The old Antminer S9's can definitely be profitable with custom firmware and time-of-use power pricing. It is possible on some power grids to get cheap energy (< 4¢) during the night or low demand periods, even if you can only maintain 60-90% uptime. But the obvious drawback is the higher demand charge and the stress on the equipment from heating/cooling cycles. Then there's the problem of getting enough electric capacity in a warehouse, vs. using the same amount of kW for more expensive miners.

Mixing/matching multiple power sources on the same device is usually a dangerous idea, at least in my experience with GPU mining. You can easily create currents that backfeed and burn out the cables.

You should buy a single power supply that is strong enough.

You probably need thermal glue, not thermal paste. Thermal paste is only useful when there's downward pressure forcing the heatsink against the die.

Correct. I calculated the absolute best case numbers possible. In the real world, you'll be lucky to get half of that 150w. Instead, you could also lower the heat production of the hashboards by 10-20% with custom firmware/undervolting at the same hashrate. That will conserve energy for both the hashboards and the fans!

There are definitely safer compounds that have a boiling point between 40° C - 80° C. Here's a [good place to start](https://en.wikipedia.org/wiki/List_of_refrigerants). It looks like [Novec](https://www.3m.co.uk/3M/en_GB/novec-uk/applications/fire-suppression/) in particular is very popular for datacenter fire suppression. If there's a leak, it is safe for workers, not flammable, and it leaves no residue. The only problem is it has 1/10th the thermal conductivity of water. Pentane has double the thermal efficiency, but it has a lower boiling point, so it'd only work in colder climates.

Another alternative is to run air through the miners as usual, but build a heat exchanger where the exhaust flows through a big radiator so the heat is transferred to the refrigerant. But this is more complex and adds cost, and efficiency will be lower because the temperature of the exhaust is always cooler than the actual silicon. However, I like this idea the most because it can be retrofitted to existing warehouses as long as they put a manifold behind the miners.

But like I said, it takes a six figure team of technicians to figure this stuff out, let alone bringing it to life, so even Bitmain or the largest farms probably don't want to venture into this idea (except for virtue signalling purposes, which might help because it would stop the greenies from pointing fingers at us miners).

Here's my (informed) opinion on this unique idea:

A replacement fan for an Antminer uses 12v/2.7A, which is 32.4 watts. Let's assume each miner has 4 fans, for a total of 130w of consumption. Assuming the motors are 100% efficient at converting electric energy --> wind energy (which they're not, because otherwise the fans would be silent, not 90 dB), you have 130w of extractable wind power. I'm also assuming the restriction of the airflow passing through the hashboard heatsinks takes out 0% of the wind energy. Finally, I will presume that the exhaust will pass through ducts, in order to bypass Betz's Law.

Assuming the average ASIC farm has a power cost of 5¢/kWh, and your turbine's motor is 100% efficient, the turbine would return $0.15/day. Let's say the turbine has a lifespan of 1000 days. The most you will get back is $150.

Is it worthwhile to build this contraption? If the rate of return on your contraption is less than if you just spent the $$ on more miners, you shouldn't built it. Also, now you have to worry about your wind power project instead of just the mining farm. Not to mention the slowed airflow through the miners, the heat stress and the higher failure rates of your miners. They were not designed with this reduced level of airflow. The more efficient your turbine is, the slower the air will pass through, and the hotter the devices will get.

Is it worth risking your equipment failing in order to save a few bucks on electricity? I think not.

---

Now let me talk about harvesting the heat energy. I actually think it is technically feasible. Of course, water won't work, since your hashboards would need to be 100°C to even generate steam (which will kill them).

But instead of water, you could use a hydrocarbon, like propane, with a lower boiling point. Let's say the hardware is 70° C while the outside air is -10 - 30° C. You can theoretically harvest that energy to spin a turbine, then cool the propane down and it flows back to the miner to get boiled again. There are power plants in the real world which use this process for certain applications. The efficiency is usually 30-40% of the heat energy, so for a 3400w miner, you can get 1200w back.

