Topic: hashespersec (Read 3742 times)

Well, it may be real. I don't know if a particular node accept block if there are missing transactions in it which are certainly known by this node. If so, transactions can't be cancelled and an attacker could generate as many blocks as he wants but chain won't grow for anybody except him. But the weaknesses page states that he can. Maybe it should be discussed. Also, I'm interested in how can such powerful attacker reverse his transactions? I suppose it's already sent and other nodes got it. If there were no blocks with this transaction inside then yes, he can send another one or just don't include tx in the generated block. But who will trust such unconfirmed txs? And if he issue some blocks with his tx in first of them how can he later cancel them all? Other nodes already got this sequence. He'd regenerate that latest blocks (fork them like there was no tx) while he should continue generating valid blocks for the current branch or somebody in the rest generate first. And that false branch should grow even faster than main. It will require more and more power.

The client would always choose the longest branch if there are two with perfectly valid blocks but different txs? Am I right? What happens with txs in the shorter branch? If they're transferred what if a conflict detected (for example different number of coins sent from the same address)? Which one will client choose?

How about 20 Radeon 5970s and OpenCL client? 40?

~$700 * 20 == $14,000
Maybe $400 for rest of computer parts * 20 = $8,000

Only $22,000 + cost of electricity + closed OpenCL client to completely pwn the network.

ArtForz?  Do it!!!  Hurry, before an open source client becomes available!
Although, it is unlikely he would do it because it would probably cause the value of Bitcoin to fall to $0.00000001.
Instead he is pursuing most profitable method while preserving value of Bitcoins. ^_^

Let's do some fun math. I have E6750 CPU with 5425 BogoMips on each core. I do 2200 khash/sec so it's 2200000/(5425*2) ~= 203 hash/Mip. Now look at the difficulty; it's 712.88486455 now. So the network does 712.88486455*2^32 ~= 3061817179056 hashes per 10 minutes. Now we can estimate its power in Mips, 3061817179056/203 ~= 15082843247 Mips ~= 15 Pips. Here is a world most powerful supercomputer: http://www.top500.org/system/performance/10184 It has only 2 Pips in peak and maximum 1.7 Pips (don't really get the difference between max and peak, maybe max is an average maximum and peak is a one time performance shot). I don't remember what's the threat of the overpowering of the network but it looks like the attacker will need about 9 such computers for successful overpowering. I guess it's highly ineffective for such tasks and want to congratulate us all for participating in such a big and powerful network (although we do pretty absurd job in mathematical sense).

Here's my stock client results:

PIII-M 1066Mhz win2k ~250.000
Atom N270 1.6Ghz winXP ~400.000
x2 5000+* 1.6Ghz winXP ~1.400.000
x2 BE-2350 2.1Ghz Win7x64 ~1.800.000
x2 5000+* 2.6Ghz winXP ~2.100.000
x3 720 2.8Ghz winXPx64 ~3.800.000

Olipro's 64bit-version squeezes a bit more juice out, last time i checked it's been around ~4M on my triple-core,
not to mention gpu-possibilities.
At current difficulty I would expect about a month at least to find 1 block/get free coins while generating on all those 5 machines 24/7. I might get real lucky and find 2 in that time though.



*black edition, usually extremly undervolted and set fix to 1.6Ghz, default clock 2.6Ghz, runs up to 3.2Ghz aircooled stable without any probs, results in noisier fans though, i kinda like it cool and quiet.

So you are running a PIII?

Just kidding, but that is kind of slow. I do ~1400000 on my mac using the CPU (that's an i5), around 6200000 using my cuda enabled GPU and custom code and on a [email protected] OC'd to 3.0, I crunch ~5000000.

I'm interested it know where I am in terms of how much CPU I'm donating.  What are other people's hashespersec number?  I'm at 350886 just now