Social media

On game theory on cryptocurrency mining with increasing DAG size

Author: LB 2021-04-28 204

If you have a powerful GPU in 2021 you can utilize it many ways. You play your favourite game with it and you can also mine cryptocurrencies. As bitcoin is not an option anymore for average cryptocurrency miners, ethereum, the second most popular cryptocurrency is getting the most attention because its mining profitability has skyrocketed in recent months. We all know that no miracle lasts forever and the ethereum community is looking into replacing their PoW verification algorithm and the ethereum 2.0 blockchain is already running in production. As a result, miners have started shifting their attention to alternative solutions. I’d like to share some useful thoughts from the cryptocurrency mining horizon in 2021.

Yesterday, I decided to perform a short verification on the data provided by This site publishes the current profitability for a given hardware, energy cost and cryptocurrency combination. When I looked at their data, ethereum was on the top of the list with $8.26 of daily profit for a given GPU combination. Currently, I don’t have any GPUs so my only option was to rent a virtual machine in the p3 resource group from AWS and mine cryptocurrencies on a Tesla V100 card.

1. Initial steps

I’ve some experience on how to mine other cryptos using GPU, so this time I decided to test 3 additional currencies. So I tried to mine ethereum using nanominer, ravencoin using kawpowminer, beam using lolminer and haven protocol using xmrig. I managed to launch a p3 instance with a fresh Ubuntu installed on it and after connecting to the virtual machine I updated the system and installed the Cuda 11.0 library, which is essential if you want to use these miners.

2. Mining ethereum with Tesla V100

Mining ethereum on a Tesla V100 card is pretty easy. You just have to download nanominer, set up your mining configuration in config.ini and you can launch the miner. The Tesla card was producing roughly 93 megahashes per second and using the current market rate this would equal to a daily profit of $8.26. The energy consumption varied around 208W when I was running nanominer. So this profit value corresponded the the one listed by, and no doubt, the mining profitability of ethereum is mindblowing.

3. Mining ravencoin with Tesla V100

The second coin I decided to mine today was ravencoin. kawpowminer is one of the miner softwares that can be used to mine ravecoin and this time I decided to compile the miner from its source. Compiling kawpowminer takes some time, but hopefully I got a more optimized version of the miner, than just running a precompiled binary on the virtual machine. Then I launched the miner and it was mining with 43.50 Mh/s and the energy consumption was around 283W. I used an online calculator to find out my mining profitability with ravencoin and it turned out to be somewhere around $5.41 per day. This is clearly less than mining ethereum, so rational miners, would never ever choose to mine ravencoin as long as they’re aware of the existence of ethereum. The electricity consumption is slightly larger than in the case of ethereum mining, however electricity cost is usually almost insignificant to miners with energy efficient new cards. On the other hand, miners with older models will have to pay attention to this factor too, as with increasing electricity costs their mining profit might completely disappear.

4. Mining beam with Tesla V100

Mining beam is also a very straightforward process if you have a GPU and a Ubuntu system. I used the precompiled lolminer from Github and the Tesla V100 card was producing 40 solutions per second. The GPU’s power consumption was around 200W and with this my mining profitability was around $3.75 daily before deducting any cost.

5. Mining haven protocol with Tesla V100

I used to be a huge fan of monero, so I enjoy playing around with monero based cryptocurrencies as well. Mining haven protocol in the AWS cloud is also very easy, I just had to download the xmrig and the xmrig-cuda source code from Github and compile both of them. As soon as the xmrig-cuda dynamic object was ready, I copied it next to the xmrig binary and I launched the miner. The miner was producing 1947.6 H/s and the energy consumption was around 150W. I used an online calculator to approximate my daily profits with 1947.6 H/s and it turned out to be around $3.57 with this hardware and software combination. This data was significantly different from what I found on The site said that mining haven protocol is somewhat less profitable than mining ethereum, but I found that on this given day its profitability was worse than all of the currencies listed above. Interestingly, the card was using the least amount of energy in this case, so we can conclude that the power consumption varies with the miner you’re running on your machine and this factor should also be considered when building a mining rig.

6. Final words

Obviously, if you have a very expensive video card, such as the Tesla V100 with 16Gb of onboard memory and very good energy efficiency, you hardly have to think about these issues. You can launch your miner and you can just watch ethereum balance grow day by day. If you have 16Gb of GPU memory, the ethereum DAG size, which has just surpassed 4Gb recently is also something you don’t have to worry about too much. But the case is altered for miners, with old cards, because they either have 4Gb cards or 8Gb cards, so for them whether it’s still profitable to mine ethereum on a 4Gb in zombie mode or switching to another cryptocurrency to mine is a more rational move is a huge question. As mining cryptocurrencies other than ethereum usually doesn’t require massive GPU memory, miners with older mining rigs are going to choose those with high probability.

So there are several parameters we have to consider in this scenario. If you incorporate game theory into the mining scene, you get a very interesting picture and we know that game theory is not at all orthogonal to the cryptocurrency world. The Miner's Dilemma by Ittay Eyal for example is a very similar topic.

If you consider cryptocurrency mining as an artefact of game theory the following assumption is common. Miners act in a rational way to maximize their profits. In other words, as soon as a miner realizes that his mining rig could earn more profit by hopping onto another cryptocurrency, he executes the necessary modifications immediately. While it’s true that it’s nearly impossible to see through the entire cryptocurrency world and to find the optimal solutions to optimize profits, miners that enable automatic switching between the mining of several cryptocurrencies are being developed and miners are able to adjust their mining rigs as a reaction to these changes very quickly.

Clearly, power consumption, the DAG size in ethereum and the profitability of other cryptocurrencies are the most important parameters that influence the decision making of the miners. While miners with older GPUs are more vulnerable to the DAG size problem, miners with new equipment can keep on mining ethereum as long as the DAG size doesn’t exceed their GPU memory. Whether it’s rational to mine ethereum on a GPU with 16Gb memory or if there is any room for improvement in these cases are good questions and the answer to these questions could lead us further on the cryptocurrency mining roadmap. Reach out on Whatsapp (+36 30 8316972) if you want to know more on cryptos.

Interesting entries


Russian miners are already testing the Nvidia CMP 30HX chip for mining


Technical details and the ROI of Chia Network Farming


Investigating monero mining on a notebook


Optimizing the mining of monero on a multithreaded CPU


On game theory on cryptocurrency mining with increasing DAG size