bitcoin home mining opportunity cost!

[philipma1957]

Gave you 20 merits you are now at 50. Very detailed info.

And it can be more complex.  I am sitting at a desk typing and looking at 35 books of us tax code while I write this.

I can only say the laws are very complex but if you understand the Model Amazon did to build its business up a USA miner can do the same.

Our Clifton operation started in Dec 2018 with 4 s9's

We now have

2ph in BTC asics
10gh in Gpus
30gh in LTC asics.

No debt. And very little tax paid. All legal.

[1714944558]

1714944558

[1714944558]

1714944558

[philipma1957]

Here in the U.S., a farm owner can deduct 100% of the equipment cost in the first year. This is a huge tax advantage worth up to 37%. The owner can keep re-investing all profit back into more miners.

Ok that makes sense, but you can't have an infinite loop, you will have to stop at one point, or else, everyone in the U.S would be mining bitcoin and you all would have been rich.

Also, a few key points are missing here, the most important being the lifespan of the miners you invest in, and that could be the physical status of the miner (dead before ROI) or it gets to the point where the power bill is a lot greater than it's projected income.

Also if I understand correctly, selling your mining gears later on for whatever reason is a taxable event, no? as well as selling your bitcoin to pay the power bill or to buy a new transformer or anything like that, even your received reward from the pool are taxable as income.

Phill's example above is a good one but there are a few things I don't get

1.9ph  in btc gear and 40gh in ltc/doge gear and 8gh in eth gear  would be 100% non taxable business expansion.

So sitting here that gear is worth 386k

So just to be clear here, and let's assume your business model isn't 50/50, and that you had to pay $$ every month for the power bills, that means, you would have to sell some coins (a taxable event) no?

The second question, your example also assumes that for all these years you did not spend a single dime and re-invested all of it in buying more miners, true?

furthermore, you now (according to the example) own 386k worth of mining gears, what happens when you sell them? you will pay a lump sum of tax?

and lastly, if you decided to stop expanding (which you will get to at one point), all the profit you make will be taxable even if you don't sell your bitcoin, and then if you sell it later on for higher a price that will be a capital gains tax, true?

Not that I could use any of these U.S tax tips, but it's an interesting topic, I would love for someone to give me a perfectly detailed example of all the tax involved with numbers.


Let's say today, You buy S19 for $10,000, for the next 12 months, it makes you $500 worth of coins monthly, every month you sell 100$ to pay the power bill, and use $100 to improve your mining farm to be able to add more capacity and $300 reinvested in mining gears, how much tax do you pay in total for the total mining profit you made per month?

Miner life span is critical this is why I have a ton of gpu's they last 3-4 years. They have warranties that get honored.

And the eth mined goes to cash/gear

the btc mined can be hodled if I chose. Yes the btc will be partly taxed. hodl or sell.

 Ie a paid for s17 makes 0.0002155 btc a day. my structure for power and hosting is 50% of that so I get 0.0010775 btc a day for that machine that is $4.31 usd a day if I choose to hodled it or sell it. . I will owe tax on the $4.31 so say $1500 in a year would be taxed. at regular income. maybe 450 tax so it clears $ 1050 .  Note the assumption is made on flat daily numbers.

so Every s17 I own nets me 1050 a year after tax. My 17 S17's net just about 17850 a year. after tax if I sell it daily
My other btc gear is a bit more so it nets me around 22000 a year after tax if I sell it daily, so btc alone is 39850 after tax net.

all the eth
all the ltc
all the doge

get rolled into new gear. We are heat and cooling capped right now. Not power capped. I can do 45kwatts  more in gpus I do 25kwatts now. Have to figure cooling out. we have limits to fans in the gpu room. we may add a chiller or 2 they would solve overheating in the summer.

[Coinfarm ventures]

Gave you 20 merits you are now at 50. Very detailed info.

I can only say the laws are very complex but if you understand the Model Amazon did to build its business up a USA miner can do the same.

Our Clifton operation started in Dec 2018 with 4 s9's

We now have

2ph in BTC asics
10gh in Gpus
30gh in LTC asics.

No debt. And very little tax paid. All legal.

That's a lot of merit!

I have spreadsheets that are more detailed than that, but they get really complicated.

Wow, it's great that you bought in at the perfect time, after the great downturn of 2018. It's excellent that you didn't have to beg outside investors for money.

