Hacker News Reader

Comment by CyberDildonics | Hacker News Reader

How? If the proof of work only takes a few seconds, what prevents someone from spamming a new transaction every few seconds?

View on HN

I am very curious about this as well. In this example [0] someone was able to precompute 30k transactions in 5 hours with a single 1070. A mining farm with 100 video cards (capital cost of $40k) could generate a sustained load of 10k transactions per second. With precomputation, you could easily do an order of magnitude more damage. All of these transactions are stored permanently on the ledger as far as I can tell.

Nanoblocks does serve some purpose in my mind, but it has major issues to solve, and in no way is a catch-all solution for the scaling problems that Bitcoin has.

[0] https://hackernoon.com/stress-testing-the-raiblocks-network-...

Your calculation is wrong. One 1070 is able to produce about 1.6 txs, so 100 x 1070 are able to produce 160 txs. For a full transaction you need to do two time POW (send and receive), so if you use the same cards to receive the tx also you are only able to produce 80 txs. Right now I do not think you would harm the network up to about 10k - 50k txs (a simple laptop is able to monitor about 1k txs). So you may want to target at least 10k txs, therefore you would need more than 10k graphic cards at a cost of about $500 each at least, this sums up to $5,000,000 and as I said it is not clear that this would harm the net in any substantial way.

My calculations are based on the numbers from the article I linked to. I'm guessing they might have changed the difficulty since that blog post was written?

Regardless though, I can precompute the PoW for each transaction, so I could preprocess for a month to get a higher tps.

It was not changed yet, 30,000tx/18,000seconds=1.6txs

You're right. I did my math wrong.

Transaction fees.

I thought the whole point is free (or very low) transaction fees?

Reply on news.ycombinator.com