Topic: [XPM] [ANN] Primecoin High Performance | HP14 released! - page 35. (Read 397657 times)
Nice! I'll keep my eyes peeled for this...Might install ubuntu and compile it myself.
I pushed some big optimizations for -hp10 to github so that more people can test them.
Thanks man!
I pushed some big optimizations for -hp10 to github so that more people can test them.
Any chance of getting a compiled Windows binary? I'd like to test on a 32 core Opteron rig and a 3930k.
I pushed some big optimizations for -hp10 to github so that more people can test them.
Why would you care about having a different wallet on each system?
My understand from reading about the VPS' is that one wallet with 10,000+ 'addresses' can be used on multiple systems and thus you only need to check 1 to see all. Since I have 5 wallets on 5 systems, I have to manually check each system to see if it mined anything or not. Call me lazy, but I prefer not ro run from computer to computer throughout the house checking balances daily.Yeah, I've been manually checking each and then exporting immature blocks to one wallet to track them (and also in case I blow through my credits on Azure before they mature). At least I'm half efficient that way.
But how does that work (the one wallet idea) since each block that's mined gets a new address?
I finally got brave today and checked all of my computers. 6 days, 5 computers (2@~.8 and 3@~.45 chains per day), 4 blocks.
Needless to say, I are happy!
Needless to say, I are happy!
Do you mean you're not checking at all? At least, you make sure that everyone is still mining, right? It would be too bad to have idle miners for days
Why would you care about having a different wallet on each system?
Just set up a script that sends the funds as soon as they mature to a central wallet. If you're on linux, it's just a oneliner to execute periodically:
Code:
b=$(primecoind getbalance); if [ $(echo "$b > 0" | bc) -eq 1 ]; then bminusfee=$(echo "$b - 0.01" | bc) && primecoind sendtoaddress $bminusfee; fi
(not tested)
I finally got brave today and checked all of my computers. 6 days, 5 computers (2@~.8 and 3@~.45 chains per day), 4 blocks.
Needless to say, I are happy!
Needless to say, I are happy!
Do you mean you're not checking at all? At least, you make sure that everyone is still mining, right? It would be too bad to have idle miners for days
Why would you care about having a different wallet on each system?
I finally got brave today and checked all of my computers. 6 days, 5 computers (2@~.8 and 3@~.45 chains per day), 4 blocks.
Needless to say, I are happy!
Needless to say, I are happy!
Do you mean you're not checking at all? At least, you make sure that everyone is still mining, right? It would be too bad to have idle miners for days
Why would you care about having a different wallet on each system?
Did the devellopement of the HP release will continue or Mikaelh is occupied with the official client?
I saw in the previous page there is a 6% improvement possible. Maybe with some other trick, we could go to +10% with HP10.
I saw in the previous page there is a 6% improvement possible. Maybe with some other trick, we could go to +10% with HP10.
Those are ypool stats I believe and have nothing to do with hp9.
He is correct. lulz, I'm in the wrong thread. I visit that thread, then this one, then the gpu and got mixed up. Sorry.
Those are ypool stats I believe and have nothing to do with hp9.
Anyone been tweaking the roundsievepercentage value ?
I seem to get a +10% increase in CPD with 75 instead of 70 with minimal impact on pps. That is on my core i7, doesn't work that good on my vps.
Still wondering what the best sievesize is too, we know it's cache dependent but are there any rule of thumb to apply on a known cache size ?
Trying to get the best of my few miners still running, mining is really harder now than it was at the beginning of the month...
I seem to get a +10% increase in CPD with 75 instead of 70 with minimal impact on pps. That is on my core i7, doesn't work that good on my vps.
Still wondering what the best sievesize is too, we know it's cache dependent but are there any rule of thumb to apply on a known cache size ?
Trying to get the best of my few miners still running, mining is really harder now than it was at the beginning of the month...
I'm using 80% w/a 25% sievepercentage, 1600000 sievesize, 1000000 primes. I've locked L1cache @ 672000 and saw a lot of 7ch, but after I let it autotune on another 24 hour run I saw an increase in 8ch w/same 7ch output or better (it's at 800000 now) and primoral multiplier at 43 usually. (i7/[email protected]/Win7)
can you send your primecoin.conf ?
Anyone been tweaking the roundsievepercentage value ?
I seem to get a +10% increase in CPD with 75 instead of 70 with minimal impact on pps. That is on my core i7, doesn't work that good on my vps.
Still wondering what the best sievesize is too, we know it's cache dependent but are there any rule of thumb to apply on a known cache size ?
Trying to get the best of my few miners still running, mining is really harder now than it was at the beginning of the month...
I seem to get a +10% increase in CPD with 75 instead of 70 with minimal impact on pps. That is on my core i7, doesn't work that good on my vps.
Still wondering what the best sievesize is too, we know it's cache dependent but are there any rule of thumb to apply on a known cache size ?
Trying to get the best of my few miners still running, mining is really harder now than it was at the beginning of the month...