I did exactly the opposite, which is why my ETH farm went bankrupt. It deserved to go bankrupt, especially because I wasted $$ on 4K monitors, gigabit internet and bought cards at the top of the market. On top of that, most of my investors were the 'get rich quick' type who bailed when the market crashed, and weren't interested in coming back even when GPUs were dirt cheap. I was a dumb college kid at the time who was desperate with FOMO to not let my dream of owning a mining warehouse go while investors were still interested. The only good decision I made was getting a 6 month lease.

Thank god I spent a few days digging through this forum and learning from the veterans. I think I've come to the conclusion that I should not build an ASIC farm until/if BTC plummets to $25k while I already have a GPU farm running.

I should probably spend the next 6-8 months keeping track of the mining market, saving up $, while trying other business ideas first (like software). By then, I will have enough funds to buy at least 250 of my own GPUs if PoS happens. If PoS doesn't happen but BTC is $25k, at least I can buy ASICs cheaper than today's prices. If my software ideas succeeds, then there's no point in building a farm in the first place.

[mikeywith]

I disagree because it looks like manufacturers these days want to make their equipment more reliable, because they know that the die shrink to 3/4nm will take a long time and customers will want the miners to last longer.

you are right about the quality recent miners of Bitmain (only Bitmain) has improved compared to the previous 17 series, but other companies like Avalon are now making worse gears, I did not pull the 20% out of no where, people have reported 80% failure rate on new Avalon gears, 25% on S19's PSUs, 10% on MicroBT 20-30 series, it's nothing but wishful thinking if you run the numbers with the assumption that failure rate would be below 10%.

The reason why I said 20% was to account for

1-The downtime
2-Repair Cost
3-Shippment fees

If you want to use 10% or even 8% then you need to take those factors into account as well, you need to consider that waiting for 6 months to get your miners back isn't something unusual, or sending a whole miner that has only 1 dead hash board and then having to wait for a month or two to get it back, many things can go wrong and using anything below 10%+ the three points above is honestly being unrealistic. 

In this spreadsheet, the coins are always sold instantly in order to get cash. The risk of losing coins from a hack is practically zero; you can only lose a day of profit. As for the network, the firewall/switch should be configured so none of the machines can talk to each other. Only a few outbound IP addresses, such as the pools, should be whitelisted. This isn't hard to set up.

Not sure why you think that keeping a hardware wallet safe without anyone knowing you have it is harder than maintaining a mining farm that's worth a few million and sucking thousands of amps of the grid.

Most people who have enough money to start a mining farm at this scale are people who earn $250k+/year (engineers, doctors, lawyers, executives) and have saved up that 1 million over the past 10-20 years or so. Their marginal tax rate is about 32%. They typically have other investments with unrealized tax gains like stock shares or real estate. Why not take this opportunity to sell those assets and deduct the $1m in miners against that profit? $250k of gains from their regular job, $350k from their vacation house, then $400k from Tesla stock they bought long ago.

This is a valid point but it's a bit of cherry-picking IMO, so this model of "using tax to your advantage" isn't fit for everyone, if I don't have unrealized tax gains that are worth 32% of whatever I am willing to spend on mining -- it's no longer 32% is it? it could be 5% or even 1%? please elaborate more on this subject without assuming that everyone who enters mining has some sort of massive unrealized tax gains.

Gave you 20 merits you are now at 50. Very detailed info.

He deserves them, a very detailed explanation of how the tax stuff works in the U.S.

Miner life span is critical this is why I have a ton of gpu's they last 3-4 years. They have warranties that get honored.

the miner's lifespan is more important than the power rate, I have free power but I have hundreds of dead hash boards  S9ks, S9, S9se, D3s, S17 pro, S17,S17 +, T17, L3s and you name the rest, recently I fired up 40 Avalon 841 miners which run like shit, glad I didn't buy more.

[Coinfarm ventures]

The reason why I said 20% was to account for

Do you mean 20% per year or 20% over the useful lifespan of the miner? If it's 20% per year, that really sucks in comparison to GPUs. I have had a farm with 200 video cards where only 2 of them failed in a 9-month period. I sent these cards to MSI warranty repair and they returned working ones within 1-2 weeks. This is why I have MSI brand loyalty.

If you want to use 10% or even 8% then you need to take those factors into account as well, you need to consider that waiting for 6 months to get your miners back isn't something unusual, or sending a whole miner that has only 1 dead hash board and then having to wait for a month or two to get it back, many things can go wrong and using anything below 10%+ the three points above is honestly being unrealistic. 

