8

Modern miners on the market have up to 110 TH/s. Is this the true hash rate (singular SHA-256 hashes per second) or the net hashes relating to Bitcoin specifically (since Bitcoin involves double hashes 'SHA256(SHA256(x))'. If so, that would set its true hash rate of 220TH/s.

Thanks

Improve this question

1 Answer 1

Reset to default
8

Modern miners on the market have up to 110 TH/s. Is this the true hash rate (singular SHA-256 hashes per second) or the net hashes relating to Bitcoin specifically (since Bitcoin involves double hashes 'SHA256(SHA256(x))'.

It refers to the number of block hash candidates the miners can try per second. While you can think of those as double SHA256, that is actually very far from what the hardware does.

It doesn't actually compute two hashes, in the sense that:

  • it cannot take as input anything but a block header (with some specific structure, of which some parts are precomputed, in the case of ASICBoost),
  • it doesn't output the result of the first hash,
  • it doesn't actually even compute the full second hash result (only whether or not it has a certain number of zero bits up front)

If so, that would set its true hash rate of 220TH/s.

You could say that, but it's a meaningless metric. The hardware cannot actually perform 220 tera-SHA256's per second, for the reasons above.

The only thing it can do is try 110 tera-Bitcoin-block-candidates per second.

Improve this answer
2
  • do you have some details about HW optimizations for hashes with leading zeros? Do mining HW implementations take advantage of the fact that for valid proof of work block hashes the first N bits are zero? If they skip part of the second hash computation they can save a few % of transistor and thus have better efficiency. Or do they actually compute only the first N bits to be able to decide if they have certain number of zeroes?
    – OptOut
    Commented Dec 15, 2022 at 7:55
  • 2
    I stopped paying attention to mining internals probably somewhere around 2011, but even back then, (GPU) mining implementations had lots of optimizations to skip bits of computation here and there. I can only imagine that's become a lot of extensive since then. ASICBoost is one particular impactful example.
    – Pieter Wuille
    Commented Dec 15, 2022 at 14:13

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.