WTF is this? Someone found a trick for fast mining? (part II)

[valiron]

This comes from the thread:

WTF is this? Someone found a trick for fast mining?


https://bitcointalk.org/index.php?topic=1045381.0

That old thread from May 2015 was locked by moderation (why?)

In that thread it was observed statistical anomalies that to my understanding were signs of use of a faster mining algorithm.

The blocks in question were mined by Antpool.

I think, in view of recent events, it would be good to comment. I am curious to know if gmaxwell has changed his views.

It is worth quoting my post #75:

I withdrew any claim. So, stop repeating the same thing over and over. Already Mr gmaxwell rated me as a scammer (at the same level as other people having stolen bitcoins to others) and has tried to bullshit my expertise, which I think at this point says more about him than about me.

But I do believe and I will repeat that some unusual patterns do deserve attention, in particular when the numbers show that these events are extremely rare.

Without being paranoic it is conceivable that some people found a boost on the mining performance (it wouldn't be the first time this happens), and they try to hide it for their own interest.

I fully agree that this possibility has to be treated with caution, but it cannot be ignored and we should be on the look up. If this appears to be the case at the end, some people in this forum will have collaborated concealing this fact. They will bear that responsibility.

On my side I would restrict my comments to statistical facts.

[dannon]

If you look on the mining subform you can see people talking about it. It's called asicboost. Antminet hardware has it built in but it is only used when miners are connected to antpool.

[European Central Bank]

i thought it was nothing to do with 'faster' mining, simply more profitable mining because of less power consumption.

[achow101]

No, Asicboost does not make mining hardware run faster and somehow generate blocks faster. Your original question, assumption, and theory is still invalid; those "fast" blocks are due to variance in block finding. Asicboost does not make miners make blocks faster, just more efficient so they end up consuming less electricity and thus have fewer costs.

[valiron]

No, Asicboost does not make mining hardware run faster and somehow generate blocks faster. Your original question, assumption, and theory is still invalid; those "fast" blocks are due to variance in block finding. Asicboost does not make miners make blocks faster, just more efficient so they end up consuming less electricity and thus have fewer costs.

That's your opinion. Thank you. We all have one. Some of us even have arguments.

Can you explain why the old thread has been locked by moderation?

[achow101]

That's your opinion. Thank you. We all have one. Some of us even have arguments.

It is my opinion, it is a fact stated in the asicboost paper itself:

Through gate count reduction on the silicon AsicBoost improves two essential Bitcoin mining
cost metrics simultaneously and by a similar factor: the energy consumption (Joule per Gh) and
the system cost ($ per Gh/s). With the system cost being proportional to the capital expenses of
a Bitcoin mine and the energy consumption being proportional to its operating expenses,
AsicBoost reduces the total cost per bitcoin mined by approximately 20%.

Asicboost does not make existing GH/s somehow mine blocks faster without change the GH/s (in fact, it can't even effect non-asicboost chips because it is partially a hardware optimization). It makes the cost per GH/s lower.

Can you explain why the old thread has been locked by moderation?

No, I can't. I was not a moderator at that time. It was probably locked because the topic was beaten to death.

[valiron]

No, I can't. I was not a moderator at that time. It was probably locked because the topic was beaten to death.

Doesn't seem to be a good reason, or valid in view of recent events.

Thank you for your opinion.

[eXpl0sive]

No, I can't. I was not a moderator at that time. It was probably locked because the topic was beaten to death.

Doesn't seem to be a good reason, or valid in view of recent events.

Thank you for your opinion.

You got me curious. When you say recent events, what are you referring to? Have these patterns been observed again lately?

[achow101]

You got me curious. When you say recent events, what are you referring to? Have these patterns been observed again lately?

The recent events he is referring to is the fact that it was recently discovered that Bitmain has implemented and possibly used the covert version of asicboost in their own mining operations.

[BitcoinNewsMagazine]

If you look on the mining subform you can see people talking about it. It's called asicboost. Antminet hardware has it built in but it is only used when miners are connected to antpool.

Correct me if wrong but my best understanding of the situation is that current production S9 miners have a covert form of asicboost embedded in the hardware, but just using AntPool is not enough to take advantage of the power savings. Proprietary software is necessary that Bitmain keeps to themselves, or perhaps shares with some good customers. The average miner does not have access to the covert asicboost hence the unfair advantage.

Edit: seems overt asicboost is available today for power savings if you use your own mining pool according to this reddit post.

[valiron]

The recent events he is referring to is the fact that it was recently discovered that Bitmain has implemented and possibly used the covert version of asicboost in their own mining operations.

You answer for me? That's part of your moderation duties? I think you have been the only one to mention "asicboost". Do you have any statistical evidence for the possibility you claim?

A healthy moderation should stay neutral and not participate in the discussions, nor lock threads.

[achow101]

You answer for me? That's part of your moderation duties?

A healthy moderation should stay neutral and not participate in the discussions, nor lock threads.

Now that is a load of bullshit. You mean to say that moderators are not allowed to participate in discussion in which they are interested and have expertise in? Moderators are participants in this forum too, and were posters before they became moderators. I am posting as myself, not as a moderator, nor am I posting on behalf of the forum.

I think you have been the only one to mention "asicboost". Do you have any statistical evidence for the possibility you claim?

Reverse engineering has found that the mining chips used in Bitmain's Antminer S9's and R4's (and perhaps more, I don't know off of the top of my head) contain the circuitry necessary for asicboost to work. It was also found that antpool and the publicly available firmware for the antminers contains the codepaths and api calls necessary for overt asicboost to work, and in fact people have managed to get their miners to use overt asicboost. See https://www.reddit.com/r/Bitcoin/comments/63yo27/some_circumstantial_evidence_supporting_the_claim/dfy5o65/

[valiron]

Now that is a load of bullshit. You mean to say that moderators are not allowed to participate in discussion in which they are interested and have expertise in? Moderators are participants in this forum too, and were posters before they became moderators. I am posting as myself, not as a moderator, nor am I posting on behalf of the forum.

You should watch your language and be respectful...at least when you use your moderator account.

Obviously anyone opinion is welcome, but better to keep one user account for moderation and another for your own opinions.

I think you have been the only one to mention "asicboost". Do you have any statistical evidence for the possibility you claim?

Reverse engineering has found that the mining chips used in Bitmain's Antminer S9's and R4's (and perhaps more, I don't know off of the top of my head) contain the circuitry necessary for asicboost to work. It was also found that antpool and the publicly available firmware for the antminers contains the codepaths and api calls necessary for overt asicboost to work, and in fact people have managed to get their miners to use overt asicboost. See https://www.reddit.com/r/Bitcoin/comments/63yo27/some_circumstantial_evidence_supporting_the_claim/dfy5o65/

That's not statistical evidence of blocks mined by that procedure.

[achow101]

You should watch your language and be respectful...at least when you use your moderator account.

Obviously anyone opinion is welcome, but better to keep one user account for moderation and another for your own opinions.

No. I only use one account for everything, both for my opinions and when I moderate. It is very clear when I speak as a moderator because the only time I do is when I moderate, and right now, I have not done any moderation action on this thread. Staff members using alt accounts is frowned upon.

I think you have been the only one to mention "asicboost". Do you have any statistical evidence for the possibility you claim?

Reverse engineering has found that the mining chips used in Bitmain's Antminer S9's and R4's (and perhaps more, I don't know off of the top of my head) contain the circuitry necessary for asicboost to work. It was also found that antpool and the publicly available firmware for the antminers contains the codepaths and api calls necessary for overt asicboost to work, and in fact people have managed to get their miners to use overt asicboost. See https://www.reddit.com/r/Bitcoin/comments/63yo27/some_circumstantial_evidence_supporting_the_claim/dfy5o65/

That's not statistical evidence of blocks mined by that procedure.

The entire point of covert asicboost is that it is covert, you cannot detect it and there is no statistical evidence that it is being used. The only evidence you can have for it is to examine the software and hardware being used by the suspected party. So far, the hardware and publicly available firmware has been examined and confirmed to be capable of using covert asicboost. Greg has examined a private software and firmware and confirmed that those software came from Bitmain and that they had covert asicboost implemented. Regardless, there is proof that asicboost (covert and overt use the same hardware) is implemented in production hardware and the overt version can be made to work with production hardware and software. Whether Bitmain is using covert asicboost privately is unknown, but likely given the available evidence.

[valiron]

Someone pointed me to that post with an explanation of why the other thread was locked. Is this true? It is disturbing to say the least...

Btw, the way valiron handled the first few people who trolled him in this thread is probably indicative of the way I should handle monsterer, but what was more shocking is how gmaxwell and his gang railroaded valiron and even apparently deleted Come-from-Beyond's post wherein CfB had linked to this white paper just today:

http://rakeshk.crhc.illinois.edu/dac_16_cam.pdf

What is incredible is to see gmaxwel (and the other huge egos over there in Bitcoin Technical Discussion) have his arrogant, totalitarian ass (their arses) handed to him (them) by valiron (who is apparently a PhD level researcher) and so what does Gmaxwell do? Today when CfB posts, he locks the thread and does his usual Hitler tactics.

Fucking amazing.