Lesson learned for me, then. If I want to buy used MicroBT M20S's or Avalon 1146's, I should price in a 20% failure rate for the unrepairable problems + the money/downtime cost of fixing the fixable ones.

Do you know if failure rates can be lowered by downclocking or running at lower power usage?

This is a valid point but it's a bit of cherry-picking IMO, so this model of "using tax to your advantage" isn't fit for everyone, if I don't have unrealized tax gains that are worth 32% of whatever I am willing to spend on mining -- it's no longer 32% is it? it could be 5% or even 1%? please elaborate more on this subject without assuming that everyone who enters mining has some sort of massive unrealized tax gains.

I do have to admit I cherry-picked the ideal scenario. If one doesn't have unrealized capital gain equal to $1m, they should depreciate the miners over 2-3 years instead of in 1 year with Section 179.

the miner's lifespan is more important than the power rate, I have free power but I have hundreds of dead hash boards  S9ks, S9, S9se, D3s, S17 pro, S17,S17 +, T17, L3s and you name the rest, recently I fired up 40 Avalon 841 miners which run like shit, glad I didn't buy more.

The impression I had at first when looking at mining during recessions was power & rent rates were important in order to be able to pay the bills while profit was low. But the more research I do, the more I feel like the residual value of the equipment is more important.

Mining farm A: pays 8 cent power, buys a M30S for $6000, then it's worth $5000 by the end of the year (including repairs). Profit = $2600

Mining farm B: pays 4 cent power, buys a M30S for $8000, then it's worth $5000 by the end of the year (including repairs). Profit = $1730

It looks like buying reliable equipment for a reasonable price is the #1 most important factor, as long as you can keep the doors open if BTC price were to fall by half.

Here's a new spreadsheet with more realistic assumptions. I assumed BTC price will slowly rise by 60% over the period, diff will rise by 25%, and the resale value of each working miner will stay the same at the end.



Unfortunately, HODLing still wins by a hair even when price growth is zero.



I tried many more price increases/declines, and in every single case, HODLing won by a small margin over mining.


However, as a bright spot, the GPU spreadsheets are a hell of a lot better. Mining almost always beats HODLing Ethereum. So long as you can sell the cards before PoS, the very high reliability & profit margin at even 7 cent power trounces ASIC mining at free power.

The best part in my view is, since GPU mining is light on power, a normal U.S. house with a 200 amp panel can hold 200 RTX 3060 cards. No need to rent a separate warehouse if power is already 9 cents.

[mikeywith]

Do you mean 20% per year or 20% over the useful lifespan of the miner? If it's 20% per year, that really sucks in comparison to GPUs. I have had a farm with 200 video cards where only 2 of them failed in a 9-month period. I sent these cards to MSI warranty repair and they returned working ones within 1-2 weeks. This is why I have MSI brand loyalty.

Per year, keep in mind that the 20% I used accounts for many other things which will be hard to predict, the global average failure rate IMO is about 10% and double that for fixing your gears, for example, one of the members here offers one working 17 series board in exchange for 2 dead boards, so if you get say a total of 20% failure it means the final outcome would be 10%, but then there is a downtime, shipping cost and a few other things involved.

But keep in mind, not everyone will have that 10-20% failure rate, some people reported a 3-5% yearly failure rate, others reported up to 90%, if you get a bad batch that makes up 50% of your purchase, you will be in deep trouble, so the failure rate is really near impossible to predict, but based on my experience and what I hear from many other miners, I believe 20% should be used as a general rule of thumb.

Keep in mind that the size of your farm is also very important because even when the global failure rate is 2-3% x model, it could be 50% for you and of course vice versa, I think a few folks around here got lucky with the 17 series, I personally had close to 100% failure rate in the first few months, a large Chinese miner (at that time) reported 40-60% failure rate in the first 6 months (Bitmain's CEO himself confirmed that high failure rate).

If you are a large player who places orders in the tens of millions, that might not be a huge issue, I remember how Bitmain offered to send their technicians to those large players in China to have their 17 series fixed on the spot, and yet they screwed the majority of other clients by forcing them to send a whole miner to get 1 board fixed, and then some people had to wait forever to get the gears back, some got back broken miners, so it's different for everyone, you might be lucky, but when you are running the numbers for a business you should not count on luck and -- instead, you should use the worst-case scenario numbers.


