Yes, 85° is a good memory temperature for GDDR6X cards. 70° or lower is a good core temperature.

Correct. Mining is a chicken-and-egg problem. You need money to make money.
If you want to become rich off of nothing, I suggest becoming a software developer and helping mining farms optimize their operations. You could profit off the 'suckers' entering the market. Of course, the failure rate is 80-90% for software companies, so be prepared to spend up to 10 years throwing ideas at the wall to see what sticks. When one does succeed, the payoff will more than compensate for the 10 or so failures.

I don't really understand why you bought solar panels... usually, the payback period on the actual ASICs is much faster than on solar arrays. What's the grid power cost in your area?

Personally, I think GPU mining will be alive until the end of the year. ETH PoS has been delayed to Q3 at the earliest. GPU returns are much higher than ASICs.

I was thinking of getting into ASIC mining, too, until I realized I would have to rent a small warehouse. Even then, the 100-150 kW of power capacity would only be enough to run 30-40 T19's. It would take 4-6 months to have the utility upgrade the power supply to 500 kW, and that could cost tens of thousands of dollars.

So now I'm considering trying to start it as a GPU warehouse, then converting to ASIC right before PoS. I have much more experience in GPU mining after all.

You're probably better off mining Monero or another CPU coin, but payback will take 5-10 years.

You are absolutely correct that the reported 'core temperature' isn't the whole picture. The VRAM can get much hotter than that. The worst I've seen is my XFX R9 290 where the core was 80° but the VRAM was 132°. This is especially true on the RTX 3000 GDDR6X series. But thankfully, VRAM temp can now be reported in Linux.

On average for my RTX 3080's/3080 Ti, the memory was 30-40° hotter than the core. The card will throttle if temp goes > 110°, but ideally, you should stay under 100°.

Other than that, I've used hundreds of video cards. As long as they don't use GDDR6X, all you need to worry about is the core temperature. Stay under 70 degrees, and if you can't, reduce power usage so you stay < 80. It's always better to increase fan speed; fans are cheap to replace.

It's usually worth paying for these accounting platforms, such as Bitwave. I would write a few lines of code to integrate Bitwave with the CRM/ticketing system, Minerstat, mining pool sites, the exchange, and the wallet API for payouts. Not to mention the ERCOT API with Hiveon / BraiinsOS / nvidia-smi for power management.

85 degrees is too high. It will reduce the lifespan of your video cards by 1-2 years.

I recommend staying below 70 degrees. There will be no benefit below that. The most important factor is that you reduce heating/cooling cycles, which destroys the silicon the fastest.

I did look at ASICminervalue.com. It says profit would be $7.25/day at 6 cents, which would be a 469 day payback time (static diff/price).

Good point, I would be mining in a hot climate. Even though the M20S is rated for 40° C, the heat may cause them to break.

Well, in that case I would rather not start a warehouse. I would just use colocation for 9-10 cents or even host them in the garage for 8-9 cents.

I'm considering buying used MicroBT Whatsminer M20S ASICs for a new farm. The electricity price is 6.5 cents.

Is this a good choice if I can buy them for $3400-3500 each? I've heard that Whatsminer is generally a reliable brand.

The other models I have looked at are the Antminer T19 (new) and the Avalon 1126 Pro (used). I don't like the S19 series because they are too expensive per TH, and my power cost will likely go down to 5.5 cents or lower after the summer.

I don't want to buy anything less efficient than 55W/TH because they are higher risk, and I'll probably have a limited amount of power capacity available for the first 5-6 months.

I believe BTC price can fall to as low as $25k this year, but it will most likely stay flat at $40-50k. Any recommendations?

Yes, it looks like I can find them, at least in the U.S.

Breakout board (fits 2 PSUs): $13 each

PCIe power cables: $19 for 10

IBM 94Y8118 900W 80+ Platinum supply: $20 each. Make sure you specifically buy the '94Y8118' version manufactured by Emerson, because it has the highest efficiency.

It looks like you can get 1800w of power for $72 in total at Platinum, which isn't a bad deal at just $40/kW. But I would only run 1500w through the breakout board, just to be on the safe side.

If you went with the 80+ Gold HP DPS-750RB, that would cost $47 for 750w of power. That's $63/kW.

So IBM is clearly a better deal, especially after you consider the $8-9/kW of savings every year from the increased efficiency.

I don't trust these 'new technology' press release articles. It usually takes years to refine the technology for mass production. Just because some laboratory at some university can produce 1 solar cell doesn't mean it is even close to being ready for the mainstream market, let alone mining farms. What is important today are trapped power sources like hydroelectricity, flared oil wells or imbalanced energy markets.

Oh, I didn't know Texas is one of the states that taxes cloud hosting.

But I don't think this applies to your situation, since the guidance states that "Settling of electronic payment transactions by certain entities" are not taxable. I couldn't find anything specifically declaring that crypto mining services are taxable, so you could default to not levying tax as long as your accountant agrees.

