Interesting information RE: Total network strength

[td]

If you read the article on supercomputers, there are a few interesting notes. I understand the Pflop comparison of what a super computer is capable of is not exactly the same as what bitcoin is doing, however, the largest super computer is capable of:


Whereas, according to the article, the largest "Quasi supercomputing" with a distributed computing system is:

2010    Tianhe-IA    2.566 PFLOPS    National Supercomputing Center, Tianjin, China

The fastest cluster, Folding@home, reported 9.739 petaflops of processing power as of mid January 2011. Of this, 7.1 petaflops are contributed by clients running on various GPUs, 1.8 petaflops come from PlayStation 3 systems, and the rest from various computer systems.[9]

SETI has a similar PFLOPS rate in the low 9's.

I just look at bitcoinwatch.com and it says the bitcoin network hash rate is currently at 11.7 PFLOPS. Bitcoin in one way could be considered the largest grid super computer in the world. Very impressive fact. Feel free to roast me if I am completely wrong. I do understand the difference between the single point and double point calculations and am not completely sure what the folding@home and seti@home are calculating if it is the same as what bitcoin is calculating.

[1714061004]

1714061004

[PLATO]

Is there a thread somewhere explaining how bitcoinwatch is converting between hashrate and petaflops? It seems to me that they are totally different operations; e.g. hashing is fixed-point, flops are floating point

[rezin777]

There is a reasonable estimation was published around here on forum that Tianhe-IA could do about 850 Ghps . This means that there are quite a few miners in bitcoin network who have more than 1% of Tianhe-IA's bitcoin mining capacity and there are some people who have more than 5%. And I am not talking about pools. Interesting, isn't it? Particularly considering that Tianhe-IA cost is, reportedly, more than 20 million $ per year to run and 60 million $ to build and it requires 200 people to run it. (http://en.wikipedia.org/wiki/Tianhe-I)

Grassroots supercomputing! I love it!

[td]

Again, all that I am suggesting is that depending on how you look at it; bitcoin could be the largest grid supercomputer in the world and still could be considered very young. It definitely makes it seems that much stronger than just a few people in their mom's basement calculating coins on their old computer.

[xf2_org]

Is there a thread somewhere explaining how bitcoinwatch is converting between hashrate and petaflops? It seems to me that they are totally different operations; e.g. hashing is fixed-point, flops are floating point

Look at ArtForz' recent forum posts, describing the conversion.

[Stephen Gornick]

Look at ArtForz' recent forum posts, describing the conversion.

These perhaps?
  http://bitcointalk.org/index.php?topic=4408.msg64596#msg64596
  http://bitcointalk.org/index.php?topic=4408.msg64633#msg64633

[Stephen Gornick]

I think it was Raulo who calculated 850 Ghps of Tianhe-IA capacity based on number and type of GPU's and CPU's it has. It seems Raulo's approach is more accurate in this specific case.

Ah, ... here:
  http://bitcointalk.org/index.php?topic=4559.msg66931#msg66931

[bitcool]

I think it was Raulo who calculated 850 Ghps of Tianhe-IA capacity based on number and type of GPU's and CPU's it has. It seems Raulo's approach is more accurate in this specific case.

Back of the envelop calculation tells me if bitcoin network reaches 3000GHPS, its computational power will surpass seti@home and folding@home.
Are we there yet?

[bitcool]

I think it was Raulo who calculated 850 Ghps of Tianhe-IA capacity based on number and type of GPU's and CPU's it has. It seems Raulo's approach is more accurate in this specific case.

Back of the envelop calculation tells me if bitcoin network reaches 3000GHPS, its computational power will surpass seti@home and folding@home.
Are we there yet?

You probably give too much credit to whatever@home. Chances are that they are shrinking big time due to likely mass defection of most ATI GPUs to bitcoin.

I thought they were sub TFLOPS anyway and were left in the dust long time ago.

We are above 2 Thps right now.

well, I took the numbers from OP and no effort was made to verify them. if bitcoin has indeed become the most powerful cluster in the world, I'd expect much more media attention other than reddit/slashes

[jo.bo.co]

Its too bad the system wasn't designed to carry out useful calculations. I know people have talked about the security cons/ logistics/ ethics of this- but I think just from a publicity standpoint, this would give bitcoins so much attention, and would make gov'ts less likely to target the digital currency.

[marcus_of_augustus]

Its too bad the system wasn't designed to carry out useful calculations. I know people have talked about the security cons/ logistics/ ethics of this- but I think just from a publicity standpoint, this would give bitcoins so much attention, and would make gov'ts less likely to target the digital currency.

I personally find supporting and securing first ever viable convenient and honest currency on the planet being quite useful. More useful than, for example, search for aliens in other galaxies.

Precisely. There is probably no more important application for networked computers right now than securing the wealth of nations and providing reliable, fair price signalling information technology.

