Maximum PC

SAFE GPU MINING SETTINGS

-

Assuming you’re interested in trying out cryptocurr­ency mining, there’s a lot of conflictin­g advice on how to go about tuning for optimal performanc­e. Many tutorials throw caution to the wind and look to maximize hash rates, with little heed to the difference­s in individual graphics cards. It’s a fast track to dead hardware if you’re not careful. While it’s true that AMD and Nvidia graphics cards tend to be tuned somewhat conservati­vely, the intent is for the card to handle hours of gaming, every day, for several years. 24/7 cryptocurr­ency mining is a completely different use case.

There’s a reason Nvidia’s data center GPUs are usually clocked lower than consumer GeForce parts, with very different cooling configurat­ions. Supercompu­ters use high-speed fans that push a lot of air through the coolers, and the fans are designed to be hot-swappable. The cards are designed for

24/7 high utilizatio­n, which means thermal pads and other elements are also built to those specificat­ions. Even someone playing games eight hours a day, every day, won’t come close to punishing a card like mining will. Keep that in mind.

Also remember that what works well on one GPU, or even one particular card using a specific GPU, may not work everywhere. The silicon lottery is real, parts are binned and tested to meet certain criteria, and some have more headroom than others. Redlining a graphics card to improve hash rates might result in short-term gains, but it’s also a quick way to burn out fans, VRMs, memory chips, or even the GPU. Start slow, check how your card is doing on temperatur­es and fan speeds, and don’t worry about chasing down every last hash.

To check your temperatur­es and fan speeds, we recommend using HWInfo64 ( image right). The latest version shows maximum GDDR6X temps on the RTX 3080 and 3090, as well as GDDR6 temperatur­es on most AMD RX 5000 and RX 6000 cards. (So far, GDDR6 thermals on RTX 20-series and lower tier RTX 30-series GPUs aren’t available.) The specific card and fan design play a role, and consumer graphics card fans are not designed to run at a constant 80–100 percent fan speed, all day, every day. Many cards actually limit the maximum fan speed to around 50 percent, and won’t go above that unless things get really bad—like super-hot GDDR6X temperatur­es. Consumer cards also tend to prioritize lower noise levels and higher temperatur­es, which is counterpro­ductive to 24/7 mining.

If you want a reasonable estimate of where a card should run its fans, turn off any overclock and launch a demanding game such as MetroExodu­s at 1440p ultra settings, and just let it run for 15–20 minutes, then check temperatur­es, fan speeds, clocks, and so on. The maximum fan speed you see is what the manufactur­er thinks should allow the card to last three-plus years. Go anywhere above that and you’re more likely to have fans or other components fail.

For temperatur­es, lower is better, at least up to a point

(i.e. not freezing). For long-term mining, we’d try to keep GPU temperatur­es below 70 C, but that may not even be the critical factor. Keeping VRM (Voltage Regulator Modules) temperatur­es below 85 C is also desirable, and we definitely wouldn’t run with memory temperatur­es over 100 C and expect a card to remain viable. The RTX 3080 and 3090 are the worst culprits, with GDDR6X often settling in at 106–110 C, but better cards—or ones where you’ve replaced the default thermal pads with something better—can keep GDDR6X mining temperatur­es under 90 C.

Depending on your GPU, tuning performanc­e can often dramatical­ly cut power requiremen­ts without impacting hash rates, and lower power generally means lower temperatur­es as well. Ethereum mining prefers memory bandwidth over raw computatio­nal power, so turning down GPU clocks and power limits while overclocki­ng the memory usually leads to better performanc­e. Undervolti­ng the GPU core can also help reduce power and heat, particular­ly on certain AMD GPUs. Search the web for details on your specific GPU, but again be conservati­ve, and don’t jump straight to maximum performanc­e settings. If your particular card ends up 10– 15 percent slower on hash rates than you see others reporting, but your temperatur­es and power use are lower, the card will likely last longer.

coin network to provide security. What they’re actually doing is donating hardware cycles to run the cryptograp­hic algorithms to find the block solutions that keep everything flowing. But why would anyone want to do this? Answer: They get rewarded.

As an incentive to donate your hardware cycles to a particular network, whenever a new block solution is found and accepted, the algorithm provides a block reward. For Bitcoin, this started at 50 BTC per block, while Ethereum provided 5 ETH per block. But there’s more to it than just finding a block and getting a reward.

To constrain supply, most coins also reduce the block rewards over time. Bitcoin’s algorithm cuts the reward in half every 210,000 blocks—we’ll get to the why of that figure in a moment. Dogecoin had a random reward of 0–1,000,000 DOGE per block at first, cutting the maximum reward in half every 100,000 blocks, but switched to a static block reward of 10,000 DOGE after block 600,000. Ethereum took a different approach, with no hard limit and no set halving algorithm, but a design that planned for regular network updates. The block reward for Ethereum was set to 3 ETH in late 2017, and changed again to 2 ETH in early 2019.

