07:37:03 exploring crypto twitter is fun: https://twitter.com/simulxxx/status/1649618696510357505 10:41:14 lol 10:43:26 https://twitter.com/hyc_symas/status/1649872617803841536 10:44:07 https://twitter.com/hyc_symas/status/1649875101871923200 10:48:28 2m numbers sounds off 10:49:41 buuuuut, i need to find my maths. I think the ratio is even worse 12:31:57 I'd figure it's hard to estimate, you just have to settle on a reasonable average CPU hashrate 12:59:47 tevador: the 1 minute is to generate a suitable v1 which will lead to a valid v2 hash, but what would prevent an attacker from reusing those v1/v2 values over and over again? Unless nodes keep track of valid v1/v2 pairs whose v1s are not valid RandomX hashes the nodes will still perform the heavy RandomX computation right? 13:02:12 they can't reuse old hashes because https://github.com/tevador/RandomX/pull/265/files#diff-495d78c7bd13cc283d13e085a654aa6627d52eb1258e6e6cc14aadb64e559174R410 13:02:33 second Blake2b hash also uses original input which is new every time 13:16:02 Oh ok, should have read the code first! Great! 14:02:17 The number of cpu cores on xmr is probably even higher, around ~5M. But so is the number of asics (2-10M, depending on the generation) 14:02:42 But I would not consider cpu cores and asic devices directly comparable 14:03:05 maybe entire cpus vs asic devices 14:03:10 yes, only Ryzens (starting from Ryzen 3000) can do 1 kh/s per core (running 2 threads) 14:03:28 most Intel CPUs and older Ryzens (1000, 2000 series) can do max 600-700 h/s per core 14:04:38 But Bitcoin's security budget is two orders of magnitude higher: 296 $/s vs 0.8 $/s for xmr 14:04:54 (right now) 14:07:19 energy budget (kWh spent) per $ of emission is higher in xmr 14:07:27 Also, the top tier asic (Bitmain S19 XP and its hydro-cooled variant) should earn ~0.14 $/kWh (going off of the nominal speed and power consumption listed on their side) 14:07:29 due to "free electricity" miners 14:07:56 Yeah, Monero gets a bit more bang for the buck thanks to the lower entry barrier 14:09:25 But the flipside of that is that it's harder for the miners to break even, due to the lower profitability 14:10:30 Right now you'd need 130 H/s/W on randomx to match the 0.14 $/kWh figure 14:11:40 my 7950X can go higher than that if tuned for efficiency 14:11:44 I tested when I bought it 14:12:10 Just the cpu, or at the wall? 14:25:31 16m epyc 7763's to use the equivilent amount of power to that of 845k s19 xp hybrid (18x more devices) 14:25:31 and i dont know current hardware prices, but old numbers had $8600 for epyc 7763 and $19.8k for s19 xp hybrid. 14:25:31 Usig these numbers, the hardware costs for epyc are 18x more expensive than the hw for thr s19 to use use the same mount of power 14:27:44 My script says 1.367M S19 XP Hyd. Asics 14:27:46 That last part might be wrong, rechecking what i did here 14:29:32 these are old nunbers, 14:29:32 "341.66eh" 14:29:32 my numbers were from 215eh 14:30:09 Yeah that checks out 14:31:23 1367/341*215=860 14:35:13 I think last # (hardware cost) is supposed to be 7.6x 14:36:06 merope at the wall 14:36:43 [2022-09-30T18:32:49.453Z] <@sech1> Ryzen 7950X running at 0.85V, 4.3 GHz all core: 18143 h/s, 140 watts at the wall = 129.6 h/J 14:36:54 This was before I tuned memory timings 14:37:14 so it will be more now, and I can save 10 more watts by removing GPU 14:38:01 it can achieve 145-150 h/J 14:38:18 Sweet! 16:52:44 feel free to reply to my tweet with better numbers 17:59:47 https://poly.cybre.city/notice/AUvOlqlGdkoJNNYLcu for Ryzen 7000 X3D users, 🔥 18:00:17 "ignored AMD's voltage recommendations to stand out from the competition and ended up overheating CPUs and *physically wrapping them*." 18:52:56 also maybe interesting for here https://chipsandcheese.com/2023/04/23/amds-7950x3d-zen-4-gets-vcache/