Bitcoin Forum

According to [1] we're on top of the world with combined cpu.

Can anyone site another network of combined cpu for a cause over 208,000 TeraFLOPS?

Keep on crunching!



[1] - short unofficial list of sources
http://en.wikipedia.org/wiki/Grid_computing
http://en.wikipedia.org/wiki/List_of_distributed_computing_projects

No nothing even comes close (like within 20% of Bitcoin).   That also applies to centralized projects too (grid computing, super computers, cloud networks, etc).  Absolutely nothing is even in the same magnitude.  

There is some controversy though.  The Bitcoin network uses integer operations.  So technically the Teraflops (floating point ops) of the network is always 0.0 and always will be 0.0.  Given almost every measurement of computing power measuring floating point the Bitoin floating point estimate exists simply to allow some kind of comparison to other projects.  That being said there is no way to calculate the TFLOPS the Bitcoin network WOULD be capable of (if it ran floating point computations).  The estimate should be seen as a "best guess" only.  That being said Bitcoin is sooooooooooooooooooooooooooooooooooooooooo far ahead of anything else the crude estimate is likely "good enough".

There is a project out there called Internet that uses a lot of CPU.

I'd bet if you took the combined computing power of the entire internet it still wouldn't be anywhere close.... It doesn't take a lot of computing power to serve or view webpages.

no, bitcoin is far and away the most powerful computing effort humanity has ever produced, it dwarfs folding at home and government supercomputers...kind of makes you think...

The only reason for this is money though.  If any of the other programs offered money as a reward for computing at the rate bitcoin does, they'd be taking a large chunk out.  Makes you think of how greedy people are ;) haha

And it's also a pity.

Might be an opportunity for me to advert for my proposal about folding molecules instead of computing sha-256:
https://bitcointalk.org/index.php?topic=108888.0

It depends if you're talking about the computing power dedicated or the power actively used at any given time. The former is still much higher, but the latter, perhaps not.

Uhhhh, the sha-256 computation secures the blockchain, kind of important

My proposal consists in doing both:  securing the block chain while folding molecules.

can a teraflop supercomputer be configured to crunch integer operations?

your proposal involves forking the blockchain for your cause.  There may be other causes more valuable to society over time (which will constantly change).  One thing we all agree has value is a device that can store our productive output in a unit to be transferred to others for their productive output.

No fork.  A brand new chain used as a secondary, alternate cryptocurrency.

That's a pretty awesome idea actually!

Thanks.  However, it's technically difficult and I'm not even sure it is possible.  I think it is, but until someone actually implements it, nobody really knows.

Sure, you can think of an integer as being a float with only zeroes behind the decimal point.

Those numbers in the Wikipedia article are from February 2012, computing power has more than doubled since. By any reasonable measure Bitcoin is the largest distributed computing effort by far.

And the best thing is, it doesn't matter how much computing power you throw at bitcoin, the result (time between coins/transactions) is going to stay the same. For other projects, throwing more computer power at a problem results in faster solving of the problem. In bitcoin, thanks to adaptive difficulty, computing power is completely irrelevant, other than to decide who gets a block.

The biggest computing effort in human history: fueled entirely by greed.

+1 to what Holliday said

And no, no other project has more computing power than bitcoin. Of course they try to hide our power by removing us from FLOPs statistics because "bitcoin does 0 FLOPs"  :D

Unless of course the 22 confirmed ASIC minirigs go online and immediately fork the chain.

Current total bitcoin computing speed: ~ 21.000 GHash/s
22 ASIC minirigs: ~ 22.000 GHash/s

Descendant of http://en.wikipedia.org/wiki/ECHELON could have more, much more CPU power.

good luck finding out for sure! :D

It would be beneficial for someone to come in behind Bitcoin and offer up a small reward based project that can utilize older mining machines which are of less use because people have moved on to faster machines.


There will likely be a bunch of GPU machines out there that will be outdated when it comes to mining. Just as CPU mining reverted to being people's regular computers.

Most of the world spends 8-12 hours average a day working to get money so they can eat and have a better life for themselves and their families. Is that greed?

Myself and many others mine bitcoin as supplemental income. I'm making a buck fifty a day mining. Is that greed? Am I greedy for wanting that buck fifty?

If you want to talk about greed, see the banksters and monopoly men. JP Morgan. The Rockefellers. Donald Trump. Warren Buffet. Bill Gates. Those men fit your labeling much more closely. And I am sure they sneer at bitcoin (if they've heard of it).

Then were is our Guinness Book of record submission at ?

Source please.  I haven't been lurking here for a long time.

Last year I did a rough estimate and came to the conclusion that if you wanted to hire super computers #1 down to whatever it takes, you would end up hiring the top 40 to run a 51% attack for as long as you want to keep the attack up.
I only looked at their GPU's and basically accounted 0 for all that only had CPUs which might be slightly incorrect.