GPUs are a different story, those companies have been building them for decades, those companies know they will be making those GPUs forever, they also operate in countries where the law protects the client to some degree, the warranty gets honored despite the small failure rate, on the other hand, with ASIC miners, you are dealing with Chinese companies that have poor quality control, only a couple of years of experience, demand to supply ratio is unbelievable, and of course, they know their business won't last forever.

Another factor is that ASIC makers are under severe pressure, they rush to make miners and sell them to the public before doing enough testing, a great example is the entire 17 series, 90% of failures were caused by bad solder/soldering, Issues like that appear in the first month or two of running, which means nearly zero testing was done on them.

Now you have the S19 series, the PSUs suck and have a high failure rate, another proof that it didn't undergo enough testing.

A bulk Chinese reseller whom I trust confirmed that the new Avalons PSUs have up to 80% failure rate, another proof that Canaan did close to nothing of testing.

MicroBT put out a new firmware that can't be reverted, and guess what? once you flash it the old PSU is no longer compatible, and thus you will have to either order a new PSU or a replacement hash board, which is out of stock most of the time, not sure if they did it intentionally or they were just careless, but the outcome remains the same for the poor folks who had to go through that trouble, and it gives you a good idea of how bad these companies manage their products.

I can go on forever, but I think I need not, it's plain simple, GPUs are more reliable than ASICs by an order of magnitude, nobody in their right mind should assume that the quality of both is similar in any way, shape or form.

[philipma1957]

The reason why I said 20% was to account for

Do you mean 20% per year or 20% over the useful lifespan of the miner? If it's 20% per year, that really sucks in comparison to GPUs. I have had a farm with 200 video cards where only 2 of them failed in a 9-month period. I sent these cards to MSI warranty repair and they returned working ones within 1-2 weeks. This is why I have MSI brand loyalty.

If you want to use 10% or even 8% then you need to take those factors into account as well, you need to consider that waiting for 6 months to get your miners back isn't something unusual, or sending a whole miner that has only 1 dead hash board and then having to wait for a month or two to get it back, many things can go wrong and using anything below 10%+ the three points above is honestly being unrealistic. 

Lesson learned for me, then. If I want to buy used MicroBT M20S's or Avalon 1146's, I should price in a 20% failure rate for the unrepairable problems + the money/downtime cost of fixing the fixable ones.

Do you know if failure rates can be lowered by downclocking or running at lower power usage?

This is a valid point but it's a bit of cherry-picking IMO, so this model of "using tax to your advantage" isn't fit for everyone, if I don't have unrealized tax gains that are worth 32% of whatever I am willing to spend on mining -- it's no longer 32% is it? it could be 5% or even 1%? please elaborate more on this subject without assuming that everyone who enters mining has some sort of massive unrealized tax gains.

I do have to admit I cherry-picked the ideal scenario. If one doesn't have unrealized capital gain equal to $1m, they should depreciate the miners over 2-3 years instead of in 1 year with Section 179.

the miner's lifespan is more important than the power rate, I have free power but I have hundreds of dead hash boards  S9ks, S9, S9se, D3s, S17 pro, S17,S17 +, T17, L3s and you name the rest, recently I fired up 40 Avalon 841 miners which run like shit, glad I didn't buy more.

The impression I had at first when looking at mining during recessions was power & rent rates were important in order to be able to pay the bills while profit was low. But the more research I do, the more I feel like the residual value of the equipment is more important.

Mining farm A: pays 8 cent power, buys a M30S for $6000, then it's worth $5000 by the end of the year (including repairs). Profit = $2600

Mining farm B: pays 4 cent power, buys a M30S for $8000, then it's worth $5000 by the end of the year (including repairs). Profit = $1730

It looks like buying reliable equipment for a reasonable price is the #1 most important factor, as long as you can keep the doors open if BTC price were to fall by half.

Here's a new spreadsheet with more realistic assumptions. I assumed BTC price will slowly rise by 60% over the period, diff will rise by 25%, and the resale value of each working miner will stay the same at the end.



Unfortunately, HODLing still wins by a hair even when price growth is zero.



I tried many more price increases/declines, and in every single case, HODLing won by a small margin over mining.


However, as a bright spot, the GPU spreadsheets are a hell of a lot better. Mining almost always beats HODLing Ethereum. So long as you can sell the cards before PoS, the very high reliability & profit margin at even 7 cent power trounces ASIC mining at free power.

