6 ms·

I am extremely glad that most comments here see the inherent craziness of this idea. I am a hardware designer who participated heavily in mining since Jan 2011,

I am extremely glad that most comments here see the inherent craziness of this idea. I am a hardware designer who participated heavily in mining since Jan 2011, and gave up completely on Bitcoin in summer of 2014. The big problem that Balaji is completely ignoring is that SHA256 cores implemented in silicon have extraordinarily high power consumption relative to their size (mm2). This is due to the incredibly high signal switch rate inherent to implementing avalanche-style ciphers [1] in hardware. Perversely, most modern semiconductor processes optimize for static (leakage) power, since modern silicon designs are more vulnerable to this than to dynamic power. No rational semiconductor maker will want to include this IP on chip, because it will drastically complicate the power distribution metal layers. Additionally, it will require more expensive packaging, since you will need many more power and ground pins to source/sink the required current. I know this first hand after building a mining ASIC myself. I would venture a guess that most SoC chips would cost significantly more due to the extra engineering time and packaging cost needed to include this IP. This entire concept is a non-starter for me. [1] http://research.neustar.biz/2012/02/02/choosing-a-good-hash-function-part-3/ http://research.neustar.biz/2012/02/02/choosing-a-good-hash-...

bryanbuckley 11y ago
I know it's not an actual mining asic/HW module, but didn't some SoCs support sha256 hw acceleration at least 3.75 years ago? https://android.googlesource.com/kernel/omap/+/android-omap-tuna-3.0/security/smc/tf_crypto_digest.c https://android.googlesource.com/kernel/omap/+/android-omap-...
lukifer 11y ago
Out of curiosity: what made you give up on Bitcoin? (I'm still bullish on crypto-economies in general, but am concerned that the electricity costs of PoW creates a downward price pressure that's unsustainable.)
- IkmoIkmo 11y ago
  Why would that be? The maximum PoW can depress the price is the inflation rate. Nobody will pay $1 for electricity to generate <$1 of bitcoin. And the inflation rate of bitcoin is set to drop below that of any large currency like dollars, euros, yen. In other words, the money supply that causes inflation (downward price pressure of currencies) is by design lower than fiat currencies. It's one of the key things supposedly we're not ought to be concerned about with bitcoin, so far I see no reason to think otherwise.
  - teraflop 11y ago
    > Nobody will pay $1 for electricity to generate <$1 of bitcoin. Nobody would willingly, deliberately do so. There are plenty of cases of cryptocurrency mining software being installed on unsuspecting users' machines, causing them to spend a large amount of energy generating a tiny amount of Bitcoin/Litecoin/whatever that gets sent to the attackers. Notably, this has happened to a number of EC2 customers. If you accidentally leak your AWS secret key (by unwisely checking it into version control, or through a buggy deployment process, for example) then more likely than not, you'll very quickly find your account being used to spin up a bunch of expensive GPU instances.
    - IkmoIkmo 11y ago
      True -- but completely irrelevant to my point that even if every single coin that was mined was immediately sold, the limit would be the supply of new bitcoins which is by design lower than that of other currencies like dollars/euros.
  - ssharp 11y ago
    > Nobody will pay $1 for electricity to generate <$1 of bitcoin. Hobbyist miners are convincing themselves that <$1 of Bitcoin will be worth >$1 the future. Major miners seem to have question business models as well, where they are either losing money or have something shady going on with their cost of electricity. BTC still seems to be very prone to pump-and-dump schemes, so major miners may be able to pump the price enough to sell their coins profitably before the next crash.
    - IkmoIkmo 11y ago
      No not really. If you believe $1 of X will be more than $1 in the future, you buy it for $1, you don't pay more than $1 for the manufacturing or mining of $1 of X, it makes no sense from a financial perspective, especially considered the time, effort and substantial risks in purchasing and running mining hardware. The argument that you mine to hold to invest, while mining and selling right away is unprofitable, doesn't really hold when you can just buy and hold instead for cheaper and less hassle & risk. The exception is of course when buying coin isn't an option, but buying electricity & hardware is. Which is why you may see shady operators mine above cost in China, because purchasing bitcoin in bulk in China can't be done easily anymore, a substantial premium cost. But generally speaking, mining expenses approach mining revenues and top out there, as with virtually every other market. There are some caveats sure but they're not the rule. Beyond that though, electricity isn't very relevant. At the end of the day if you have 1 million bitcoins, and the annual mining rewards are 50k bitcoins, the most coins that can be sold is 5%. Regardless of electricity prices, the new supply is 5%. Whether it ends up in the hands of miners or others, that's the inflation and the amount of downward pressure. It's true if electricity is expensive and margins are tight, miners sell more quickly. But in the long-run, the supply on the market is still averaging 5%, there's no magic way to go above that number. You could see miners sell 0% the first half of the year and then the entire 5% the second half of the year when margins are tight and miners sell to get cashflow, but it's still 5% a year, regardless of electricity prices. So 5% is just a hypothetical number, but the reality is that by design it's set to go to 0%, again regardless of whether electricity is cheap or expensive. Which is why I contest the earlier statement by the guy I replied to that electricity prices is something to worry about.
wmf 11y ago
Perhaps power density could be "fixed" by padding out the circuit to take more area. This would obviously increase cost, but probably less than changing the overall chip design.
nilnullzip 11y ago
There are many degrees of freedom in chip design: speed, power, real estate, slew rate, clock rate, etc. The core can have more moderate power density (trading off real estate and speed) if that's what is needed in the application. Probably the figure of merit that is most important is energy/hash, not circuit density or speed.