Current total hashrate: http://bitcoin.sipa.be/

BFL ASIC minirigs are specced @ 1000 GHash/s (the one to the right, not the middle one): http://www.butterflylabs.com/products/

… of course this does not answer the question

Those super computers are programmable to do all kind of shit while in the age of ASICs we will have very dumb nodes that will not be suitable for anything but bitcoin, which is kind of sad.

here: https://bitcointalk.org/index.php?topic=29722

Unfortunately stalled.

With more and more fpgas and asics coming we need to formulate the record differently, though.

Is there 3rd party proof of their existence/performance (in the wild), or is it anther scam (no shortage of those)?

The net total speed of the bitcoin network in FLOPS is exactly zero.  I'm not sure that anyone has ever bothered to measure or record the fastest integer system.  We kinda suspect that we might be it, but as far as I know, there is no official record.

That is a very interesting question indeed.

Considering the way this whole thing is structured (full prepayment for preorders, no refunds until start of shipping, no real adress, incorporation and domain setup through proxy agents, telephone # is assigned to a mobile phone) it's perfectly set up for very very interesting results. I'll let everyone draw their own conclusions what kind of results.

Maybe a credit with 0% interest? We'll just get our coins back and some excuses like "sry, we failed to create the product, here is ur money back".

I just need to say that you are thinking of the world wide web.  Technically bitcoin is part of the internet so therefore the internet as a whole has to me more powerful.

sure. but if you have a fast car and live in place where you are never allowed to go faster than 70miles/hour you might still want to brag about how fast your car could go.  ;D

I kinda feel sorry for all the spent GPU's out there forced to calculate a problem to which the solution is worthless. it would be cool if the network was used to calculate something like the next prime number...

well the calculation ITSELF is kinda useless, but then it's purpose is to secure the network and all transactions of a (currently) hundred million dollar and growing economy. that aspect is far from useless. indeed, it's the core of bitcoin in many ways.

Wow, I can't think of a better thing to devote my computing power to, than open-source, decentralized, power-to-the-people, currency.

Centralized money is the root cause of much misery and suffering in the world.

and it could be done with 10 Pentiums if there wasn't an arms-race going on for purposes of "Dude, I need more of dem bitcoins than George next door."

Said arms race secures the blockchain against attack, it is a good thing.

No, it doesn't. If the mythical ASICs should proove to be true, the 22 confirmed orders for BFL SC ASIC Minirigs will have more hashrate than all current miners put together.

And it's a good think that this transition is happening now, and not 10 years later when everyone is using bitcoins.

So the attacker posed as 22 distinct separate entities?

If you allude to the thoery that an attacker could simply pay miners to hash the way the attacker wants, I don't see that ASIC has anything to do with it, unless the mind control beams they emit slacken miner morals more than the beams emitted by GPUs do?

-MarkM-

my point was...if we were doing this on 10 pentiums governments would crush us like a bug

as it stands we operate the most powerful computing effort that has ever existed

 8)

i don't know if this has been answered, is this once again a thread resurrection?

your argument is simply not possible. the reason is, that although protein folding is an optimization process and you could replicate this measure of difficulty in a way of finding a solution that is at least of some certain quality, it still won't do the necessary trick. this kind of optimization has an unknown optimal value and the amount of work is in no way predictable. the reason why it is unpredictable is based on the fact, that you optimize real-world data which is not homogeneous and contains additional information (in an abstract sense). this means, this is a calculation based on processing data and extracting information (the structure). the sha256 hashing done for the blockchain is homogeneous and does not incorporate any additional information. therefore it's progress and outcome (after trying hashes) can be predicted even before the procedure has even started.

also, imagine the inverse: you somehow (by physical experiment) already know the solution to a protein folding problem and now you get it as your workload. this would make you the immediate winner and you get the block reward. hence, the pre knowledge saved you cpu power and that's not desirable in any way.

thus, securing the blockchain (in the way bitcoin does) must not depend on any higher information, just raw data plus a "useless" computation.

I'm considering using a relative progress in the progression of the action as a measure of difficulty.

difficulty = - ln(A_new/A_previous)

Historical data about difficulty progression should, at least in a statistical sense, give some predictability.


Well, this is not much different from SHA-256, is it?  I mean, I could know some stuff you don't know about SHA digests and then this "pre-knowledge" would save me some CPU.

Edit.  Even if it was possible to know the solution of a protein folding problem (I doubt it is:  you can know the geometrical structure of a folded molecule by crystallographic methods I guess, but you won't know the folding process), that wouldn't be a problem, unless you're ready to do that all the time for all molecules.  You would just trade CPU against the real life time and efforts you'd put in your experiments.  To me, this would be fair enough.

