[XPM] [ANN] Primecoin High Performance | HP14 released!

[mhps]

I have a total of about 7 CPD (HP11) crunching. Since about a month ago I'm experiencing much, much lower blockrate. Currently it's about one block per 2-3 days. Wondering who finds all those blocks?

Also I noticed (since day 0) that my machines often find a few blocks in a row (over 1-3 days) and then find nothing for days/weeks. I even had 3 blocks within a few hours a few times. Can anybody confirm, that it is more likely to find a block when a block has been found in the past hours?  I can't believe this is pure luck.

Yes. They come in groups. It's not totally random. There are hidden parameters that are not well understood.  In the hidden parameter space there is tide of "luck". I can only guess the parameters are related to network speed, performance of the whole network, and the condition of your peers who are network neighbors at the moment.

[patapato]

I have a total of about 7 CPD (HP11) crunching. Since about a month ago I'm experiencing much, much lower blockrate. Currently it's about one block per 2-3 days. Wondering who finds all those blocks?

Also I noticed (since day 0) that my machines often find a few blocks in a row (over 1-3 days) and then find nothing for days/weeks. I even had 3 blocks within a few hours a few times. Can anybody confirm, that it is more likely to find a block when a block has been found in the past hours?  I can't believe this is pure luck.

Yes. They come in groups. It's not totally random. There are hidden parameters that are not well understood.  In the hidden parameter space there is tide of "luck". I can only guess the parameters are related to network speed, performance of the whole network, and the condition of your peers who are network neighbors at the moment.

Very interesting.

[muto]

and the condition of your peers who are network neighbors at the moment.

So do you expect it to be a bad thing or a good thing, that they are network neighbours? I made connection between them manually with "addnode=", but I´m not sure if this is a good idea.  

[Trillium]

I have a total of about 7 CPD (HP11) crunching. Since about a month ago I'm experiencing much, much lower blockrate. Currently it's about one block per 2-3 days. Wondering who finds all those blocks?

Also I noticed (since day 0) that my machines often find a few blocks in a row (over 1-3 days) and then find nothing for days/weeks. I even had 3 blocks within a few hours a few times. Can anybody confirm, that it is more likely to find a block when a block has been found in the past hours?  I can't believe this is pure luck.

Yes. They come in groups. It's not totally random. There are hidden parameters that are not well understood.  In the hidden parameter space there is tide of "luck". I can only guess the parameters are related to network speed, performance of the whole network, and the condition of your peers who are network neighbors at the moment.

Sorry, I can't agree with this, and I have found several thousand XPM in total now. I think you are falling into the trap of seeing patterns where there really are none: http://en.wikipedia.org/wiki/Apophenia

Similarly a gambler remembers theirs successes more than their failures (http://en.wikipedia.org/wiki/Survivorship_bias), especially when they also perceive a 'lucky streak' or 'a round of wins': http://en.wikipedia.org/wiki/Hot-hand_fallacy

Our brain evolved to allow us to find and kill things really well, while not being killed by other things ourselves, to do well socially and to have lots of sex offspring. Interpreting temporal occurrences like found XPM blocks, and indeed any kind of mathematical concept was never really what the brain was evolved for. To get around these limitations of our own minds we created the field of statistics.

I suspect if you perform rigorous statistical testing you will see there is not really the phenomenon you describe.

[vinne81]

I have a total of about 7 CPD (HP11) crunching. Since about a month ago I'm experiencing much, much lower blockrate. Currently it's about one block per 2-3 days. Wondering who finds all those blocks?

Also I noticed (since day 0) that my machines often find a few blocks in a row (over 1-3 days) and then find nothing for days/weeks. I even had 3 blocks within a few hours a few times. Can anybody confirm, that it is more likely to find a block when a block has been found in the past hours?  I can't believe this is pure luck.

Yes. They come in groups. It's not totally random. There are hidden parameters that are not well understood.  In the hidden parameter space there is tide of "luck". I can only guess the parameters are related to network speed, performance of the whole network, and the condition of your peers who are network neighbors at the moment.

