Topic: [ANN][RIC] Riecoin: constellations POW *CPU* HARD FORK successful, world record - page 132. (Read 685207 times)

Awesome! Thanks Gatra! Cryptoworld is a bit volatile these days but Riecoin will surely come out on top! Any update this week btw?

Thank you!

I am quite happy with my flavor of b14.  I do some fiddling with prime numbers occasionally  and spent a lot of my spare time this summer trying to figure out the sieve. (Part 2 of Fast Prime Cluster Search could have saved some time, but it was an interesting exercise tracking the ingenious code!)

I left out all the xpt and Riecoin block handling, the starting target was simply 2^difficulty,  with a smaller random number added.  With no Riecoin restrictions, I could use larger primorials. My only other change  to the sieve was  when I check the five remaining tuplet members for primality, I break the loop at the first non-prime. I know I won't get a full  6-tuplet and searching for a 4 or 5-chain with a "hole" in it is of no use. The mpz_powms in the Fermat tests are quite time consuming!

BTW, while I was thinking about where to go next *) the program was idling away here the other day and (almost accidentally!) found a 600-digit 6-tuplet. I'll stop there and leave the rest to the Riecoin Superblock.

Regards, Vidar


*)  Fortunate numbers at present,  definitely not Riecoin stuff, probably both impossible and useless, but fun!

Could consider as well adding post-fix to the difficulty numbers with KB, MB, GB, TB, PB, etc (as in Kilobyte, Megabyte, Gigabyte, Terabyte, Petabyte) of expected storage size needed on the prime number - it can represents all or just a single prime number itself at the given difficulty found by Riecoin project thus so far. The post-fix serve as indication on the “frontier” digits-size of prime numbers the project is presently venturing.

Great!

But maybe for the sake of obsessive compulsive disorder, would it be possible to tweak the difficulty numbers to have rewards be 45 & 1485 ? (ie. round numbers)
Or even 45 & 1500 (with extra superblock difficulty), wouldn't affect network velocity much, but would be neater to look at and announce

Difficulty is a number with many decimals anyway, so extra tweaking on the difficulty numbers will have no psychological impact, but having weird reward figures might ^_^

Nice!

I still mine testnet on a semi regular basis.  If you need someone else to help test (even if it is only testnet mining) let me know.

Regards,

PS.  I've over 24k coins on testnet, let me know if anyone wants one ;-)

--
bsunau7

don't worry, I read some people had problems compiling my fork so I'll review that and try to generate all the binaries.


This is the final implementation that I have for the superblocks:
Difficulty adjustments were performed every 288 blocks which should take 12 hs, so once every 14 adjustments (ie once a week) we will have a special interval. In these special intervals the difficulty will be adjusted a little bit down for all blocks except one which will be the superblock. The expected behavior is that 287 blocks will be shorter (136 seconds instead of 150), lending 14 seconds each to the superblock which will in total accumulate almost 70 minutes. This is the most elegant solution I could find since every interval will still have 288 blocks and be targeted to last 12hs.
So, if in a few days you get blocks with a reward of 45,33333333 instead of 50, then it is expected behavior. Don't complain: your block had less reward because it was easier to find, it's still fair. And if you get a block with 1389,33333333 RIC, then congratulations: you solved the superblock.

This is implemented, under testing, and working fine so far. Now what remains is to decide the block number for the hard fork (around 135000?), generating all the binaries and issuing alerts for everyone to upgrade.

Regards!
Gatra

Anything new this week Gatra?

Awesome!  Congratulations - and thanks for doing it.  Note that I fixed a bug in the b15 version that was breaking things under windows (so it's now called b16) so you might have an easier time getting it running on VS now.

Urp - I've fallen behind on time to be able to do the code integration for stratum.  I apologize.  If someone wants to fork the miner and do it, you've got my blessing...  I still hope to do it if nobody else can, but I've got a bunch of day job things between.

The tdp is 80w, and I tested the board power consumption as a whole(include CPU + board + memory, booting from lan), it's about 85w.

My cpu is actually Xeon E3 1230v3, it's mining power && power cost is very close to I7:)

Is that 80W from intel's TDP numbers?  It is very hard to get CPU only numbers as Samsung don't release power specs on their chips (actually, they list very little aside from marketing fluff).

