Ellieo
Member
Offline
Activity: 60
Merit: 10


October 23, 2014, 09:06:05 AM 

Hey kids. So why is it that I'm always embarrassed about asking simple questions? Don't answer that, it's a simple question. I just saw the latest 64 builds in all their variety and wanted to make sure I was running the version optimized for the latest AMD architecture. Obviously the "xptminer_k10_cphrg1" version is for K10, but is there specifically one for the FX8350 or bulldozer/piledriver architecture? Is there a limit to the s parameter, either in coding or effectiveness. I've tried up to s 1600000000 but haven't noticed any improvement however I'm not too familiar with interpreting the output from the miner program. Would someone be so kind as to give me a quick rundown on what's being shown here? How do these numbers compare? I'm running 8 cores (7+1) and a pretty vanilla 4.8Ghz overclock, not pushing ram or bus speeds. Thanks in advance. Getting as big as possible on 4Ch/s if you are comparing versions of xptminer or optimizing your mining hardware. It takes awhile for 4Ch/s to stabilize before you could come to any conclusion assuming the difficulty didn't vary much while 4Ch/s stabilized. The 4Ch/s varies indirectly with difficulty. I've tried various setting on s but didn't notice any difference in performance on my system limited to 8GB only. I think the "s" stands for sieve size on RAM for mining the sextuplet prime constellation. Larger sieve size definitely gives better chance of finding the prime.







GET 25 FREE SPINS AT REGISTRATION GET 100% BONUS ON FIRST DEPOSIT PLAY NOW



Advertised sites are not endorsed by the Bitcoin Forum. They may be unsafe, untrustworthy, or illegal in your jurisdiction. Advertise here.


bsunau7
Member
Offline
Activity: 114
Merit: 10


October 26, 2014, 05:30:57 AM 

Anyone know anything about the difficulty dip to 1280 (and the rebound to 1409)?
Ypool's pool speed doubled without much of a worker increase... Just curious.
Regards,
 bsunau7




vancsj
Member
Offline
Activity: 86
Merit: 10


October 26, 2014, 05:50:02 AM 

Anyone know anything about the difficulty dip to 1280 (and the rebound to 1409)?
Ypool's pool speed doubled without much of a worker increase... Just curious.
Regards,
 bsunau7
Ypool is down for a few hours and the diff went down, then it bounced back after ypool was back.

RIC solo mining with XPT miner @ zjuer.net:10034



Ellieo
Member
Offline
Activity: 60
Merit: 10


October 26, 2014, 09:21:38 AM 

Anyone know anything about the difficulty dip to 1280 (and the rebound to 1409)?
Ypool's pool speed doubled without much of a worker increase... Just curious.
Regards,
 bsunau7
Ypool is down for a few hours and the diff went down, then it bounced back after ypool was back. Previously when Ypool was down no mining or block generation was possible at all. The 1280 dive and 1409 rebound phenomenon can be attributed (just a hypothesis) to sudden drop of large miners and subsequently recovered in short period. Current total of Riecoin miners on Ypool is about 5320. There seem to be some new confidence regained among miners that are usually less than current total when the difficulty ranged between 1400 ~ 1450.




Ellieo
Member
Offline
Activity: 60
Merit: 10


October 26, 2014, 12:09:42 PM 

Current total of Riecoin miners on Ypool has now increased to about 5400. Perhaps it's more towards divided hash power among the increasing miners hence the difficulty now is still maintained at 1409.




fairglu
Legendary
Offline
Activity: 1096
Merit: 1025


October 27, 2014, 08:22:59 AM 

Any chance of the ARM miner working on upcoming online.net ARMbased dedicated servers? ( http://labs.online.net/tryit) Pricing have not been announced yet, but since they intend to compete with VPS clouds, I assume it'll be cheap.