Sorry, I can't agree with this, and I have found several thousand XPM in total now. I think you are falling into the trap of seeing patterns where there really are none: http://en.wikipedia.org/wiki/Apophenia

Similarly a gambler remembers theirs successes more than their failures (http://en.wikipedia.org/wiki/Survivorship_bias), especially when they also perceive a 'lucky streak' or 'a round of wins': http://en.wikipedia.org/wiki/Hot-hand_fallacy

I suspect if you plot your found block frequency against time and perform rigorous statistical testing you will see there is not really the phenomenon you describe.

Agreed, having mined 10K+ XPM.

You really can't draw conclusions from mining on average a block per 2-3 days.

[muto]

Well, a block every 2-3 days is my current frequency. In total I mined about 2k XPM. Saying that, it is possible that I see Patterns and in fact there are no patterns. That's why I asked here if someone is able to confirm it.

[mhps]

I have a total of about 7 CPD (HP11) crunching. Since about a month ago I'm experiencing much, much lower blockrate. Currently it's about one block per 2-3 days. Wondering who finds all those blocks?

Also I noticed (since day 0) that my machines often find a few blocks in a row (over 1-3 days) and then find nothing for days/weeks. I even had 3 blocks within a few hours a few times. Can anybody confirm, that it is more likely to find a block when a block has been found in the past hours?  I can't believe this is pure luck.

Yes. They come in groups. It's not totally random. There are hidden parameters that are not well understood.  In the hidden parameter space there is tide of "luck". I can only guess the parameters are related to network speed, performance of the whole network, and the condition of your peers who are network neighbors at the moment.

Sorry, I can't agree with this, and I have found several thousand XPM in total now. I think you are falling into the trap of seeing patterns where there really are none: http://en.wikipedia.org/wiki/Apophenia

Similarly a gambler remembers theirs successes more than their failures (http://en.wikipedia.org/wiki/Survivorship_bias), especially when they also perceive a 'lucky streak' or 'a round of wins': http://en.wikipedia.org/wiki/Hot-hand_fallacy

I suspect if you plot your found block frequency against time and perform rigorous statistical testing you will see there is not really the phenomenon you describe.

What you say is that statistics can show people see non-existing patterns. But the logic doesn't go the other way round. There are plenty of examples that simple statistics is blind on existing patterns, especially if the hidden process is cyclic. Averaging too long you will claim there is no seasonal change on Earth.

I don't knoe either way. Only controlled detailed tests on the main net can help. Maybe someone can fund a PhD student to try to crack it. Heck someone may already have. We know a grad student has produced a GPU miner prototype for less than 100BTC  

[Trillium]

I have a total of about 7 CPD (HP11) crunching. Since about a month ago I'm experiencing much, much lower blockrate. Currently it's about one block per 2-3 days. Wondering who finds all those blocks?

Also I noticed (since day 0) that my machines often find a few blocks in a row (over 1-3 days) and then find nothing for days/weeks. I even had 3 blocks within a few hours a few times. Can anybody confirm, that it is more likely to find a block when a block has been found in the past hours?  I can't believe this is pure luck.

Yes. They come in groups. It's not totally random. There are hidden parameters that are not well understood.  In the hidden parameter space there is tide of "luck". I can only guess the parameters are related to network speed, performance of the whole network, and the condition of your peers who are network neighbors at the moment.

Sorry, I can't agree with this, and I have found several thousand XPM in total now. I think you are falling into the trap of seeing patterns where there really are none: http://en.wikipedia.org/wiki/Apophenia

Similarly a gambler remembers theirs successes more than their failures (http://en.wikipedia.org/wiki/Survivorship_bias), especially when they also perceive a 'lucky streak' or 'a round of wins': http://en.wikipedia.org/wiki/Hot-hand_fallacy

I suspect if you plot your found block frequency against time and perform rigorous statistical testing you will see there is not really the phenomenon you describe.