System numbers are much easier.  The board as a whole (all draws accounted for; memory, CPU, USB, ethernet, eMMC, stock fan, power converter loss etc..) under CPU load (no GPU) is about 6W.  Now for the numbers...

I am running @ ~ .7 coins/day or ~.11 coins/watt-day.

To get a feel for total system draw my 80W Xeon is drawing ~200W when idle (bulk of my power draw isn't CPU).

Regards,

--
bsunau7

To be accurate: it runs at 80w, and generates about 35 coins per day in the last week.

Do you have some numbers on the mining efficiency/power costs? Like coins/watt under some difficulty?

As to my test, i7 maybe the most power saving cpu mining RIC, it runs at 80w, and generates about 35 coins per day.

Hi All,

Another slight (~5%) improvement.  Changes can be seen in bitbucket:

https://bitbucket.org/bsunau7/riecoin-armv7/commits/6c6d25a333e05a5f12910bf124a9c6ef76d40f99

Binaries matching this code from:

https://mega.co.nz/#F!6Y8ywRoZ!5ogMJEc2Teq5ckt9WClNpg

Regards,

--
bsunau7

Power low single digit watts, I believe it is one of the few CPU's which can turn a profit (more coin than electricity costs) mining this coin.  I just don't have the time to make those improvements (thus this release) needed to reach break even.

The CPU I use is the same as in the Galaxy Note 2 phone.  Also there are some nice DSP-ish features which could be leveraged.

Regards,

--
bsunau7

Glad to see you working on this bsunau7!
I have a question: what's the advantage to mine on arm?

Hi All,

I've had very little time recently and it looks like the next 3-4 months will afford me even less time.  As such I've released the code for my ARMv7 miner to the public.  The link is:

https://bitbucket.org/bsunau7/riecoin-armv7

There are lots of ideas I would like to try out, some of which are:

• Seeing if the '0' heavy primes I generate can be used to speed up modular exponentiation (initial tests got within 5% of GMP performance without using montgomery)
• fast_recip might be faster than the fastmod code, making more use of fast_recip might get performance gains.
• It might be quicker to calculate the reciprocals inline as opposed to fetching from memory.
• Interleaving primes/reciprocals/offsets should be more efficient from a memory access viewpoint, worth investigating.
• I've only optimized fast_recip for memory access, should be some wins in the fastmod code.

Hope this is of interest!

Regards,

--
bsunau7

Any updates this week Gatra?

Wow, 11 cores running for 1 month, that needs patience man, cong!

Hi,
A few days ago I submitted to Anthony Forbes a 597-digit 6-tuplet and had it verified as a new world record:

http://anthony.d.forbes.googlepages.com/ktuplets.htm

This number is only a few digits above the previous record, and it should be quite possible for the superblock project to beat it!

The program used was based on the b14 version of dga's xptminer (which was the latest version I was able to compile using Windows and Visual Studio) and ran on 11 cores for about a month before it suddenly surprised me with a find. I'll give some further details later.

The 6-tuplet is

1268360451482944051031693571608646702803337857417059923179003396370803464447945 9213486050679397935104622682138475913438101505200313946821621485384179510473697 9433691754086534863323734928743959323221865529501237655633762351765417920461874 1258860678645822691792395092270199088414732238122470003707838866453023159582123 7351808046531772170786266328462890225388980932631840153394721839158254583153068 7931209705774579874734193952405536748117312631969874545797052973507394025122061 8770487542157843118646752735712130590319004160226700306792249093957168851751545 30232372583285744387264936607664813582967597
+d, d = 0, 4, 6, 10, 12, 16

or

4921035090333221245281056600469203091910280140820896181940314619657756374547612 4680711202733554189605614783506825475691404888752144184576829642013820327715134 5176983584805931408105542588834038851652656017513154942170087395616822290168504 6046380856370538924059371410139041704192384855299316496758299832292941057155474 6087497007967825846220015325654831770578435398632279854720 * 547# + 8061997 +d, d = 0, 4, 6, 10, 12, 16

Regards,
Vidar