Now of course, the average person buying ASICs and setting up a farm can't do this. It takes hardware engineers, mechanical engineers and construction workers to design a miner that is propane-cooled. Then there's the problem of the flammability of propane, especially next to a heat source like the ASIC. If the pipe has a leak and the propane meets oxygen, the entire warehouse can go up in flames. The farm owner needs to build an entire piping system, not to mention fire safety regulations. It would take an entire team of engineers to manage the construction. Why not just spend that money (six-seven figures) on more harwdare instead?

So this idea probably won't work even for Bitmain or Innosilicon themselves. Why go through all this trouble just to save 30% on power?? Why hire 10 new engineers just to design a new ASIC that is propane cooled? It only makes sense if the miners are inefficient and power costs are higher, which is at the end of the hardware's lifespan. Of course, the vendors care about selling the miners cheaply, not how long they will last.

Hello, I'm wondering if there are any commercial ASIC mining farm operators located in the U.S., especially Texas, who have negotiated real time (hourly) or day ahead pricing with their utility. I'd also like to know if any farms have old ASICs like the Antminer S9 or Avalon 1146 that are usually unprofitable.

I'm thinking of writing software that takes in real time or next-day grid prices (preferrably ERCOT) and can remotely turn less efficient ASIC miners on/off depending on what the power cost is. I'm hoping that these old miners can be profitable if they run overnight, or when power is cheap enough.

I have done some preliminary calculations. For a medium-sized farm using at least 100 kW, it looks like you can get a 200-day ROI on a used $350 Antminer S9 as long as power costs < 4¢/kWh (uptime = 60-70%). Basically, it would be turned off during the afternoon and evening or during heat waves in the summer. This accounts for the higher demand charge for having a 60-70% load factor instead of running 24/7 and getting 99%. The demand charge averages 2¢ and the energy charge averages 2¢ during that 60-70% period of runtime.

You would probably have to plug in a small Raspberry Pi to the same LAN as the miners, and my backend would broadcast commands to perform the switching. Yes, the miners are less efficient, but heat is less of a concern because the weather tends to be cooler when electric prices are lower.

Would anybody with these older ASICs be interested in such a service instead of recycling them? If not, I'm thinking of starting my own warehouse with a few dozen used 16nm ASICs and seeing how it goes. 200 days seems really good for SHA-256 mining, and with ETH going to PoS, this might be a good investment.

I don't like to use mining calculators at all because they can't predict the future. If I buy equipment today, I care about what the profit will be in the next 1-2 years, not what Whattomine says the profit is today. Instead, I like to look at profit data from the past 5 years, like Bitinfocharts.

If anything, tax reduces the cost of equipment and electricity, at least in the U.S. (Section 179). The failure rate of hardware is no more than 4%/year. As for rent, leasing a small warehouse costs extra, but it pays off when you consider the cheaper power as long as you have at least 300 video cards.

With so many GPU coins switching profitability every few hours, it's become hard for me as a miner to keep track of the wallets on multiple exchanges as well as the mining pools. I have to log in and manually trade ETC, CFX and ERG into LTC, then withdraw the LTC. I'm worried about price fluctuations; I want to sell these coins at the best price.

I know that multipools like Prohashing or Nicehash already exist, but the problem is they don't support all algos and they charge high fees.

I'm thinking of developing a website that takes in withdrawals from various mining pools (ETH, ETC, CFX, RVN, ERG, FLUX, SERO, FIRO, etc.) and auto-trades them to one output coin, like ETC or LTC. The website then instantly deposits the single coin to your output wallet for a small fee (~0.3%).

Under the hood, all the wallets will be hosted on an exchange like Hotbit and trades will be performed on that exchange. Security shouldn't be a problem because all incoming deposits are traded and sent out in seconds.

With this service, you can always mine the most profitable coin (which will more than compensate for the 0.3% fee) and receive the earnings in the coin of your choice. You can also track the revenue history/prices of each algorithm over time. You can use any combinations of pools that you want, like choosing those with the smallest fee or the best reliability. When the market becomes unstable after ETH PoS, mining the best coin out of several choices will only become more important.

I want to know what the community thinks of this idea. Is it unnecessary or is it convenient?

This absolutely is a trap. A lot of n00b miners will build rigs right now and assume these levels of profit will continue. We already know what the ETH dev roadmap is: as of right now, they are on schedule for the PoS merge in June. It'll take a major bug to delay this timeline, and even then, the delay would probably be 4 months at most.