The best part in my view is, since GPU mining is light on power, a normal U.S. house with a 200 amp panel can hold 200 RTX 3060 cards. No need to rent a separate warehouse if power is already 9 cents.

Its complex to full understand asic mining.

But asic builders are very much "fuck you blow me bitch" Sorry for being crude but it is very often the case.

Does running a bitmain  asic at a lower clock work. Yes to a large degree it does.

Especially for the psu's.

remember s15 s17 s19 and the variants all have bespoke custom psu's.

Running an s17 pro on turbo can drop the one chip heatsink that tends to fall off with too much heat. It can also blow the psu.

A dead hashboard = 33% hash drop
a dead psu = 100% hash drop.

so running at a lower speed can help
having extra psu's = bigly. A very important item to stock.

I killed 2 s17 psu's
I killed 1 s15 psu's

I had extra ones so down time = zero

[Coinfarm ventures]

I can go on forever, but I think I need not, it's plain simple, GPUs are more reliable than ASICs by an order of magnitude, nobody in their right mind should assume that the quality of both is similar in any way, shape or form.

Wow, that's a lot of good information. I think you made a good point about bad batches of ASICs, where one batch can have a 50% failure rate but other batches have hardly any issues. I would guess a way to mitigate this luck effect is to buy small amounts of miners at a time from different places (ex. feeBay, Reddit Minerswap, or this forum). But they'll probably be more expensive in smaller quantities.

I do have a question about the failure curve. Is it a 'bathtub curve', like with hard drives, where lots of miners die in the first 4-6 months, but after a year, the survivors last for years before they eventually die of old age? Or is the curve more linear, where a small number die consistently every month? The reason I'm asking is that if I bought T19 units that have been fully working for 1 year, are they going to have a lower failure rate than if I bought brand new T19's?

Yeah, GPUs can be extremely reliable after even 10 years of mining especially with the low-power algorithms after 2016. Even during the high power Bitcoin/Litecoin/Doge days, the failure rate was < 5%. Warranties are honored 90% of the time for free, and repair service is quick. The other rig components are almost as durable, as long as you don't buy some cheap knock-off mining MOBO from Chyyna. I prefer 'gaming' MOBOs because they have better resale value even if I can only attach 8-10 GPUs instead of 12.

The power supplies for GPU rigs, whether they're gaming ATX supplies or server ones, are built to last long as long as it's a mainstream brand. They have all kinds of safety protections. Efficiency is usually as good as the ASIC PSUs.

[mikeywith]

Does running a bitmain  asic at a lower clock work. Yes to a large degree it does.

It sure does, however, there is a downside to it which is the cost per hash,  an S9 will most likely last a lot longer running underclocked at 9.5th instead of the default 13 to 14.5th, but then assuming the miner costs $400 shipped, it becomes 28.5$ per th vs 42$ per th, you do save on power but the most often the lost hash rate surpasses all that.

I can go on forever, but I think I need not, it's plain simple, GPUs are more reliable than ASICs by an order of magnitude, nobody in their right mind should assume that the quality of both is similar in any way, shape or form.

I would guess a way to mitigate this luck effect is to buy small amounts of miners at a time from different places (ex. feeBay, Reddit Minerswap, or this forum). But they'll probably be more expensive in smaller quantities.

Buying different batches might be a good idea, but not in the way you mentioned, batches come out in quantities of thousands, usually monthly, so buying from different people does not mean you will get a different batch, they could all have bought that one bad batch you were trying to avoid.

Usually, the first a few batches are the worst, when a few months have passed and the manufacturer starts getting a ton of warranty repair tickets,  they "might" fix what seems to be the issue, but then missing the first batch might also mean paying a lot more for the same miner, it's pretty hard to time this, luck is a major factor.

I do have a question about the failure curve. Is it a 'bathtub curve', like with hard drives, where lots of miners die in the first 4-6 months, but after a year, the survivors last for years before they eventually die of old age? Or is the curve more linear, where a small number die consistently every month?

It's both, for an example antmine S9 are known to have a dead middle board due to a lack of airflow around the first a few chips facing the front side caused by the fan motor itself, so the longer it runs the likely to fail, so this gives you a linear curve more or less.