I will do my damn best to make the Bitcoin killer and dethrone Blockstream. I hope you all have noticed that Blockstream's Segregated Witness proposal is a Trojan Horse takeover of Bitcoin.

[DannyHamilton]

The entire point of covert asicboost is that it is covert,

I don't think that is true.  It's one of the problems with calling it "covert AsicBoost". It gives the false impression that it is "wrong" or "bad" or "secretive".

The entire point of AsicBoost is to generate SHA256 hashes as efficiently as possible. The point of "covert" instead of "overt" is that with "covert" you don't need to make a mess of the block version number (which is better used for actually keeping track of the block version).  The fact that it is not easy to tell that a block was mined using the "covert" method is a side effect, and not the point.

Ignoring AsicBoost for a moment, you'll notice that it is impossible to tell if any given block was mined with an ASIC (instead of a GPU).  People use ASIC because they are more efficient.  Nobody calls them "Covert ASICs". The point of using an ASIC isn't to hide the fact that you are using it, that's just a side effect.  The point (just like with AsicBoost) is to mine as efficiently as possible.

[achow101]

The entire point of covert asicboost is that it is covert,

I don't think that is true.  It's one of the problems with calling it "covert AsicBoost". It gives the false impression that it is "wrong" or "bad" or "secretive".

The entire point of AsicBoost is to generate SHA256 hashes as efficiently as possible. The point of "covert" instead of "overt" is that with "covert" you don't need to make a mess of the block version number (which is better used for actually keeping track of the block version).  The fact that it is not easy to tell that a block was mined using the "covert" method is a side effect, and not the point.

Well covert asicboost requires more computing power than overt asicboost. You need to find several merkle roots which collide in the last 32 bits and this requires more computing power and more memory than overt asicboost. That additional requirement makes covert asicboost less efficient than overt asicboost. I would think that a miner would want to use the more efficient overt asicboost method to get the most performance gains as possible rather than using the less efficient and more taxing covert asicboost unless they are trying to hide the fact that they are using asicboost.

[nibor]

Reading:
https://lists.linuxfoundation.org/pipermail/bitcoin-dev/2017-April/013996.html

It states:

To find multiple roots with the same trailing 32-bits the miner can
use efficient collision finding mechanism which will find a match
with as little as 2^16 candidate roots expected, 2^24 operations to
find a 4-way hit, though low memory approaches require more
computation.

So does anyone know if anyone has done the statistical analysis on the number of transactions in blocks to see if there is an above expected number of blocks with 2x a prime number of transactions?

Also for this to work with publicly available kit when the miner connected to the pool it would be very obvious from the work it was sent to processes. Just a wireshark on the packets would tell you straight away.

Also I am still not clear on the maths of this. Take an antminer S9 @ 11.5TH/s. This will grind through the nonce 2600 times a second! So even if you allow +/-5 on the timestamp that still means 260 collisions you have to find a second and send to the ASIC - so to feed it 4 way hits you need 4Ghash dedicated to that. Which seems hard as this assumes high memory - so a GPU - but 4Ghash is more early ASIC speed.... So really not sure how it all adds up. Unless there is an ASIC for the collision search too.

[ddosamerica7]

This comes from the thread:

WTF is this? Someone found a trick for fast mining?


https://bitcointalk.org/index.php?topic=1045381.0

That old thread from May 2015 was locked by moderation (why?)

In that thread it was observed statistical anomalies that to my understanding were signs of use of a faster mining algorithm.

The blocks in question were mined by Antpool.

I think, in view of recent events, it would be good to comment. I am curious to know if gmaxwell has changed his views.

It is worth quoting my post #75:

I withdrew any claim. So, stop repeating the same thing over and over. Already Mr gmaxwell rated me as a scammer (at the same level as other people having stolen bitcoins to others) and has tried to bullshit my expertise, which I think at this point says more about him than about me.

But I do believe and I will repeat that some unusual patterns do deserve attention, in particular when the numbers show that these events are extremely rare.

Without being paranoic it is conceivable that some people found a boost on the mining performance (it wouldn't be the first time this happens), and they try to hide it for their own interest.

I fully agree that this possibility has to be treated with caution, but it cannot be ignored and we should be on the look up. If this appears to be the case at the end, some people in this forum will have collaborated concealing this fact. They will bear that responsibility.

On my side I would restrict my comments to statistical facts.

How on earth does that work? What method could possibly increase mining probability? As far as I know, the only method I have ever seen was mining when the network hash rate goes down....

[DannyHamilton]

. . . if you allow +/-5 on the timestamp that still means 260 collisions you have to find a second and send to the ASIC . . .

Why would you only allow +/-5 on the timestamp?  The protocol allows a much larger range than that.  In most cases, you should be able to get somewhere close to between -3300 and +7200.