Most accurate formula for comparing Hardware


#1

I have moved this discussion out of the cloud mining thread, we need a accurate formula for comparing Hardware that is easy for users to understand.


Zeropond Cloud Mining
#2

They're selling hashes.

No; they're selling solutions....


#3

They're selling hashes. A hash is only a potential solution. There's no promise it will result in any coins.


#4

A hash is only a potential solution.

No; a hash is just a very ambiguous term, see

There's no promise it will result in any coins.

I never said it did.

They are selling solution rates....


#5

[edit: I'm deleting stuff because there is disagreement about the number of solves in 1 run of the benchmark]

The end use is mainly for comparison to other miners, all using the benchmark, so for comparison purposes all seems OK.

But then what is the conversion from solves per second to hashes per second?


#6

To be clear, each execution of the Wagner algorithm may generate 0, 1, or more solutions. Each solution may be hashed and checked against the difficulty target. So solutions / second is the same as hashes / second. The confusion arises because the CPU miner's solveequihash benchmark shows the average time to run the Wagner algorithm. Maybe we should make a new unit Wagners :slight_smile: So 1 Wagner per second gives, on average, about 2 solutions per second, which is also 2 hashes per second.


#7

Wagners per second ~= 2 solutions per second = 2 hashes per second

@daira has pointed out that it's slightly less than 2, but by convention we have been using a factor of exactly 2.


#8

The 2.8 in his link appears to be wrong and I posted there an explanation of why. In any event your results appear to be 10x my CPUs.


#9

You're adding to the confusion by focussing on number of wagner runs. There are many possible ways to implement wagner, all making different trade-offs in memory versus discarding solutions. This makes the number of algorithm runs irrelevant, and useless for comparison. We want to only talk about solutions/sec, which may be sensibly compared across all algorithms and hardware.


#10

Ok fair enough. But I sense confusion when people use the term "solutions" and then apply a multiple to get "hashes." Every solution gives exactly 1 hash.


#11

@tromp @tim_olson Then the conversion from solutions/second to hashes/second is 1.0? That's not even semantics. It's an identity relation. (as far as users of the software are concerned)

Maybe your beginning post should have started with 'There is no such thing as a hash rate in Zcash"?


#12

No, actually Tim explained himself and made sense instead of just gain-saying other people over a trivial distinction.


#13

gain-saying other people over a trivial distinction.

But solutions versus hashes is not a trivial distinction, because the term hash is ambiguous.
Sure, it's equivalent to solutions if you consider it to refer to the final hash
that's applied to each solution for checking whether it meets the difficulty threshold.
But the problem is that many people use hash to refer to a single run of the wagner
algorithm, as I've explained before. It could also refer to all the blake hashes performed within wagner's algorithm. So, with the term "hash" being ambiguous
and having little hope to get everyone to understand the correct meaning, we are
better off using the much less ambiguous notion of solution.

'There is no such thing as a hash rate in Zcash"?

The problem is that there are too many things that can be called hash rate in Zcash.