I'm using 80% w/a 25% sievepercentage, 1600000 sievesize, 1000000 primes. I've locked L1cache @ 672000 and saw a lot of 7ch, but after I let it autotune on another 24 hour run I saw an increase in 8ch w/same 7ch output or better (it's at 800000 now) and primoral multiplier at 43 usually. (i7/[email protected]/Win7)
That surely is a profit, but how much did you pay for the xeons? The ones you listed that get 2+ cpd range from $1000 to $1700 alone.
i'm not looking for ROI !!! , cause i didn't invest the money !, these servers would have been idle otherwise .
well as far as i know they've paid some 120-150K$ for all of them , they have all the bells and whistles you can imagine , HP DL380 G7 and G8 servers with SSDs and SAS disks some supermicros with SAN and such and such.
all i'm saying is that the servers are idle for another month , it's better to put them to good use
I finally got brave today and checked all of my computers. 6 days, 5 computers (2@~.8 and 3@~.45 chains per day), 4 blocks.
Needless to say, I are happy!
Needless to say, I are happy!
Do you mean you're not checking at all? At least, you make sure that everyone is still mining, right? It would be too bad to have idle miners for days
That surely is a profit, but how much did you pay for the xeons? The ones you listed that get 2+ cpd range from $1000 to $1700 alone.
2 million is typically seen as the best although I use 2000001 as it is prime :wink:
I did not see any consensus regarding 2M being the best.
Sources ?
You should check the testnet if you pursue fine optimization of your miners.
Anyway I'm not really sure why isn't it a bit more efficient to use a prime number as size of the sieve. Do you know the basic principle behind how the sieve works?
i doubt that using a prime number helps at all !!! , my brother explained it to me , but i'm not as a math geek as he is
, yet from my understanding it's for eliminating numbers in a fast manner , it tries to find if a number is prime or not by dividing the number by the first X primes , let's say you'll use 1M , it means that it'll have (imagine a database of) first 1Million primes , and divides your number very fast to them if the remainder is 0 , it just skips the number , the sieve size ofcourse matters but from my understanding larger sieve doesn't mean better performance , as the prime number get's bigger the calculation time rises , and so does your performance drop , i don't know how much is a sweet spot for sieve , after i've dropped from 2M to 1M , my chains/day increased by a hofty 20% !! i have small farm of servers , and i'm using them to mine prime coins , the servers differ from dual 2650s to dual 5680s down to 2*5620s as well as core i5 , in total there are 20 servers , and i earn like from as low as 3-4 blocks to 8-10 blocks a day ,
on 5680s and 2650s i have cent 6 64 , so i've compiled hp9 with ease , but on 5620s i only have el5.5 64 , and i couldn't compile hp9 , so i've installed xen on them and on xen installed cent 6 and hp9 on them they give .
first of all is there anyway to install hp9 directly on el 5 ? does anyone have a compiled version even ?
second of all , what sieve size would give best performance , i didn't see that much of difference between 1m and 2m so i'm using 2m .
also, is there a way that i can install a server like ypool and connect everyone to it and work like a cluster together ?
on 5680s and 2650s i have cent 6 64 , so i've compiled hp9 with ease , but on 5620s i only have el5.5 64 , and i couldn't compile hp9 , so i've installed xen on them and on xen installed cent 6 and hp9 on them they give .
first of all is there anyway to install hp9 directly on el 5 ? does anyone have a compiled version even ?
second of all , what sieve size would give best performance , i didn't see that much of difference between 1m and 2m so i'm using 2m .
also, is there a way that i can install a server like ypool and connect everyone to it and work like a cluster together ?
hey man sorry to hear that, you look like a total bro... Are you sure you are not paying more for electricity than actually earning?
2 million is typically seen as the best although I use 2000001 as it is prime :wink:
i don't care that much about the electricity bill , since the servers are colocated elsewhere and mostly are idle , i just renice the process , though in about a month or two i'll need to stop mining
, cause we'll enter another phase in our project still with electricity prices as low as here , i guess i'm still earning something , let's say the farm is earning 50XPM/day , and each server is using some 450Watts excluding cooling ofcourse , 20*450=9KVA since it's UPSed it'll be like 9*1.2 = 10.8 , and with the electricity prices of here which ranges from 0.01$/KWH to 0.05$/KWH depending on your usage , let's assume 0.05$/KWH , it all equals to 12.96$ electricity bill , and with current prices of 0.735 i'd have total earnings of 36.75$/day net profit of 23.79$/day.
i agree that it's not that good , but considering that other cryptocurrencies don't pay any better , it's this or nothing
yet i'm adding a 5639 to my mining pool , i cross my fingers for better luck , FYI i've changed the sieve size from 2M to 1M , and got 20% boost for my chains per day
right now i'm getting 1.26191589 ch/day on virtualized 5620s
, 2.64356530ch/day on 2650s and about 2.59502632 on 5680s
2 million is typically seen as the best although I use 2000001 as it is prime :wink:
I did not see any consensus regarding 2M being the best.
Sources ?
You should check the testnet if you pursue fine optimization of your miners.
Anyway I'm not really sure why isn't it a bit more efficient to use a prime number as size of the sieve. Do you know the basic principle behind how the sieve works?
I wonder if splitting up the sievepercentage and the round between the cores if that would be more efficient then them all doing the same type of work
Jump to: