Another performance boost to Funakoshi-Miner


#42

You appear to be comparing a machine with four GPUs to a machine with two GPUs.

What kind of GPUs does each have?


#43

First of all, I want to apologize for what my words may be straightforward.

I do not have enough English to choose nuances or expressions.

I am always cheering for Funa miner.

long ago, In <200, 9>, EWBF was almost never updated.

It stopped at version 0.3 since its performance over zcash’s award-winning algorithm, zennoncat.

And everyone knows that there is a minor that over and over speed than 0.3 EWBF
However, it could not be obtained by a general method.

I saw the source of <200, 9>.
api is not used at all, only the shift operation to handle the blake solver, which has superb great optimization.

I think very few people can make an implementation that can outperform its speed.

And that’s why it stays in v0.3. I thought.

And … there is a minor of funakoshi.

I have seen.

Every time the speed of funakoshi catches up with EWBF
EWBF has been speeding up.

Of course it could be a coincidence, but the speed improvement of EWBF v0.6 two days ago is really huge.

At 1080ti, it maintains almost 80sol/s

From my point of view, I have been brutally or easily told that “competition is good”.

But … In fact, I am very thankful and thankful and respectful to Funakoshi.

Thank you very much.


#44

Stay tuned in just a few days I am going to release a much faster version of Funakoshi-Miner.


#45

99b983892094b5c6d2fc3736e15da7d1

oh my god!!!

very exciting!!!


#46

Gpu are identical. Hashrate and on 4 gives the same speed. The pool is identical. Settings of dispersal are identical. Everything is absolutely identical, except for the number of cards.


#47

EWBF - msi 1080ti armor = 87 sol. (max)


#48

Ok, so the results should be similar.

A better test would be to show results on the 4-GPU rig with one miner and then the other… or the on the 2-GPU rig with one miner and then the other. Otherwise we must wonder about the influence of other factors.

That being said, looks like 69 versus 77… so about a 10% difference. Once is clearly faster, but I don’t think it’s fair to call the other “very slow…”

:slight_smile: Competition among miner-makers is good for the miners. :slight_smile:


#49

Frankly speaking, to recustomize laziness. But. I promise to lay out tests of diggers after a release of the new funokoshi version. I ask to take me a word. Funokosha good digger, stable but ewbf quicker so far.
I apologize for my English, I write as I can. Not the native speaker.


#50

Release v4.11 has been published.

It contain a small speed boost plus a bug-fix.

Avg 51 - 52 sol/s on GTX 1080


#51

I have released v4.12 with more performance :slightly_smiling_face:

To learn more about the miner and to download it visit my github repository.


#52

Release v5.1 push performance 1% higher compared to v4.12

~ 52 sol/s on GTX 1080


#53

in release 5.1 && 1080ti && power limit 70%

EWBF 6.0 73sol/s
funakoshi 5.1 61sol/s

but this score is closing.
Thank you!!


#54

I have meassured performance of v5.1 on 1080ti:

Avg 66-75 sol/s for GTX 1080ti

To be exact the results should be expressed as a range.

Funakoshi Cuda Kernels are executed in less time compared
to EWBF Cuda Kernels as shown by Microsoft Visual Studio
profiler.

Miners should only trust the pool (after running for a few hours).
The pool has no reason to prefer any miner software.


#55

Release v5.2 improves performance of equihash 144,5 by 0.5%

To check out the solver visit my github repository.


#56

Hi, i have 2 x 6 1070ti rigs on one Funakoshi miners works normaly and on other when i start it says cuda failed to alloc memory, and crashes why is this??


#57

Please open an issue on my github issues page and tell
me as much information about both rigs as possible.

Many users are running funakoshiMiner on large rigs without
any problem. I will try to help you after you open an issue.


#58

I’m telling the exact numbers.

no overclocking
70% power limit
Various Hardware

A lot of people will test it.

It is the average value of 30min.
The range is so wide that it does not make much sense.

I think, 66 ~ 75sol / s is probably a 100% power limit.


#59

The statistical nature of the equihash algorithm forces the number of sol/s to
change widely all the time. If the range is too small we can conclude that the
miner software eliminates from its report the lower numbers. Theoreticaly
there can’t be an implementation without a wide range.

Miners, trust only the pool. The pool is neutral to mining software. From the
pool you earn your money. The pool has no hidden interests regarding
which mining software is better.


#60

your # of accepted/rejected shares keeps resetting. It should keep the count going until the session has closed.


#61

it is “average value within 30min”

it is not per every sol/s.

And the compensation of the pool is constantly changing.

But for the same speed of different accounts in the same pool,
They make exactly the same compensation.

You can compare this method.
Compare them.
I have already done it.

Money is sensitive to people.
So many people are testing it in a variety of ways.

In fact, there are a number of factors that affect speed
I have tried many things.

Again, I mentioned the 30 minute average.
Not every time.

thank you

ps. I have already written that every sol/s have very lare range.