Does a miner choose which transactions to include in a block?

[res]

I heard miner who solved the block can decide to arbitrarily discard some or all transaction from the block. Is that right?
If so, I think someone with great resources can destroy whole bitcoin system anytime if they want it to be happened and there are some people who could actually do that.

Let's check some number first.
7970 GHz has 4300 FP32 GFLOPS and 1080 FP64 GFLOPS and the hashrate of GPU is 0.65GH/s. (which means 1 FP32 GFLOPS = 0.00015GH/s or 1 FP64 GFLOPS = 0.00060GH/s)
2600K has 50 GFLOPS and the hashrate of CPU is 0.02GH/s (which means 1 GFLOPS = 0.0004GH/s)

Judging by above data, 1 GFLOPS can roughly translated into 0.0004GH/s.
Todays network hashrate is 55,000GH/s, therefore 137,500,000 GFLOPS = 137,500 TFLOPS = 137 PFLOPS would be needed to achieve 55TH/s hashrate.
And US Government only has 17.59 + 16.32 + 8.162 + 1.257 + 1.11 + 1.054 + 1.042 (and so on) PFLOPS supercomputers.
So, Now we know that even the US government cannot beat the network's hashrate with current supercomputer especially considering the fact that we are on the verge of ASIC explosion.

However, If someone with great resources try to build their own version of ASIC, It could be very easy to beat the hashrate of entire network by wide margin.
I heard avalon use 110nm lithography, so if someone use 22nm lithography to build their own ASIC, ten fold increase of speed is easy.
If someone develop much better architecture and use 22nm lithography to fabricate much bigger chip, 50x speed bump would not be far fetched idea.

If so, the "improved" chip's hashrate will be 3,150GH/s. 1 million chip means 3,150PH/s which is 63,000 times greater than today's hashrate of entire network.
In other word, he now holds 99.998417% of network hashrate, so he can deny at least 99.998417% of transaction.
If that happen, bitcoin miners and users has two option. Option A, Try to beat him by increasing their hashrate. Option B, Accept the reality and leave.
I think 63,000 times difference cannot be overcome easily, so the most of miner will leave and bitcoin system will be destructed forever.

And i don't think 1 million ASIC's price will be unthinkablely expensive. I don't know nothing about ASIC, but if the die size of ASIC is 400mm^2, you can get 137 of them from one 300mm wafer.
You need minimum 7300 wafer to get one million working chip, and typical 300mm wafer cost $5k-ish, so it's $36.5M.
$36.5M is not little money to average joe, but US DoD's direct spending on Iraq War was at least $757,800M and $36.5M would mean practically nothing for them.

So, Once bitcoin become break certain critical point that could harm the real world economics, "bitcoin is enemy" kinda definition can be made by government.
and since sweep process is relatively easier than real war(find some ASIC gurus, make a $40M worth fabrication deal with intel), destruction will be inevitable.

Long story short, bitcoin will never gain wide market-acceptance because once they do, whole bitcoin system will be wiped into history.
Am i correct or not?

[christop]

FLOPS = FLoating-point Operations Per Second

Bitcoin does not use floating-point operations (only integer operations), so FLOPS performance is irrelevant.

[res]

FLOPS = FLoating-point Operations Per Second

Bitcoin does not use floating-point operations (only integer operations), so FLOPS performance is irrelevant.

Thank you for your correction. But it was never a main issue. The calculation was made to determine whether current supercomputer can beat the network or not.
And this doesn't matter because massive ASIC invasion is true problem.

BTW, Changing the numbers with integer performance is easy.
To get a hashrate of 1GH/s, you will need 5000 GIPS (for nvidia kepler gpu).
In order to get 55TH/s speed, 275 PIPS would be required.

Cray titan(no.1 supercomputer) has 18,688 Opteron 6274 and 18,688 Nvidia Tesla K20x(approx 2 TIPS).
It will get 37,376 TIPS from K20x and this is not even close to current hashrate of network.
(i didn't count for cpu because cpu speed doesn't really matter i think)

Same conclusion.

[christop]

It would be feasible for a large government to build ASICs to take over more than half of the network hash rate, but what would they gain from it? In any case the Bitcoin economy would likely collapse as people lose trust in it and exchange their Bitcoins for other currencies, and the government would be out however much they put into destroying it.

Other cryptocurrencies that may be better at resisting this type of attack than Bitcoin could pick up where it left off. Litecoin, for example, is designed to be difficult to implement on an ASIC.

[res]

It would be feasible for a large government to build ASICs to take over more than half of the network hash rate,

What I'm trying to say is it's not just feasible, it can be done very easily.

but what would they gain from it?

They would gain nothing for now, only because bitcoin is nothing for now.
However, If Bitcoin continues to grow, Bitcoin will be something at some point, something that actually could do some damage to real world economy.

In any case the Bitcoin economy would likely collapse

If people lose faith on Bitcoin, It will fall all by itself. = DOOMED
However, If people don't lose faith on Bitcoin, government will kill it at some point. = DOOMED

That's my point that there is no hope.

Other cryptocurrencies that may be better at resisting this type of attack than Bitcoin could pick up where it left off. Litecoin, for example, is designed to be difficult to implement on an ASIC.

I don't know nothing about other crypto-currencies. But I can guess some.
Other currencies which give same freedom(discard some or all transaction from the block that he cracked) to miner are all vulnerable and the total destruction might be easy.

If there is no FPGA or ASIC(no faster way to mine), one who want to destroy crypto-currency do not need to build a special machine because regular supercomputer can do the job.
Top notch supercomputers usually have 10~100k of CPUs+GPUs or 200+k of CPUs. So, If someone managed to rent multiple top-notch supercomputers (which is easily possible for some government),
He will have more than a million CPU and a few hundreds of thousands of GPU. There is no way to match these kind of processing power.

For example, before ASIC hit the network, total network hashrate remain 10-15TH/s range. It can be achieved with 16~25k of 7970.
In other words, If other currency's managed to attract same number of miner as bitcoin, their combined hashrate cannot stand a chance in the battle with a group of supercomputers.