08:20:55 It looks like recent Windows update removes xmrig from exceptions and it has to be manually added again 08:21:12 This and github actions fix probably contributed to hashrate drop from 2600+ to 2200 MH/s 09:27:19 hey guys, anyone have benchmarks for RX 6800 XT? 09:27:44 or any Radeon with infinity cache? 11:20:54 pointless to run on GPUs 12:08:20 I know but I wonder if the infinity cache helps over nvidia’s GPUs 12:08:37 Not looking for profit just academic curiosity 13:09:12 I haven't seen any benchmarks of RDNA2 GPUs 18:29:20 looking at https://xmrchain.net/randomx/2346463, each program has ~30 branches 18:30:05 if it's 1/256 a branch will be taken, that gives you a 1-(255/256)^30 = 10% chance of any branch being taken per program 18:30:22 .9^8= .40 18:30:25 per iteration 18:30:27 =60% chance of failure 18:30:31 each program runs 2048 iterations 18:30:39 as long as you can speed up by more than 2.5x, should be worth it? 18:31:04 each branch is taken 8 times per program, on average 18:31:45 sech1: do the programs just loop? 18:31:53 yes 18:33:00 how can they run 8 times in expectation? 18:33:10 10% of any branch for each execution 18:33:18 how many times over are programs looped? 18:36:33 2048 18:36:43 each program 2048 times 18:37:12 > RANDOMX_PROGRAM_ITERATIONS 2048 18:37:14 well there you go 19:02:23 how dastardly, hiding such info like that 19:10:59 Yeah, imagine you're talking to someone about the movie you're going to see, and instead of saying, say, "what time is the movie tonight", you say "what time is the ENTERTAINMENT_OBJECT ENTERTAINMENT_SUBJECT, please refer to my earlier definitions before replying, thank you". 19:11:26 "You're so not gonna go see the movie with us", is gonna be the reply.