[zumth]

Top500 computers could have done this sort of attack when CPU and/or GPU mining was all the hype, but now with FPGAs and ASICs, the hardware required to have an impact is too specialized for any sort of general-purpose computing.

[mrb]

That is correct. All of the Top500 computers combined, all 500 of them, would not be sufficient to perform a majority attack on the Bitcoin block chain.

The global network hash rate is currently 35 Thash/s.

A typical GPU found in a supercomputer (Nvidia Tesla K10/K20/C20xx) does roughly 100-200 Mhash/s.

A typical modern CPU core does roughly 1-2 Mhash/s.

You would either need 175k-350k GPUs, or 17.5M-35M CPU cores to attack Bitcoin. If you look at the Top500 list, they have maybe a few tens of thousands of GPUs combined, and about 10M CPU cores (watch out, the core counts in the list combines CPUs and GPUs, I found this out when making sense of the numbers years ago: http://blog.zorinaq.com/?e=14 )

Plus the network hash rate is predicted to increase by at least +20 Thash/s (to 55 Thash/s total) in the next month due to Avalon finishing delivery of their first batch of ASICs. Bottom line, no, all Top500 computers combined could not attack Bitcoin.

[glomph]

Do we not think that if the NSA were so inclined they could fire up a few thousand ASIC like computers?

(Just curious)

[mrb]

Yes the NSA can, and has, developed custom ASIC chips. They currently do it via http://www.trustedfoundryprogram.org However I doubt the NSA will ever care about Bitcoin (their primary mission is intelligence, not breaking decentralized currencies). And by the time they do (if they do) it might be too late for them to be able to attack it (the network hash rate might have grown too much, making such an attack too costly).

[waps]

If anything Bitcoin provides a full transaction record with transaction participant IDs (even if those IDs are not names). It probably helps intelligence gathering.

[glomph]

Thanks.