How do you set the miner to do this? ...
I am quite interested.
#crysx
If you wish to enter the competition, below is what I wrote in #competition in Discord.
If you want to take part of this competition, the important parameters in the Config file are
- Threads, BenchmarkDifficulty, ConstellationType (set how many threads you want to use to find a new record and your goal: which type of constellation and the length of the prime numbers)
- PrimeTableLimit: higher is better until a given point; the more the Difficulty, the higher the PTL should be. But with over 16 Threads or very low Difficulties, you may need to reduce this value a lot to avoid CPU Underuse
- PrimorialNumber: higher is better but the primorial must be smaller than the prime numbers you are looking for by some margin (~65 binary digits), otherwise the search will not work or loop
- PrimorialOffsets: pick numbers compatible with your Constellation Type (normally, you can simply use the first constellations' base primes)
- SieveWorkers: increase if you are affected by the CPU Underuse, lower is better
- SieveBits: higher is better; 2^SieveBits times SieveWorkers should be a number of bits that fits your L3 cache
Developers are welcomed in order to simplify the tuning by improving rieMiner (especially by definitively fixing the CPU Underuse Bug).
Currently, the easiest record to beat is still the 8-tuples. One can beat my record in about 25 days with a 3700X using a Difficulty of 1075 (my record is 1025).
The 7-tuples record can be beaten in about 50 days at Difficulty 1800 (current record is at ~1750)
For 6-tuples, it is much harder and it would take 1.5 year to find a one at Difficulty 3500 (current record is at ~3445)
For 5-tuples, it would take ~250 days to find one at Difficulty 5400 (current record is at ~5335)
Starting from 9-tuples, the CPU Underuse Bug strikes back, forcing you to use very low PTLs. It could still find a 9-tuple at Diffitculty in about 30 days at Difficulty 700 (current record is at ~675)
For 10-tuples, it would take ~40 days to find one at Difficulty 480 (current record is at ~455)
For 11-tuples, it would take ~150 days to find one at Difficulty 380 (current record is at ~360)
rieMiner is not really viable for 4-tuples and shorter and 12-tuples and longer as it would take about a decade or more to find anything...
There is a small room if you want to beat more easily records, though records by small margin will be less interesting. And of course if you have ThreadRippers or other computer beasts, you could beat records in much less time. But there is also a lot of variance so it might take much more time than estimated, or in contrary you might be lucky and find a tuple much faster than expected.
Good luck!
I will add that it would be better to try to beat records that are not already owned by someone that used rieMiner, I suggest to try to beat the 9-tuples. Vidarn told me that he is trying to beat the 6-tuples record (good luck!).
If many records were broken with rieMiner, it would surely attract more interest in it and Riecoin.
Example of rieMiner.conf for 9-tuples that works well for me (3700X), adapt it to your system
Mode = Benchmark
Threads = 16
PrimeTableLimit = 25165824
BenchmarkDifficulty = 700
Benchmark2tupleCountLimit = 0
TupleLengthMin = 8
PrimorialNumber = 89
ConstellationType = 0, 2, 4, 2, 4, 6, 2, 6, 4
PrimorialOffsets = 182403491, 226449521, 910935911, 1042090781, 1459270271, 2843348351, 6394117181, 6765896981, 8247812381, 8750853101, 11076719651, 12850665671, 17140322651, 22784826131, 24816950771, 33081664151
SieveWorkers = 10
SieveBits = 24
For example if you had a 3900X, you would use 24 Threads and probably have to lower the PrimeTableLimit (or increase the SieveWorkers)
The tuning is really Trial and Error in order to get the best efficiency.
The PrimorialOffsets can be filled with the numbers found
here (for other tuples, just search for things like "septuplet", "octuplet",... in the website, be sure to pick the right ones)
Don't discard the 8-tuples (or (k - 1)-tuples) as you can submit them as well and get easy 5th place.