What you say is that statistics can show people see non-existing patterns. But the logic doesn't go the other way round. There are plenty of examples that simple statistics is blind on existing patterns, especially if the hidden process is cyclic. Averaging too long you will claim there is no seasonal change on Earth.

I don't knoe either way. Only controlled detailed tests on the main net can help. Maybe someone can fund a PhD student to try to crack it. Heck someone may already have. We know a grad student has produced a GPU miner prototype for less than 100BTC  

Yes if you do statistics wrong it can mislead you. If you do it right it can enlighten you. The same can be said for most things.

[muto]

Well, most likely it's really pure luck. Got 9 blocks in the past 2 days after getting nearly nothing for weeks now. Every finding was >18 hours after the last block on the respective machine.

I have an i5 3570 @4x3,4 Ghz (1.56 CPD HP11) which found only one block after nearly 4 weeks 24/7 mining. Seems to be badluck for this machine Saying that, I found a total of 22 blocks on a crappy notebook with 2,5Ghz dualcore i5 (0,64 CPD HP11). I guess that's all pure luck

[mhps]

Saying that, I found a total of 22 blocks on a crappy notebook with 2,5Ghz dualcore i5 (0,64 CPD HP11). I guess that's all pure luck

C'mon many have found their slow machine turns out to be a block magnet. It can't be all coincidence. We don't fully understand the mining process of this coin.

[mechs]

Saying that, I found a total of 22 blocks on a crappy notebook with 2,5Ghz dualcore i5 (0,64 CPD HP11). I guess that's all pure luck

C'mon many have found their slow machine turns out to be a block magnet. It can't be all coincidence. We don't fully understand the mining process of this coin.

I think luck is a significant factor in mining this coin on a small scale due to variance

[teknohog]

Yes. They come in groups. It's not totally random. There are hidden parameters that are not well understood.  In the hidden parameter space there is tide of "luck". I can only guess the parameters are related to network speed, performance of the whole network, and the condition of your peers who are network neighbors at the moment.

Sorry, I can't agree with this, and I have found several thousand XPM in total now. I think you are falling into the trap of seeing patterns where there really are none: http://en.wikipedia.org/wiki/Apophenia

If the distribution of blocks is "totally random" (which, btw, is pretty hard to define in itself) then they won't be coming at equal intervals. Think about the following bit pattern, which would correspond to blocks coming in regularly every 4 days:

00010001000100010001000100010001

This is obviously not random, at least in the Kolmogorov complexity sense, that you can write a very short program to produce this string of bits.

If you move the 1s around to make this more "random", then you're bound to see some clusters of 1s, separated by longer stretches of 0s, with the occasional lone 1. However, if the process behind the bits is "totally random", then there is no particular reason why given 1s would cluster up. Yes, they come in groups, but not necessarily because of any hidden parameters.

It's interesting how shorter repetitions are considered lucky patterns, but a more distributed pattern like the one above is not usually noticed, unless you actually keep track over many days.

I've thought about this aspect of randomness in many areas of human life. When things are distributed randomly (though uniformly in a large scale), some people are bound to get more than others. This is why some people are considered "lucky", they are just the ones where things cluster up. If you want to eliminate luck, you need a command economy where everyone is forced to have the same amount of everything.

[Ifaistos]

In the pursue of randomness  
https://fbcdn-sphotos-h-a.akamaihd.net/hphotos-ak-ash3/537045_10151426897309808_230206220_n.jpg

5 days and 8h with no block found...but who's counting

[masterOfDisaster]

Regarding randomness:



Source: http://xkcd.com/221/

But I'm also posting because I have some trouble getting my Opteron 3280 to proper speed running primecoind. I've bought this CPU hoping that I would get some 1.x chainsperday with the 64 bit version of mikaelh's compile of hp11 and default settings. But the speed is not really making me happy although the CPU is running on fullspeed:

Code:

primecoind getmininginfo
{
    "blocks" : 173181,
    "chainspermin" : 5,
    "chainsperday" : 0.83579907,
    "currentblocksize" : 1374,
    "currentblocktx" : 1,
    "difficulty" : 9.86138940,
    "errors" : "",
    "generate" : true,
    "genproclimit" : -1,
    "primespersec" : 2033,
    "pooledtx" : 1,
    "sieveextensions" : 9,
    "sievepercentage" : 10,
    "sievesize" : 1000000,
    "testnet" : false
}

top
PID  USER   PR  NI  VIRT  RES  SHR S  %CPU %MEM    TIME+    COMMAND
4345 xpm    20   0 2272m 152m  34m S  798   7.8    52:10.17 primecoind
[...]

cpufreq-info (C) Dominik Brodowski 2004-2009
Bitte melden Sie Fehler an cpufreq@vger.kernel.org.
analysiere CPU 0:
  Treiber: powernow-k8
  Folgende CPUs laufen mit der gleichen Hardware-Taktfrequenz: 0
  Die Taktfrequenz folgender CPUs werden per Software koordiniert: 0
  Maximale Dauer eines Taktfrequenzwechsels: 8.0 us.
  Hardwarebedingte Grenzen der Taktfrequenz: 1.40 GHz - 2.40 GHz
  mögliche Taktfrequenzen: 2.40 GHz, 2.10 GHz, 1.90 GHz, 1.60 GHz, 1.40 GHz
  mögliche Regler: conservative, ondemand, userspace, powersave, performance
  momentane Taktik: die Frequenz soll innerhalb 1.40 GHz und 2.40 GHz.
                    liegen. Der Regler "ondemand" kann frei entscheiden,
                    welche Taktfrequenz innerhalb dieser Grenze verwendet wird.
  momentane Taktfrequenz ist 2.40 GHz.
  Statistik:2.40 GHz:97,46%, 2.10 GHz:0,00%, 1.90 GHz:0,00%, 1.60 GHz:0,00%, 1.40 GHz:2,53%  (1148)
[...] more or less the same for the next 7 cores; all running on 2.4 GHz

Sorry for the output of "cpufreq-info" in german, but I guess the report speaks for itself.
My Phenom II X6 1090T makes more than double the chainsperday (6 cores, 3.2 GHz).

I tried compiling primecoind myself with default settings; no increase in CPD.
I tried compiling primecoind with CFLAGS="-O3 -mtune=native -march=native"; no increase in CPD.
And if I set a genproclimit to below the number of cores, the CPD decreases.
I have not tried different settings for sieveextensions, sievepercentage or sievesize as I know these are the tested optimum settings for solving blocks (and I'm not interested in blindly increasing CPD for the cost of losing probabilty to solve blocks ).

Is an Opteron 3280 really that lame? I mean, this CPU consists of 4 Bulldozer modules, each carrying a "Clustered Multi-Thread" (CMT) running with 2.4 GHz. This CMT should by design be able to act like 8 cores in case of integer operation, which is the case at primecoin mining...

Does anybody have a suggestion what to try?

Kind regards
masterOfDisaster

[Trillium]

But I'm also posting because I have some trouble getting my Opteron 3280 to proper speed running primecoind. I've bought this CPU hoping that I would get some 1.x chainsperday with the 64 bit version of mikaelh's compile of hp11 and default settings. But the speed is not really making me happy although the CPU is running on fullspeed:

I know it's not perfect, but you can usually get a reasonable estimate of performance from comparing CPU that you might already have to their cpu mark (http://www.cpubenchmark.net/cpu_list.php).

In this case your Opteron 3280 is not in the list. Sadly Opteron CPU are not quite as common as Xeon and so there aren't results for most Opterons on the site.

If you download PC Mark from their site and run all tests, and upload the results it would be useful (probably to many people, not just yourself). If you can tell me what CPU mark it gives you, then I could let you know if it agrees with marks from similar scoring CPU that I might have.

Generally though we are seeing a few people overestimating the CPU's they buy. And there is not much you can do to optimise your results as long as you are sure it really is using close to 100% of CPU cycles at all times and also you have used the 64 bit version of HP.