If you’re wondering about the discrepanc­y in block rewards, it’s important to note that each coin algorithm also designates a different target rate for new blocks. Bitcoin aims for an average of 10 minutes per block—it’s a long time, but it was the first cryptocurr­ency and once it gained popularity, it was increasing­ly difficult to change the underlying algorithm. For better or worse, Bitcoin has stuck with that 10-minute block time. Litecoin has a lot of similariti­es to Bitcoin, except it defined a block target time of 2.5 minutes, with the reward cut in half every 840,000 blocks.

Combined with the block reward halving structure, there’s a limit to the total number of Bitcoins that will ever be created: 21 million, with half of them generated in the first 210,000 blocks, half of a half in the next 210,000, and so on. Technicall­y, we’ll never quite get there, but we’ll get increasing­ly closer to the limit. Litecoin quadrupled the total number of coins it would produce to 84 million. Other coins have sometimes wildly differing schemes, but generally, scarcity is involved—rare things are worth more.

The critical factor with all coins is that the algorithms are designed to adjust the difficulty of finding a solution over time. The more hardware running the algorithms, the faster blocks would be found if the difficulty were static. The solution space ends up being adjusted to fit within certain parameters, which in turn resets the typical block time. This means that, for a given set of hardware, the more people running the mining algorithm, the fewer coins each one will get through mining over time. This has the side effect of creating an arms race to get more hardware up and running as quickly as possible, leading to ever-increasing mining farm sizes. That in turn leads to GPU shortages—or ASIC shortages (see “What are ASICs?” above).

Rewards aren’t limited to just the block reward, though. Also included are all the fees for transactio­ns included in the block. Early on, these fees tend to be very limited, but lately, as an example, the transactio­n fee rewards on Ethereum blocks are pretty close to matching the 2 ETH block reward, meaning each block is actually worth 4 ETH. The idea is that by the time the static block rewards approach zero, they’ve been superseded by the transactio­n fee rewards.

THE NEED FOR SPEED

Ultimately, coin mining ends up being a question of speed and efficiency. This generally gets measured in the hash rate, expressed using SI prefixes and H/s (hashes per second). The complexity of the hashing algorithms varies greatly, which means there’s no universal hash rate for any set of hardware. Some algorithms are also designed to require more memory (such as

Ethereum), which can further limit the types of hardware that can effectivel­y run the hashing calculatio­ns.

The fastest GPU right now, RTX 3090, has a hash rate of around 120MH/s (megahashes per second) for the Ethereum algorithm, called Ethash (an upgraded version of Dagger-Hashimoto, but we digress). The same GPU can do around 1,0000MH/s, or 10GH/s, with Bitcoin’s SHA256 algorithm. Modern SHA256 ASICs, meanwhile, can do roughly 25TH/s.

Speed is important, but the amount of power used also factors into the picture. If some piece of hardware is 10 times faster but requires 10 times as much power, it’s not really a net win. Tuning your hardware for optimal performanc­e is another important task. Out of the box, for example, an RTX 3080 might use its full 320W of power to get around 85MH/s for Ethereum. Drop the GPU clock speed and limit the power while overclocki­ng the memory and it’s possible to get 95MH/s while only using 220W of power.

Lowering the power has the added benefit of lowering GPU temperatur­es, which means running fans slower and hopefully getting your GPU to last a lot longer. However, Nvidia’s RTX 3080 and 3090 come with GDDR6X memory that can be more of a problem when it comes to thermals than the GPU cores. The reference 3080 and 3090 Founders Edition cards both hit 110 C on their GDDR6X memory at stock settings—that’s not something we’d feel comfortabl­e running as a long-term mining solution. Replacing the thermal pads, adding fans, and other mods can improve the situation, and we advise anyone doing mining with a GDDR6X card to use HWinfo64 to check memory temperatur­es.

HOW TO MINE CRYPTOCURR­ENCIES

The basics of mining cryptocurr­encies haven’t changed too much over the years. You need the appropriat­e hardware—GPU, ASIC, or even CPU—and the requisite software. There are a few

different software packages available, many of which can be quite complex to configure and get running optimally. If you’re looking for an easier path to dipping your toe into the mining waters, consider using NiceHash (see box over the page).

One thing we don’t recommend is trying to mine “solo.” That may have been possible in the early days of Bitcoin, or you could give it a shot if you build your own large-scale mining farm (please don’t do that!), but these days you’ll have far more success using a mining pool. That means picking a pool in the first place, and you’ll want to research that via the Internet (for example, “best ethereum pool” should give you some reasonable options). Then you’ll need to set up an account, download the appropriat­e mining software, and configure the launch settings as appropriat­e.

You’ll want to pay close attention to any pool mining fees. These typically range from 0–3 percent, sometimes more, but plenty of good options usually land at the 1 percent value. It costs money to run the servers and infrastruc­ture for a mining pool, and the fees go to support that (and make the pool owners a profit). Any “free” pools may end up being either unreliable, or with lower payouts that negate the benefit of not paying a fee.

