2 ms·

Yes, that's the point. To compete against them directly. Get enough funding to override their advantage. > And you’ll be starting from scratch. I haven't bu

Yes, that's the point. To compete against them directly. Get enough funding to override their advantage. > And you’ll be starting from scratch. I haven't built a mining ASIC (yet), but I built the first open source code for mining FPGAs. I wouldn't be starting from scratch. It'll take a little bit of re-architecture to update that code to multi-core designs (the FPGAs were single core, fully unrolled loops), but that's relatively easy. The more aggressive optimization will be around power consumption, which I plan to hand off to the engineers working with the target fab's libraries doing the HDL->layout conversion. This isn't rocket surgery. Mining ASICs are perhaps the easiest ASICs to build. It's even feasible to do manual layout, since the design is so repetitive. Just need to hand layout one round of SHA-256 which amounts to only a couple adds and mix of binary ops. Done. > That’s half a million dollars per day, or $15m per month Minus NRE, manufacturing costs, on-going maintenance, on-going electrical costs, and all the money they wasted mining BCH and playing other games.

Zarath 9y ago
Don't forget AsicBoost!!!
runeks 9y ago
The inherent challenge, as I see it, is that if you manage to produce e.g. an electric car that’s not quite as good as a Tesla, you still have a car that can transport people from A to B. And if you have trouble selling your cars, you just lower the price until all your cars are sold. With mining chips, if you have a chip that’s just half as efficient as the best one, your customers have zero incentive to buy your product, as efficiency is all that matters. An inferior car can still transport people, even though it’s not the best car, but a mining chip whose electricity cost is $5,000 to mine a Bitcoin that costs $3,500 is worse than useless (you might as well throw it out). That’s quite a difference. I’m not saying this makes your venture impossible, just that it’s an important aspect to consider.
- fpgaminer 9y ago
  Good points. Efficiency isn't actually all that matters; it's total cost versus revenue. Every mining device has some sort of lifetime to it. Its total cost is cost of the device itself + (electrical costs + maintenance + cooling) * lifetime. Its revenue is a function of its hashrate. So, all else being equal, if someone makes a mining chip with less efficiency than the best, they simply have to sell the chip for less $/TH. But usually all else _aren't_ equal. Say you're mining in the far northern regions of the planet. You can take advantage of the colder climate to reduce your cooling costs, so you don't have to expend money designing chips and hardware with good cooling. But you can't take advantage of hydro or solar power, so power efficiency becomes more important. Or, let's say you mine near the equator. Now you can probably build out a solar farm and get cheaper power than anyone else. But cooling is a major issue. You have to spend more on good chip packages and thermal designs. On top of that, the design of the chip itself can dictate a lot of cost. For example, when I designed my FPGA miners, I made it so that the FPGAs could be both programmed and communicated with by a single, cheap FTDI chip. This made the boards super simple and cheap. Basically, it's a complex equation that can't be boiled down to simple power efficiency.
  - runeks 9y ago
    Sounds like you know a lot more about this than I do. Have you considered leasing as a way to reduce risk for operators? So, rather than mining operators needing to take the risk of standing with outdated hardware, you — as the manufacturer — would take on that risk by allowing mining operators to stop the lease if it becomes unprofitable. This changes the incentive structure so that you stand with a loss if your hardware becomes obsolete.

Zarath 9y ago

Don't forget AsicBoost!!!

runeks 9y ago

The inherent challenge, as I see it, is that if you manage to produce e.g. an electric car that’s not quite as good as a Tesla, you still have a car that can transport people from A to B. And if you have trouble selling your cars, you just lower the price until all your cars are sold. With mining chips, if you have a chip that’s just half as efficient as the best one, your customers have zero incentive to buy your product, as efficiency is all that matters. An inferior car can still transport people, even though it’s not the best car, but a mining chip whose electricity cost is $5,000 to mine a Bitcoin that costs $3,500 is worse than useless (you might as well throw it out). That’s quite a difference. I’m not saying this makes your venture impossible, just that it’s an important aspect to consider.

fpgaminer 9y ago
Good points. Efficiency isn't actually all that matters; it's total cost versus revenue. Every mining device has some sort of lifetime to it. Its total cost is cost of the device itself + (electrical costs + maintenance + cooling) * lifetime. Its revenue is a function of its hashrate. So, all else being equal, if someone makes a mining chip with less efficiency than the best, they simply have to sell the chip for less $/TH. But usually all else _aren't_ equal. Say you're mining in the far northern regions of the planet. You can take advantage of the colder climate to reduce your cooling costs, so you don't have to expend money designing chips and hardware with good cooling. But you can't take advantage of hydro or solar power, so power efficiency becomes more important. Or, let's say you mine near the equator. Now you can probably build out a solar farm and get cheaper power than anyone else. But cooling is a major issue. You have to spend more on good chip packages and thermal designs. On top of that, the design of the chip itself can dictate a lot of cost. For example, when I designed my FPGA miners, I made it so that the FPGAs could be both programmed and communicated with by a single, cheap FTDI chip. This made the boards super simple and cheap. Basically, it's a complex equation that can't be boiled down to simple power efficiency.
- runeks 9y ago
  Sounds like you know a lot more about this than I do. Have you considered leasing as a way to reduce risk for operators? So, rather than mining operators needing to take the risk of standing with outdated hardware, you — as the manufacturer — would take on that risk by allowing mining operators to stop the lease if it becomes unprofitable. This changes the incentive structure so that you stand with a loss if your hardware becomes obsolete.