Just look at what happens when it is mismanaged by greedy bankers and incompetent politicians. It can become a crime on a vast scale that goes unpunished. We have not seen the last of the "financial crises" yet. The wealth has all been squandered already, spent on hookers and blow by Wall St. for the last 25 years most likely, the crises are only the aftermath to distribute the theft evenly to the plebes.

[againey]

Its too bad the system wasn't designed to carry out useful calculations. I know people have talked about the security cons/ logistics/ ethics of this- but I think just from a publicity standpoint, this would give bitcoins so much attention, and would make gov'ts less likely to target the digital currency.

If bitcoin depended upon calculations that served some alternative purpose in addition to securing the bitcoin network, then there is a very real likelihood that the strength of the security would be compromised.  The value of cryptographic hashing is that for all practical purposes it is so close to random that it cannot be reversed.  If the calculated output of confirming a block were an actual useful value for some other purpose, then except for maybe some very specific cases (and I'm not confident such cases even exist), that output would have significantly more of a pattern than a pure cryptographic hash.  And that pattern could be used to find exploits that could weaken the desirable features of bitcoin.  People find some awfully clever ways to defeat various cryptographic security schemes.  Even the slightest move away from focusing on pure security can open up some very subtle weaknesses in an algorithm.

Offhand, the only thing I can think of is the following (and it is highly dependent on details that I don't know, but others might):  Let's say that there is some unrelated application that needs to make a lot of hashes during its ordinary operation.  Whenever it needs to do a hash, it takes the data that it is going to hash, uses it as a nonce for an unconfirmed bitcoin block, and then hashes the bitcoin block.  If the nonce comes at the very beginning, then given most hashing functions, you should be able to grab the hash value out of the algorithm after just the nonce portion has been hashed, send that back to the application, and continue with the hashing until the full bitcoin block has been hashed.  Then compare the final hash to see if it passes the difficulty or not.  If so, you've confirmed a block.  Either way, you've hashed your unrelated data.

If the nonce does not come at the very beginning of the bitcoin block, then you can't do the above.  If the hash function in use cannot give a current hash at any arbitrary point in the process, then it won't work either.  And regardless, the above probably wouldn't generate nonces at a high enough rate to be anywhere close to competetive, even for applications that generate a lot of hashes.  (If you know of programs that routinely hash hundreds of millions of independent pieces of data per second for hours on end, then such application might be candidates for this scheme.)

And there might be other issues that prevent the above idea from being implemented; I'm certainly no expert, and I could have overlooked any number of complications.  But if there were no complications other than that the current bitcoin algorithm doesn't allow it for reasons such as the ones already considered, then that could be an opening for an alternative to bitcoin, one that has all of bitcoin's advantages (other than existing market share), plus an additional benefit:  the ability to do useful work while mining.

But again, I'm not optimistic.  I only take the time to write this idea as I thought it up in case it actually happens to have any merit, and someone goes off and does something useful with it.

[jo.bo.co]

Haha sorry- I didn't mean to turn this into another thread of this type. I do understand the security concerns, and I don't know of a way this could be technically implemented. Thanks for the detailed responses, though. I'm a newbie, and theres a lot of think about here.

My point really was just that, if a large portion of the world's scientific research (or something else equally important) were carried out on the quasi super computer that is the bitcoin network, bitcoin would have few enemies.

Offhand, the only thing I can think of is the following (and it is highly dependent on details that I don't know, but others might):  Let's say that there is some unrelated application that needs to make a lot of hashes during its ordinary operation.  Whenever it needs to do a hash, it takes the data that it is going to hash, uses it as a nonce for an unconfirmed bitcoin block, and then hashes the bitcoin block.  If the nonce comes at the very beginning, then given most hashing functions, you should be able to grab the hash value out of the algorithm after just the nonce portion has been hashed, send that back to the application, and continue with the hashing until the full bitcoin block has been hashed.  Then compare the final hash to see if it passes the difficulty or not.  If so, you've confirmed a block.  Either way, you've hashed your unrelated data.

Againey, I do find your idea interesting.  I have no idea if the bitcoin infrastructure could ever be modified to this extent even if this were deemed to be a practical way of doing things.

My instinct- and this is still vaque, so pardon me if this is totally unreasonable- would be to keep the difficulty at a fixed level that the community considers totally secure, and require GPUs to perform an increasing ratio of calculations for some outside application (be it figuring out how proteins fold or looking for earth-like planets) to each random block generated. Thus, as more and more people mine, each miner carries out more calculations for the application and does less mining, but the mining difficulty hasn't changed.

[marcus_of_augustus]

Haha sorry- I didn't mean to turn this into another thread of this type. I do understand the security concerns, and I don't know of a way this could be technically implemented. Thanks for the detailed responses, though.


My point really was just that, if a large portion of the world's scientific research (or something else equally important) were carried out on the quasi super computer that is the bitcoin network, bitcoin would have few enemies.

Or more probably the inevitable enemies it has would be PC'd into silence by public perception ... point taken though.

I'm kind of surprised that the scale of the network, > 20 Pflops, as a hardware achievement has not made it into the news, /. or similar. This alone is a feat given the speed with which it came together.