bsunau7
Member
Offline
Activity: 114
Merit: 10


October 27, 2014, 01:12:07 PM 

Any chance of the ARM miner working on upcoming online.net ARMbased dedicated servers? ( http://labs.online.net/tryit) Pricing have not been announced yet, but since they intend to compete with VPS clouds, I assume it'll be cheap. Sorry for the numerous disjointed threads in my reply... I didn't know about them (well know of Iliad from their Tmobile bid)! I've been expecting a few ARM based VPS' to pop up, but I was expecting them to be ARMv8 based with a XEN/KVM virtualization layer over the top. My guess is they will be using the low power to ensure good margins while undercutting linode/AWS/DO/rackspace prices esp when you price up a 4 CPU 2GB VPS. I guess I am saying I don't expect them to be ultra cheap. With a quick look, 1.2GHz CPU and a smallish L2 cache compared to what a phone SoC board would have (which will hurt). It has NEON and unexpectedly it looks like if might have hardware integer division. I don't have time right now aside from a quick look at the hardware, but I doubt my code will be cost effective to run. As an aside, I did a quick and dirty port of my code over to x86. My sieve code on ARM is significantly faster than the "place holder" C version I knocked up even if you factor in bigger cache, memory and clock speed. Ignoring the sieve, the biggest (unexpected) difference is in the binary exponentiation (Fermat test); GMP on ARM is abysmal in comparison (I mean REALLY bad, which also explains why I need/use a big sieve). Regards,  bsunau7





gatra


October 28, 2014, 01:02:35 PM Last edit: October 28, 2014, 01:34:03 PM by gatra 

The concept is interesting... it's not better at finding primes than RIC or XPM because it doesn't look for primes, it looks for prime gaps which are relatively large ranges of numbers without any prime number inside. Kind of complementary to what we do. Their primes are at the start and at the end of the gaps. This part of their "paper" is not accurate: In Riecoin, it is even worse. Riecoin searches for prime tuples of length 6. Pools do accept tuples with less primes, but a 6tuple only can occur in certain places. (Look at this post for a detailed explanation.) 4tuples, by comparison, are more frequent. There are places, where a 4tuple can occur, but no 6tuple, which Riecoin truly needs. So pools have to check every submitted share whether the miner really searches for 6tuples or not. Those facts are what make it so hard to create a Prime or Riecoin pool. This was true for the first version of pools and miners, but we've been using the "first + any 3" rule for pool shares for a long time now. This doesn't have this problem as looking for 4tuples in places were 4tuples can occur but 6tuples cannot (to use their language) is actually slower than actually looking for 6tuples (because we not look for strict 4tuplets but ask for "first + any 3" of the numbers to be prime, if you strictly look for 4tuples you lose a lot of combinations). Also, they don't provide an analysis of the complexity of mining vs the complexity of verification. The complexity of the verification of a proof of work in RIC is O(log(p)^3) while that of GAP looks like O(log(p)^5), meaning their verification is slower. This may be a problem when the size of the primes grows (in GAP it may need to grow as difficulty increases at a much faster rate than in RIC since the largest gap is expected to be O(log(p)^2) ). edit: verification seems O(log(p)^5) instead of O(log(p)^4) as I originally wrote. They will eventually have problems with verification performance. This will affect pools too. Too bad... hope I'm wrong, I liked the concept. edit2: as I read user Supercomputing wrote, GAP looks more gpufriendly than XPM or RIC. Just pointing out this fact, this may be good or not, idk. cheers! Gatra




Supercomputing


October 28, 2014, 01:26:28 PM 

It is a GPU only coin. It has AMD R9 and Nvidia Maxwell written all over it. Riecoin requires specialized GPU programming skills to deal with prime numbers > 1024bit. It is for this reason that there are no GPU miners released to the public for Riecoin mining.




gatra


October 28, 2014, 01:39:06 PM 

It is a GPU only coin. It has AMD R9 and Nvidia Maxwell written all over it. Riecoin requires specialized GPU programming skills to deal with prime numbers > 1024bit. It is for this reason that there are no GPU miners released to the public for Riecoin mining. I read your comment in the GAP thread and I was adding the "edit2" part to my post here while you posted here too we agree GAP would be easier to mine with a GPU




cinnamon_carter
Legendary
Offline
Activity: 1148
Merit: 1018
It's about time  All merrit accepted !!!


October 29, 2014, 03:52:20 AM 

interesting concept (gap) however I think personally riecoin is the best introduced by far for anything to do with primes ect..... .
primecoin is good too however an 'endless' supply of coins is not the best distribution model and the other differences (which have all been debated and discussed previously several times on this thread) to me stand riecoin up a notch from primecoin.
as always with anyone launching a coin project , i wish my best, it certainly is not easy being a developer at worst it feels like a thankless job , at best it makes you feel good to know you built something that works and other people respect ..... especially when you built something different , original and/or useful which expands on the technology of bitcoin vs. just copying it......