[markm]

I have three eight-core servers that were getting blocks with version 10 but haven't gotten any since version 11.

(Sometimes a day or even two might go by without any of them getting a block but most pairs of days at least one often two got a block or even more than one block.)

I am starting to wonder if maybe there is some github weirdness or something such as maybe git pull isn't really getting me the latest code even though the last time I pulled all I got* was a change in version.cpp, which changed the actual version number displayed to 11. (What I'd already installed on my servers seems to have been version 11 minus that afterthought change of the version.cpp file to actually have it admit it was version 11...)

* All I seemed to get, anyway. Git doesn't sometimes not mention every file that it pulled changes for or anythign like that, does it?

-MarkM-

[mhps]

If you move the 1s around to make this more "random", then you're bound to see some clusters of 1s, separated by longer stretches of 0s, with the occasional lone 1. However, if the process behind the bits is "totally random", then there is no particular reason why given 1s would cluster up. Yes, they come in groups, but not necessarily because of any hidden parameters.

It's interesting how shorter repetitions are considered lucky patterns, but a more distributed pattern like the one above is not usually noticed, unless you actually keep track over many days.

That is fine. But the intervals between clustered members should have a normal distribution if the apparent clustering is caused by randomeness. This discussion came up because many people have reported that blocks came up within hours after days or weeks when there were nothing. It;s these high sigma events appearing to happen too often that makes me suspect it's not random.  I think rigorous measurement is needed to determine if it's truely random. Statistics arguements without data cut both ways.

A probably related phenomena is apparently slower machines get many more blocks than faster machines.

[Trillium]

A probably related phenomena is apparently slower machines get many more blocks than faster machines.

I could provide my own anecdotal evidence that this is not the case (for me).

I have four Core 2 Duo machines that in total are about twice as fast as my main core 2 quad system. The 4 slower machines have found 9 blocks in a month and my faster machine has found 5 which is what you would expect.

But again its meaningless with so little data.

[mhps]

A probably related phenomena is apparently slower machines get many more blocks than faster machines.

I could provide my own anecdotal evidence that this is not the case (for me).

I have four Core 2 Duo machines that in total are about twice as fast as my main core 2 quad system. The 4 slower machines have found 9 blocks in a month and my faster machine has found 5 which is what you would expect.

But again its meaningless with so little data.

I agree the quality of our data is poor. I wonder if mining the blockchain and associated addresses could tell anything.

[Ifaistos]

I've been going for almost 7 days now with HP11 and 0 blocks found with a number of machines running 24x7
1 with 1.5 cpd
1 with 0.9 cpd
4 with 0.2 cpd

all machines run the same version (ubuntu 12.04,primecoid build from source) on the same wallet.dat and made sure they started with a fresh pull of the blockchain

{
    "version" : "v0.1.2.0xpm-hp11-unk-beta",
    "protocolversion" : 70001,
    "walletversion" : 60000,
    "balance" : 0.00000000,
    "blocks" : 173976,
    "moneysupply" : 2167557.13000000,
    "timeoffset" : 14,
    "connections" : 8,
    "proxy" : "",
    "testnet" : false,
    "keypoololdest" : 1378963175,
    "keypoolsize" : 101,
    "paytxfee" : 0.00000000,
    "errors" : ""
}
{
    "blocks" : 173976,
    "chainspermin" : 16,
    "chainsperday" : 1.56431248,
    "currentblocksize" : 1227,
    "currentblocktx" : 1,
    "difficulty" : 9.86041021,
    "errors" : "",
    "generate" : true,
    "genproclimit" : -1,
    "primespersec" : 2826,
    "pooledtx" : 1,
    "sieveextensions" : 9,
    "sievepercentage" : 10,
    "sievesize" : 1000000,
    "testnet" : false
}


I was expecting to have found at least 1 block during that period and it seems others are also having issue finding blocks compared to earlier versions.


I plan to let them run for 1 more week to see how things go