With something like 17 series where very often the issue is just one bad solder ball shorting the shit out of the board, it could be safer to buy a used miner which has been repaired, if your 17 series miner makes it through the first 6 months, it's more likely to survive another 6 months, but then again, different batches will have different results, the ones that later came from Malaysia were a lot better, it seems like the robots/people that did the soldering were doing it in a better way.

I also remember buying an early batch of S9k where I got miners without heatsinks attached, the heatsinks were sitting in the case doing nothing, it seems like they forgot to heat the thermal , on the other hand, I recall Phill got a different batch that arrived in good shape and worked for a while (not sure about the current status).

So ya, it's just a mess, very hard to predict and it's part of the risk you have to take when you are looking at a business than can get you a 1-2 year ROI, I highly doubt that even drug dealers have such a short projected ROI, , this is why when you invest in mining, you can win big just in the same way you can lose big, it's a very risky business, especially for those who are just starting, I always tell the majority of people NOT to invest in mining, at least not before knowing the full story, because things look pretty damn good on paper, you go to an online calculator and think to yourself that you will get your money back in 6 months, reinvest and double your hashrate in 3 months, get that many PH worths of hashrate and start making 10k a month while doing nothing, it's exactly how things look like in an excel sheet like the one you made, but that couldn't be further from the truth.

[Coinfarm ventures]

Buying different batches might be a good idea, but not in the way you mentioned, batches come out in quantities of thousands, usually monthly, so buying from different people does not mean you will get a different batch, they could all have bought that one bad batch you were trying to avoid.

Usually, the first a few batches are the worst, when a few months have passed and the manufacturer starts getting a ton of warranty repair tickets,  they "might" fix what seems to be the issue, but then missing the first batch might also mean paying a lot more for the same miner, it's pretty hard to time this, luck is a major factor.

In that case I guess a decent strategy is to never buy a miner that has just been released onto the market. Always buy a model that has been in production for at least 1 year, whether it's new or used. Avoid the new Whatsminer M50S, Bitmain S19 XP or

I also remember buying an early batch of S9k where I got miners without heatsinks attached, the heatsinks were sitting in the case doing nothing, it seems like they forgot to heat the thermal , on the other hand, I recall Phill got a different batch that arrived in good shape and worked for a while (not sure about the current status).

Imagine what would happen if a manufacturer like ASUS or EVGA or Sapphire had a heatsink or solder ball issue on even 5-10% of the units. There would be mass outrage among gamers and they would go bankrupt.

Looks like ASIC companies can get away with murder for whatever reason, while GPU companies build stuff that is meant to last...

So ya, it's just a mess, very hard to predict and it's part of the risk you have to take when you are looking at a business than can get you a 1-2 year ROI, I highly doubt that even drug dealers have such a short projected ROI, , this is why when you invest in mining, you can win big just in the same way you can lose big, it's a very risky business, especially for those who are just starting, I always tell the majority of people NOT to invest in mining, at least not before knowing the full story, because things look pretty damn good on paper, you go to an online calculator and think to yourself that you will get your money back in 6 months, reinvest and double your hashrate in 3 months, get that many PH worths of hashrate and start making 10k a month while doing nothing, it's exactly how things look like in an excel sheet like the one you made, but that couldn't be further from the truth.

The funny part is that I used to know a 'former drug dealer' who got into mining during the 2017-2018 gold rush, buying dozens of video cards because he thought ETH profit was so great. Then he quit after ETH crashed, then probably went back to selling drugs for a while. So you're probably not exaggerating.

The mining 'calculators' are complete garbage because the market constantly reacts to difficulty & price changes. No spreadsheet I can make is much better than that, because the best I can do is account for different costs and BTC price levels. Such spreadsheets are useless for the #1 most important factor for profit; which is the price and reliability of the equipment, which is the hardest one to predict.

I used to think the electricity rate was more important, until I did the math. I realized a 10 cent farm can easily beat a 3 cent farm if their timing is better by just 3-4 months. That's exactly what happened to me in the past. The rigs I built in late 2017 for $200/card did well, the rigs I built in early 2018 at $370/card were total losses, and the rig I built in early 2019 for $140/card returned 4x my money back. Of course, the 'all-knowing' mining Whattomine said early 2018 was the most profitable while early 2019 was a bad time to invest. That's when I figured out mining calculators were garbage.

After all this research, I have decided to stick to GPU mining for now. If I build a warehouse, it will be mostly for GPU rigs. Now I don't even need to move to Texas since saving 1-2 cents on power isn't important for GPU mining. I think it's a good time to build GPU rigs since video card prices are relatively low and PoS has been delayed (to Q3 at the earliest).