I 100% agree. Successful operators balance the amount of equipment, cash and coins that they own.

Right now, we should probably hold the equipment, sell the coins, and save cash. In 1-2 months, we should sell the equipment in order to build up cash. After the PoS aftermath is over, we should buy as much equipment as possible with cash, then (depending on coin prices) either hold the coins or slowly replenish our cash reserves.

I recommend the RTX 3060 12GB LHR even though it has a PCIe power connector. It costs $500 used on feeBay or $530 on Newegg. Each card can get 44 Mh total (18 Mh ETH + 26 Mh CFX) at 130w DC.

I agree, mining profit has gone up by 15-25% in the last 2 weeks. Video card prices have stabilized on feeBay and Newegg.

Same here! I cannot wait for PoS. I'm going to be ready to buy hundreds of video cards and start a warehouse again.

1: Purchasing mining motherboards, which were unreliable and one even caught on fire
2: Bricking a video card with a failed BIOS flash. I later un-bricked it though
3: Not mining from 2015-2016 and saving the coins
4: Buying a cheap Raidmax power supply, which exploded after mining DOGE with two R9 270's
5: Powering a Radeon RX 470 USB riser with Molex instead of PCIe. The card drew 60-70w from a single 18AWG wire, which burned the wire until it glowed red hot
6: Not monitoring VRAM temperatures. I allowed an XFX Radeon R9 290 to run at 132° C VRAM, which left black spots on the PCB. I fixed the problem by installing tiny heatsinks on the memory chips

A mining calculator isn't the best answer here. The profitability will change a lot in the coming months. The OP wants to know if GPU mining is a good idea in the long run.

To answer the question, 10¢/kWh is a decent price for power when it comes to GPU mining. That's better than average for U.S. residential prices. You'll do perfectly fine even during a coin recession or after ETH PoS. The only rate where GPU mining doesn't make sense in the long run is > 16¢/kWh.

However, 10 cents sucks for ASIC mining, so don't buy any ASIC machines except for low-power Goldshells. 5.5-6¢/kWh is average for ASIC farms.

If you're on Windows, I don't see why you want to get rid of MSI afterburner unless you have a farm with dozens of GPU rigs (why would you even use Windows in that case?).
If you're using Linux, I would switch to a standard mining OS like Minerstat, SMOS or HiveOS, which can handle many other things as well as overclocking. Minerstat is the best at OC.

Like others have said, the other coins will never compensate for the loss of block rewards due to ETH disappearing. ETH makes up 95% of the total revenue available to GPU miners. Profit on the other coins after PoS will collapse to barely breaking even above power cost. The get-rich-quick miners who were expecting a 400-day payback time will panic once their profit margin gets squeezed to 0. They will exit the market while they can still get decent prices for their hardware on feeBay.

But those kinds of projects can take years to become big enough to support millions of video cards. Developers are probably biased against using PoW in the first place, let alone ASIC-resistant PoW.

That's why my business plan is to wait for the big ETH crash, buy up cheap video cards, then mine whatever breadcrumbs are left on the other coins. If I buy the equipment cheaply, I don't care if it takes 2-3 years until the next big GPU coin comes along.

---

Good point. This is why the second half of this year will be the perfect storm for gamers as well as miners who want to get into the market at a cheap price. According to Moore's Law is Dead, Shintel will probably sell their video cards for below cost in order to aggressively take market share from AMDead / nGreedia.

[moderator's note: consecutive posts merged]

I don't know about hashrate fluctuations due to USB risers. I've never experienced this before across 300 different video cards.

But I can definitely tell you that you shouldn't buy a UPS (uninterruptible power supply) to deal with the power outages. It will be very expensive. Just make sure your rig can boot up within 2 minutes and get back to mining (which you say it does). That's the best you can do.

The RX 580 is from mid-2017. The only reason it's profitable is high ETH prices. This is probably one of the worst cards to own post-PoS. The efficiency on any non-ETH coin is awful.

The best video card models are probably going to be the RTX 3080 Ti, RTX 3060 and RX 6800 non-XT.