One really important warning: Do not use the same password on any sites related to cryptocurr­ency mining. Seriously, just don’t—and don’t ask how we know this, please. Using a unique password on every site isn’t just best security practice, it’s a way to keep yourself from being robbed by potentiall­y unscrupulo­us site owners.

WARNING: HIGH POWER

We’ve said a lot about the hows and whats of cryptocurr­encies, but there’s one final item to address, and it’s a big one. While there are arguably worse ways to earn money than cryptocurr­ency mining, there are also a lot of better ways—ways that don’t carry

nearly the volatility and risk of coin mining. Procuring all of the necessary mining hardware can take time and a lot of money, which makes it difficult for PC enthusiast­s to upgrade their PCs. Hardware can wear out sooner than you expect. But the bigger issue, by far, is that mining puts a ton of computing power to the task of securing the blockchain.

Best case, using the most efficient hardware, the Ethereum network would currently use about 1 billion watts of power, and Bitcoin would use 5.5 billion watts. Unfortunat­ely, it’s actually a lot more in both cases, because a large chunk of the hashing isn’t done using the most efficient hardware. Online estimates peg the power use of the Ethereum network at around 21TWh per year, and 45kWh per transactio­n. Put another way, Ethereum hashing uses 60GWh every day, which costs around $6 million. That works out to $4.50 in power costs (using $0.10 per kWh) just to send Ethereum from one wallet to another.

Bitcoin is even worse. The network currently uses an estimated 78TWh per year, with a cost for a single transactio­n of around 690kWh. That’s about as much power as a typical US household uses in 24 days! That means Bitcoin uses nearly $8 billion per year just keeping the blockchain secure. That figure continues to grow, and while Bitcoin and Ethereum are the two largest cryptocurr­encies, there are dozens of others that together require another large chunk of power.

Looking at the costs and power going into these networks, it’s difficult to remain optimistic about the long-term potential. We’re strip-mining digital coins, and that’s unsustaina­ble. That’s part of the cyclic nature of Bitcoin, where it finally reaches a level that can’t be sustained, companies and people stop investing in it, and there’s a massive correction. A few years later, enough people have forgotten the last time this all happened that we see a repeat. There are other proposed alternativ­es to current reward schemes, such as proof-of-stake instead of proof-of-work, but they have their own drawbacks.

Imagine if all of the effort being put into cryptocurr­ency mining was instead put into research. While Bitcoin’s SHA256 hashing isn’t useful for many other tasks, the most powerful supercompu­ter in the world—Japan’s Fugaku—has a peak power consumptio­n of around 30MW. That’s a modest city’s worth of power, and yet the Bitcoin network uses roughly 3,000 times as much energy. What could scientists do with another 3,000 of the fastest supercompu­ters? We don’t know, but we’d love to find out. The money that’s pouring into cryptocurr­encies could be put to far better uses.

These drawbacks won’t stop people from mining, though, at least not without help. Government­s have started to crack down on mining farms, however—China’s Inner Mongolia recently issued a ban on cryptocurr­ency mining. Other areas of the world may follow suit. Regulation seems inevitable, and for something that’s supposed to free people from the oppression of the banking system, it’s difficult to get past the fact that every coin eventually derives its worth from how much it is worth in dollars, pounds, yen, and so on.

If you still want to give mining a shot, particular­ly if you already have a modest PC and you live in an area where power isn’t too expensive, or the power grid isn’t already having severe issues, we won’t object too much. But please don’t bet the farm or your retirement on PC hardware and upgrades solely used in pursuit of mining. There’s a real likelihood of chasing diminishin­g profits, potentiall­y for years. You have to keep on mining in the hope of striking it rich with a future surge in cryptocurr­ency prices. We doubt we’ve heard the last of Bitcoin or Ethereum mining, but there’s a good chance anyone trying to start now has already missed the current wave.

 ??  ??
 ??  ?? Fromleftto­right: Ethereum, launched in 2015; Bitcoin Cash, which split from Bitcoin in 2017; Litecoin, launched in 2011, making it an oldtimer; Dash, which is similar to Bitcoin, but uses a two-tier structure of miners and masternode­s; Bytecoin—not to be confused with Bitcoin.
Fromleftto­right: Ethereum, launched in 2015; Bitcoin Cash, which split from Bitcoin in 2017; Litecoin, launched in 2011, making it an oldtimer; Dash, which is similar to Bitcoin, but uses a two-tier structure of miners and masternode­s; Bytecoin—not to be confused with Bitcoin.
 ??  ??
 ??  ??
 ??  ??
 ??  ??
 ??  ?? These examples are probably a bit more glamorous than your average mining setup, but this is what it could look like.
These examples are probably a bit more glamorous than your average mining setup, but this is what it could look like.
 ??  ?? All that power has to come from somewhere, and electricit­y generation isn’t always the greenest of industries.
All that power has to come from somewhere, and electricit­y generation isn’t always the greenest of industries.

Newspapers in English

Newspapers from United States