Check out my coin Photon Merge Mine 5 other Blake 256 coins  6x your hash power https://www.blakecoin.org/The obvious choice is not always the best choice. LOOK DEEPER  Look into the Blake 256 Family  CC



Ellieo
Member
Offline
Activity: 60
Merit: 10


October 29, 2014, 05:42:47 AM 

Has the claim by Yitang Zhang that the gap between two primes would never exceed 70,000,000 being refuted already?




gatra


October 29, 2014, 12:52:11 PM Last edit: October 29, 2014, 02:06:13 PM by gatra 

Has the claim by Yitang Zhang that the gap between two primes would never exceed 70,000,000 being refuted already?
Yitang Zhang's famous work is about the existence of infinite pairs of primes separated by at most 70000000. It never says that "the gap between two primes would never exceed 70,000,000" because for numbers large enough that gap is expected to exceed any bound. Moreover, Yitang Zhang's work is not a claim: it's a proof, so don't expect it to be refuted any time soon. Yitang Zhang's work was improved from 70000000 to 600. This means that there exists a number k, such that k<600, and such that there are infinite pairs of numbers p and p+k where both p and p+k are prime. This is the closest we are to proving the twin prime conjecture, which would be similar but for k=2. In the meantime, Riecoin hasn't run out of primes, so it's looking good




popticu
Newbie
Offline
Activity: 14
Merit: 0


October 29, 2014, 06:28:00 PM 

Keep up good work Gatra




Ellieo
Member
Offline
Activity: 60
Merit: 10


October 29, 2014, 06:57:46 PM 

Has the claim by Yitang Zhang that the gap between two primes would never exceed 70,000,000 being refuted already?
Yitang Zhang's famous work is about the existence of infinite pairs of primes separated by at most 70000000. It never says that "the gap between two primes would never exceed 70,000,000" because for numbers large enough that gap is expected to exceed any bound. Moreover, Yitang Zhang's work is not a claim: it's a proof, so don't expect it to be refuted any time soon. Yitang Zhang's work was improved from 70000000 to 600. This means that there exists a number k, such that k<600, and such that there are infinite pairs of numbers p and p+k where both p and p+k are prime. This is the closest we are to proving the twin prime conjecture, which would be similar but for k=2. In the meantime, Riecoin hasn't run out of primes, so it's looking good Yup, there's no proof yet that prime numbers are bounded on the real line. It's a fact by Prime Number Theorem (PNT) that said the primecounting function that gives the number of primes less than or equal to x by: And since then




Ellieo
Member
Offline
Activity: 60
Merit: 10


October 30, 2014, 07:49:25 AM 

oops...didn't notice the incorrect primecounting function typed earlier. I was thinking of something else and got confused while at it. The correct primecounting function argument of unbounded prime number on real line should be: and when therefore




fairglu
Legendary
Offline
Activity: 1096
Merit: 1025


October 30, 2014, 08:06:08 AM 

Side question: how do you generate those images? using oldfashioned copypaste/saveas from a software + manual upload or is there a convenient online tool somewhere that would generate them and upload them automagically to imgur?




Ellieo
Member
Offline
Activity: 60
Merit: 10


October 30, 2014, 08:17:11 AM 

Side question: how do you generate those images? using oldfashioned copypaste/saveas from a software + manual upload or is there a convenient online tool somewhere that would generate them and upload them automagically to imgur? I'm not aware of any automated tool for uploading the images into imgur but it would great to know if there's any. But those images were generated from http://www.codecogs.com then uploaded into imgur.




bsunau7
Member
Offline
Activity: 114
Merit: 10


October 31, 2014, 06:27:23 AM 

It is a GPU only coin. It has AMD R9 and Nvidia Maxwell written all over it. Riecoin requires specialized GPU programming skills to deal with prime numbers > 1024bit. It is for this reason that there are no GPU miners released to the public for Riecoin mining. Gapcoin's testnet's difficulty is relatively high. The GPU prophesy might be coming true, or someone found the obvious speed up. Regards,  bsunau7




