131–136 of 136 posts
I'm at about ~22GH/s per Xilinx VU9P UltraScale+ FPGA - at least 40GH/s is possible (for around 250W/device.) The nonce alphabet being limited makes the whole thing quite a bit more expensive.
Anyone got a guestimate as to the energy used to make that leaderboard?
Running this for 1.5 hours would consume 1KWh of energy, and compute 1.134e14 hashes (or 2^46.7 if you prefer).
Assuming SHA-256 hashes are essentially normally distributed, that means that finding a hash with 13 leading hex zeroes should be a 1-in-(16^13) aka 1-in-(2^52) odds.
This gives 2^(52-46.7) or 39.7KWh on average to generate a single 13-zero hash, and ~16x that (635KWh) for a 14-zero hash. I got very lucky finding a 14-zero hash quickly, vastly earlier than average expectation, so my energy usage is quite a bit lower than that.
Earlier quoted context omitted.
I couldn't resist, so here's the GPU hasher: https://gist.github.com/grishka/c1be1c035f39564debfd4b195bdf... It gets around 232.5 MH/s on the M1 Max. There's also an optimization that exploits the fact that SHA256 is incremental, so the state for the common prefix could be pre-computed and then copied and only updated with the variable part. This yields around 30% performance boost.
I'm able to get 250MH/s with this. M1 Pro
I'm sitting on just under 40MH/s on an M1 Pro with a multithreaded Rust implementation I whipped up. I doubt I can go much faster without using the GPU.
Also, Oi tudo bem @quirino?!? Will you be posting the number of competitors at any point? I would love to see how good my 24 is in the grand scheme!
Sorry for the late reply.
I've added a "/data" endpoint where you can get the full database in json format.
I'm at about ~22GH/s per Xilinx VU9P UltraScale+ FPGA - at least 40GH/s is possible (for around 250W/device.) The nonce alphabet being limited makes the whole thing quite a bit more expensive.
That's a beefy FPGA! Wish I had access to one to revive my 2009 era FPGA coursework knowledge but it seems the dev boards start at 5 figures which is too rich for a side project.