Claymore’s bins are protected with an off-the-shelf binary encrypter. I found it difficult to find a cracking tutorial for the specific software and version he used, but I’m sure that if you spend a little bit on it by hiring a pro cracker, it should still be relatively easy.
What impresses me most is the assumption that you are the best OpenCL programmers in the community by 30% - 10,000%… That is one way to look at it and maybe it will work out but there are far too many people with the right skillset in this space that would release one just to make a point, I don’t know for sure but I suspect it will come…
30% is the comparison of our code to the other optimized GPU miners whose performance has been published or hinted at. 10,000% is the comparison to the only open source implementations that are currently out there (which are CPU based).
We’re not particularly good OpenCL programmers, but we had a few ideas during development that seem to have been good ones.
Genoil’s miner is around 170 ms per run, last I heard (roughly 11 Sol/s). Tromp’s miner gets around 18 Sol/s on hardware that’s about 1.7x as expensive as the R9 290s that we get 31 Sol/s on. Zeropond has kept their current performance a secret, but their first published values were 1.2 Sol/s. Et cetera. But none of that software is public. The best public option is about 0.06 Sol/s per core, or maybe 0.2 Sol/s per CPU. We are 150x faster than that.
zawy, can you check the second test, in which we mined 52 blocks in about 2 minutes on 5 GPUs? There may have been a problem during the 1st multi-GPU test, and I think the difficulty was higher then anyway.
The difficulty should have been rising 16% per block after the 3rd block, so you would have to break it out into a sum, using each time-to-solve and each D:
A diff 1 block has the same difficulty as the genesis block. The genesis block for beta1 testnet had a difficulty target of 0x0f0f0f000000… The chance that any given hash is smaller than that number would be float(0x0f0f0f00…) / float(0xffffffff…) = 0.05882. The expected number of hashes or solutions to find a diff1 block would thus be 1/0.05882 = 17.
At diff 77 (which is what I think the difficulty was during the 7 block streak), it would take 77*17 = 1309 hashes on average in order to find a block. Finding seven blocks would take an average of 9163 hashes. Over 123 seconds, that translates to about 74.5 H/s (or Sol/s).
At diff 30 (the average for the 52 block run), it would take 30*17 = 510 hashes per block. For 52 blocks, that would be 26520 hashes. Over roughly two minutes (I don’t think I ever looked up the timestamps), that would be around 220 H/s. We should have only been getting 150-180 H/s, so I’m probably remembering one or more of those numbers incorrectly.
It’s not per card. It’s per rig. Stratum adjusts the difficulty so that you send a fixed number of shares per minute regardless of your hashrate.
We aren’t actually using stratum right now, though. We’re using a combination of getblocktemplate, submitblock, and a custom binary UDP protocol that we threw together because it sounded easy and fun. We’re adding stratum support for our cloud hashing customers, but our internal solo mining won’t use it.
The difficulty should have been increasing 16% per block after 3 blocks (or at least after 9 blocks), so by block 28 it should have been D=30 x 1.16^(28-9) = 503. For some reason it did not increase much. Why is that?
Your 17 H/block at D=1 means 17/150 = 0.113 H/s for D=1, so the D/10 people have said to use should be D/8.823.
Your hash rate:
Starting at block 17029, you got 32 blocks in 52 seconds at an average difficulty of 27.6. Your H/s is measured to be:
I de-selected the last 10 data points because it seemed to get a lot slower, down to 100 H/s. For the full 42 points I can see, it’s 250 H/s.
The 2 time-stamps following the series were about 600 seconds slower than the expected average expected which is not an anomaly, but it’s enough that if you had manipulated the 600 seconds, your actual hash rate could have been as low as 50 H/s
Here’s the data, seconds and difficulty starting with 17029