If you think many folks entering ASIC mining who just look at the calculator sites are useless... you don't want to know how stupid some of the GPU miners can be. The #1 biggest mistake is mining ETH on every single model of video card all the time, yet 90% of miners do that. These people don't harness the true power of GPUs, which is that they are Swiss army knives that can mine on 100 different algo's and switch in a few seconds.

IMO, the stupidest investment these days is to buy Ethereum ASICs. Why pay the same price per MH as GPUs when you have to deal with 20% failure rates instead of 1-2%, then take a huge risk with resale value? If something happens to ETH, you're screwed, since you can't even sell it to a gamer; let alone mine another algo. "Bbbuuuttt the Antminer E9 is 3x as efficient as an AMD RX 6800!" Power cost hardly matters for GPU mining, dummy! Yet I see so many used Antminer E3's and Innosilicon A10's being sold on feeBay. Those suckers will never get their $ back even if PoS takes 12 months and ETH is $5000.

[philipma1957]

Does running a bitmain  asic at a lower clock work. Yes to a large degree it does.

It sure does, however, there is a downside to it which is the cost per hash,  an S9 will most likely last a lot longer running underclocked at 9.5th instead of the default 13 to 14.5th, but then assuming the miner costs $400 shipped, it becomes 28.5$ per th vs 42$ per th, you do save on power but the most often the lost hash rate surpasses all that.

I can go on forever, but I think I need not, it's plain simple, GPUs are more reliable than ASICs by an order of magnitude, nobody in their right mind should assume that the quality of both is similar in any way, shape or form.

I would guess a way to mitigate this luck effect is to buy small amounts of miners at a time from different places (ex. feeBay, Reddit Minerswap, or this forum). But they'll probably be more expensive in smaller quantities.

Buying different batches might be a good idea, but not in the way you mentioned, batches come out in quantities of thousands, usually monthly, so buying from different people does not mean you will get a different batch, they could all have bought that one bad batch you were trying to avoid.

Usually, the first a few batches are the worst, when a few months have passed and the manufacturer starts getting a ton of warranty repair tickets,  they "might" fix what seems to be the issue, but then missing the first batch might also mean paying a lot more for the same miner, it's pretty hard to time this, luck is a major factor.

I do have a question about the failure curve. Is it a 'bathtub curve', like with hard drives, where lots of miners die in the first 4-6 months, but after a year, the survivors last for years before they eventually die of old age? Or is the curve more linear, where a small number die consistently every month?

It's both, for an example antmine S9 are known to have a dead middle board due to a lack of airflow around the first a few chips facing the front side caused by the fan motor itself, so the longer it runs the likely to fail, so this gives you a linear curve more or less.

With something like 17 series where very often the issue is just one bad solder ball shorting the shit out of the board, it could be safer to buy a used miner which has been repaired, if your 17 series miner makes it through the first 6 months, it's more likely to survive another 6 months, but then again, different batches will have different results, the ones that later came from Malaysia were a lot better, it seems like the robots/people that did the soldering were doing it in a better way.

I also remember buying an early batch of S9k where I got miners without heatsinks attached, the heatsinks were sitting in the case doing nothing, it seems like they forgot to heat the thermal , on the other hand, I recall Phill got a different batch that arrived in good shape and worked for a while (not sure about the current status).

So ya, it's just a mess, very hard to predict and it's part of the risk you have to take when you are looking at a business than can get you a 1-2 year ROI, I highly doubt that even drug dealers have such a short projected ROI, , this is why when you invest in mining, you can win big just in the same way you can lose big, it's a very risky business, especially for those who are just starting, I always tell the majority of people NOT to invest in mining, at least not before knowing the full story, because things look pretty damn good on paper, you go to an online calculator and think to yourself that you will get your money back in 6 months, reinvest and double your hashrate in 3 months, get that many PH worths of hashrate and start making 10k a month while doing nothing, it's exactly how things look like in an excel sheet like the one you made, but that couldn't be further from the truth.

I grabbed 4 s9k. they still all work. but two of the four would over hear. simple solution was detach the center board. they are working as i type. since sept 2018.

I dca gear. ie 4000-7000 a month in buys every month.

it usually means 1 or 2 pieces a month.

depending if it is asic or gpu it could be 1 or 2 or 3 pieces a month.