bryanbuckley 11y ago

I know it's not an actual mining asic/HW module, but didn't some SoCs support sha256 hw acceleration at least 3.75 years ago? https://android.googlesource.com/kernel/omap/+/android-omap-tuna-3.0/security/smc/tf_crypto_digest.c https://android.googlesource.com/kernel/omap/+/android-omap-...

lukifer 11y ago

Out of curiosity: what made you give up on Bitcoin? (I'm still bullish on crypto-economies in general, but am concerned that the electricity costs of PoW creates a downward price pressure that's unsustainable.)

IkmoIkmo 11y ago
Why would that be? The maximum PoW can depress the price is the inflation rate. Nobody will pay $1 for electricity to generate <$1 of bitcoin. And the inflation rate of bitcoin is set to drop below that of any large currency like dollars, euros, yen. In other words, the money supply that causes inflation (downward price pressure of currencies) is by design lower than fiat currencies. It's one of the key things supposedly we're not ought to be concerned about with bitcoin, so far I see no reason to think otherwise.
- teraflop 11y ago
  > Nobody will pay $1 for electricity to generate <$1 of bitcoin. Nobody would willingly, deliberately do so. There are plenty of cases of cryptocurrency mining software being installed on unsuspecting users' machines, causing them to spend a large amount of energy generating a tiny amount of Bitcoin/Litecoin/whatever that gets sent to the attackers. Notably, this has happened to a number of EC2 customers. If you accidentally leak your AWS secret key (by unwisely checking it into version control, or through a buggy deployment process, for example) then more likely than not, you'll very quickly find your account being used to spin up a bunch of expensive GPU instances.
  - IkmoIkmo 11y ago
    True -- but completely irrelevant to my point that even if every single coin that was mined was immediately sold, the limit would be the supply of new bitcoins which is by design lower than that of other currencies like dollars/euros.
- ssharp 11y ago
  > Nobody will pay $1 for electricity to generate <$1 of bitcoin. Hobbyist miners are convincing themselves that <$1 of Bitcoin will be worth >$1 the future. Major miners seem to have question business models as well, where they are either losing money or have something shady going on with their cost of electricity. BTC still seems to be very prone to pump-and-dump schemes, so major miners may be able to pump the price enough to sell their coins profitably before the next crash.
  - IkmoIkmo 11y ago
    No not really. If you believe $1 of X will be more than $1 in the future, you buy it for $1, you don't pay more than $1 for the manufacturing or mining of $1 of X, it makes no sense from a financial perspective, especially considered the time, effort and substantial risks in purchasing and running mining hardware. The argument that you mine to hold to invest, while mining and selling right away is unprofitable, doesn't really hold when you can just buy and hold instead for cheaper and less hassle & risk. The exception is of course when buying coin isn't an option, but buying electricity & hardware is. Which is why you may see shady operators mine above cost in China, because purchasing bitcoin in bulk in China can't be done easily anymore, a substantial premium cost. But generally speaking, mining expenses approach mining revenues and top out there, as with virtually every other market. There are some caveats sure but they're not the rule. Beyond that though, electricity isn't very relevant. At the end of the day if you have 1 million bitcoins, and the annual mining rewards are 50k bitcoins, the most coins that can be sold is 5%. Regardless of electricity prices, the new supply is 5%. Whether it ends up in the hands of miners or others, that's the inflation and the amount of downward pressure. It's true if electricity is expensive and margins are tight, miners sell more quickly. But in the long-run, the supply on the market is still averaging 5%, there's no magic way to go above that number. You could see miners sell 0% the first half of the year and then the entire 5% the second half of the year when margins are tight and miners sell to get cashflow, but it's still 5% a year, regardless of electricity prices. So 5% is just a hypothetical number, but the reality is that by design it's set to go to 0%, again regardless of whether electricity is cheap or expensive. Which is why I contest the earlier statement by the guy I replied to that electricity prices is something to worry about.

wmf 11y ago

Perhaps power density could be "fixed" by padding out the circuit to take more area. This would obviously increase cost, but probably less than changing the overall chip design.

nilnullzip 11y ago

There are many degrees of freedom in chip design: speed, power, real estate, slew rate, clock rate, etc. The core can have more moderate power density (trading off real estate and speed) if that's what is needed in the application. Probably the figure of merit that is most important is energy/hash, not circuit density or speed.