There's probably a way you could make it work, such as taking wallet addresses from customers and having the pool mine directly to their wallet.

I don't think so, since it would be an S-corporation or LLC with just one shareholder, and the shares are not being publicly sold. The sub-corp wouldn't even have to issue more shares for more investment. The customer could simply write a check to the sub-corp, which I'll then use to buy more equipment. No stock or securities transactions involved here since they already own 100%.

But like I said, I'll talk to an attorney. I don't care what the structure is, as long as the customer can take the depreciation (since I have no use for it) and the parent corp is the exclusive hosting provider.

It would only work if the rate of return on the solar panels is higher than on just buying ASICs and running them on the grid at 5¢/kWh. I don't think that's true, even close to the equator.

Solar farms with current technology take 5 - 10 years to pay for themselves, compared to 1-1.5 years for an ASIC miner. Even though that's a decent return, why not just buy more miners instead?

The new solar cell technology will need to be at least 5x cheaper per watt in order to be viable for mining. It still has to compete with hydropower, which is currently the cheapest, most reliable source of all.

I don't think services are taxable in most U.S. states. You're probably providing a service, not a good. If not, then try structuring your business so it's obviously a service.

Think of it like a cloud hosting provider. I've never paid sales tax on VPS hosting even from providers based in the U.S., even though both parties live in states with a sales tax.

Worst case, tell your customers to provide reseller certificates so that they pay the tax 12-15 months later instead of today. Hopefully they'll make a return on the extra 5-10% of hashrate they can buy, you get 5-10% more revenue, and the sales tax owed is diluted 10% by inflation.

Personally my business model will not be to sell hashrate or mining contracts. I would set up a new corporation for the customer so the full tax benefits of depreciation pass through to them. The customer would be a non-voting 100% owner of that corp. My parent company would charge the mini-corp a fixed $/kWh rent rate or something + a setup fee. Whatever profit is left is paid out to them as a distribution every month or so. Still need to talk to a lawyer to make sure this is a sound idea.

This also gives them a powerful incentive to re-invest that distribution into more shares (a.k.a. more equipment), in order to keep Uncle Sam away and not pay 22%-37% the next year. If they want to quit and liquidate, they owe 22-37% from depreciation recapture. That's one way to retain customers without forcing them into 12+ month contracts.

Update:

I realized it would be a bad idea to buy very old ASICs, because not only will they break, but because the cost of building that capacity is too high. It makes no sense to buy an Antminer S9 for $280 if the buildout cost (electric + ventilation) is $200/kW. The newest gen is also a bad idea, because it's harder to insure and I risk more capital per terahash.

The T17 or S17 seem to be better choices. They have reasonable efficiency while I can still take advantage of changes in the energy index.

I wrote code to analyze ERCOT rates for the past 30 days. For the T17, the all-in rate would be 6.72¢ with 100% uptime, or 6.08¢ with 95% uptime. That comes out to 3-6% more profit on average by powering down during peak prices.

Once I can get rid of the pesky sales tax and buy transformers instead of using ONCOR's secondary service, the rates would be 5.7¢ for 100% uptime, or 5.13¢ for 95% uptime.

Of course, these rates would be 1-1.5¢ higher during the summer. There is no way to avoid that.

You mean power supplies with multiple rails? This is why I prefer PSUs with one huge single 12V rail instead of those with multiple 12V rails. Server supplies are good because they usually have just one beefy rail.

At the very least, the number of rails needs to match the number of video cards. Let's say I have a 1400w PSU with 4 12v rails @ 30A each. I can either power one RTX 3080 Ti on each rail (4 cards total), or two RTX 3060's on each rail (8 cards total). I would not use RTX 3070's unless they are downclocked to RTX 3060 levels.

I'm not worried about this problem with ASICs, since I just use Bitmain or server PSUs.

Here we ago, another 'free energy' scam device...

The physics behind this idea have been known for over 200 years. If it were possible to generate lots of power cheaper than existing sources, somebody would've tried it by now. It would probably cost trillions of dollars to build a bunch of flywheels that had enough mass to generate even 1 megawatt.

It's probably a better idea to build more hydroelectric dams, or make nuclear power cheaper.

It looks like the customer needs to go through a long process:

https://electricityplans.com/electricity-sales-tax-exemption/

Probably only worth the time/expense for farms larger larger 1 MW. The extra 8% probably comes out to 2-4% less profit.

I agree that ASIC mining, especially SHA-256 mining, is a very competitive industry. I already know that a small operation can't compete on SHA-256 in the long run. The ASICs are a means to an end. Power consumption needs to be shored up at first to get good electric rates.

My end goal is to sell SaaS, consulting services, GPU colocation, speculative GPU/CPU hosting, etc. These are products that I think Coinfarm can actually compete on, and I have way more experience mining small coins.

Nice! There are definitely opportunities that pop up from time to time to buy equipment at below-market prices. That's why I prefer slowly building up more equipment over time, instead of buying it all at once. I also like ordering 1 unit of a GPU/ASIC first to learn how it works, then doing a bulk order.