale shares.
It means that Phoenix and NBMiner overstate their hashrates which is in no way beneficial to you.
Testing AMD Mining Software
Mining rig used for testing:
- Motherboard: MSI Z270 A-pro
- Processor: Celeron G3930
- RAM: 4 GB
- SSD: 120 GB
- PSU: Cougar CMX 1200W
- 4 GPUs: AMD Power Color RX 5700 Reference with Micron memory, flashed into XT
- Overclocking: core 1300, mem 930
- Driver: AMD driver 20.40
- OS: Linux
Mining software featured in the experiment:
Best Ethereum mining software for AMD. Test results
TeamReadMiner came first just like last time. This time it’s even more distant from its competitors with a hashrate 5% higher than that of the runner-up. TRM hashrates in the miner and in the pool are almost identical (it is possible that TeamRedMiner doesn’t count the dev fee when it calculates hashrate).
Phoenix and lolMiner came second and third with a tiny difference. Phoenix hashrate was too high as always, while effective speed was extremely accurate: 214.58 MH/s.
We remind you to look only at Effective Speed if you use Phoenix Miner.
NBMiner came fourth lagging behind TeamRedMiner by a margin of 10%.
We wanted to test GMiner with AMD video cards as well but it failed to start most likely due to the high memory overclocking. We’ve sent a bug report to the GMiner developers and we hope to see this mining software in our future AMD tests.
Confidence Level and Future Testing
This time our experiment was 5 times more accurate than last time (we tested 5 times more shares). It is still not enough to hold a highly accurate experiment. TeamReadMiner shared their thoughts about comparing mining software. They say that it’s important to consider mining as a Poisson process with all consequent measurement errors.
In our experiment with 50 thousand shares we have a confidence level of slightly less than 99%. If you look at the results, it’s clear that 1% could make a huge difference.
If we want to have a truly accurate experiment, mining software should send 10 million shares each. It means that testing for each piece of mining software would take a month! Way too much. By the time we finish such an experiment, developers will release upgrades and the experiment will become useless. With that being said, we do want to hold new experiments in the future, with upgraded mining software and higher accuracy.
It’s worth mentioning that if mining software developers report their hash rate honestly and don’t have errors in the code, you can trust the hashrate in the program when you are choosing mining software.
Our experiment is supposed to help you, but don’t take it as absolute truth. It would be wrong to say that Gminer and TeamRedMiner are always the best, and NBMiner is the worst. You may get different results depending on your GPUs, overclocking settings, and rig configuration.
Remember to follow us on Twitter to get all the news as soon as possible and join our Telegram chat with more than 3000 active users.