It can rely on higher information, as long as nobody has such information.  If nobody has it, it's just as if it didn't exist.  And nobody can magically solve the problem of folding molecules without doing brute force attempts, just as nobody can find an arbitrary small digest without brute force attempts.  I'm just not convinced by your argument, as I don't see any fundamental difference between digests and folding molecules.

We can scale the difficulty of the SHA-256 problem over a huge range, with constant effort required to verify the results.  I don't think that protein folding has this property.  Actually, I don't think that any "useful" work has that property.

I don't see why it wouldn't.  In the scheme I propose, we would use many molecules.  Actually, miners could incorporate new folding challenges in each new block.  There are certainly lots and lots of molecules scientists would be interested in knowing how they fold.

And even if there are not enough molecules, I wouldn't be surprised if there were many other physical problems that require a brute force method to solve their least action formulation.

I'm sorry dude, I think you are confused about the properties of either SHA-256, protein folding, or both.  I've read a ton of threads on these forums started by people that wish that we could use a different algorithm, and so far not a single person, including you, has shown an algorithm that has the properties that we need.

Just a quick recap:

1.  The output must be provably linked to the input, and the input must contain proof of the block (the Merkle root).
2.  The output must not be guessable by any means other than doing the attempt.
3.  The chance of success must be scalable over a huge range (probably at least 128 bits, bare minimum, and that might even be pushing it).
4.  The output must be easily verifiable.

Protein folding seems to satisfy condition 2, and maybe condition 4.  But conditions 1 and 3 conspire against all of these efforts.

That's a nice summary.  It will certainly be useful for me if I want to prove that my proposal works.  Thanks.


I've proposed something for condition 1:  encoding the hash of the block with a permutation of at least 58 molecules.  The actions would be searched in a decreasing difficulty order.


As for 3, I'm pretty sure we can scale the chances of success as much as we want.  We just need to increase the number of challenges.

Good luck man.  I think you are wasting your time, but I wouldn't mind being shown that I'm wrong about this one.

I might be a bit wrong with what I wrote above, but it's certainly in the right direction, i hope.
About this: I don't get it. What do you permutate?! A protein is just one molecule. What you could permutate are the hydrogen atoms, but that's take care of anyways.
Have you looked into the newer developments of how those foldings are calculated? The best methods, and those which are actually useful, have several layers of abstraction and don't deal with atoms at all.
Maybe, you find some algorithm ideas here: http://www.tbi.univie.ac.at/~ivo/RNA/ ...

I would encode the hash using a list of molecules.

Given a list of molecules, and assuming there is a standard way of serializing their chemical formula in a unique, unambiguous manner, this list has a unique basic order A, B, C, D, ...
If this list is at least 58 long, then it is possible to encode any number from 0 to 2^160 using a permutation of A, B, C, ..., for instance F, G, T, Z, A, C....  This is due to the fact that 58! > 2^160.
So the miner must find solutions in a decreasing difficulty order.  This means that he must first work on F, then he must find a better solution for G, then an even better solution for T, and so on.
Then a block is not considered valid unless its hash encodes into a permutation of the molecules in their decreasing difficulty order.

The way the miners would solve the problem does not matter much.  It's a bit like some people with bitcoin found better technical solution to compute hashes, either software or hardware.

what? the hash encodes into something? how that?
and does someone know the difficulty order?

I told you, you can make a correspondence between hashes (there are 2^160 of them) and permutations of 1, 2, ..., 58 (there are 58! of them).  See https://en.wikipedia.org/wiki/Permutation#Numbering_permutations

For a molecule, the performed difficulty is the logarithm of the ratio between previous action and new action.
You make a list of given molecules in their decreasing performed difficulty order.  It has to correspond to the corresponding hash value (given correspondence mentioned above).

Parts of this are very unclear because he talks about two different constraints using the same terms.

I think he's saying that you have at least 58 different molecules, and you do work on all of them.  The order that you do the work is defined by the the hash of your potential block.  Then, for your block to be valid, the difficulty is computed as the sum of the change in least action of your proposed paths on all of the molecules.

What he means by "decreasing difficulty order" is a mystery to me.  The ordering obviously can't satisfy both the hash bits, and the change in difficulty at the same time.

Just my guesses and attempts to make it clearer for everyone.  I'm sure he'll correct me when I'm wrong.

Yes.


Yes.


Not exactly the sum of changes, but otherwise yes that's the overall difficulty.

If the action for block n and molecule M_i is A(n, M_i), then the difficulty D_{n,i} is

D_{n,i} = ln( A(n-1, M_i)/A(n, M_i) )

And the global difficulty D_n is the sum for all i.


Well, the permutation is defined by the hash, and you use it to search paths so that D_{n,i} is decreasing with i.


Yes it can.  The miner just has to work very little on the molecules "to the left" and harder with molecules "to the right".  He can just set his requirements to match those of the previously computed path in the list.

The hash defines a permutation of the molecules.  For the block to be valid, this permutation must match the one defined by the decreasing difficulty order.