Author

Topic: [XPM] [ANN] Primecoin High Performance | HP14 released! - page 113. (Read 397647 times)

sr. member
Activity: 301
Merit: 250
I understand that.  It kills my GPU mining when I use all cores.  But I think you missed something in my post.  I'm getting 5k PPS on 7 cores in the windows client but only 1k PPS in the Linux one.

Look up the "nice" command.

Any CPU mining, whether built in to a coin or using minerd aka cpuminer or whatever, use "nice coind" or "nice minerd", it runs at lower priority so does not seem to impact GPU mining nor FPGA mining nor ASIC mining yet still gets to use most of the cores most of the time.

-MarkM-


For debug purpose, I've shut down my GPU mining and tried all cores.  I still only get about 1200PPS in linux vm but I'm getting 6000 in Windows.

I was chitchatting with someone else and he said his Linux PPS got halved when he switched to this client too.

Which Linux distribution are you using? I suspect some distributions might not be supplying a properly optimized version of GMP. I may eventually start releasing my own Linux binaries.

Ubuntu 12.04

Looks like Ubuntu 12.04 comes with GMP version 5.0.2. That's from Aug 2011. It may fail to recognize newer CPUs. In that case, it won't choose the optimal code path.
legendary
Activity: 1484
Merit: 1000
Excellent!! found first block in 1 hour time with this windows build..... Grin

What build? link?

Its a mikaelh's modified client. I downloaded it from the first page of this thread. Windows X64 built binaries.
newbie
Activity: 37
Merit: 0
I understand that.  It kills my GPU mining when I use all cores.  But I think you missed something in my post.  I'm getting 5k PPS on 7 cores in the windows client but only 1k PPS in the Linux one.

Look up the "nice" command.

Any CPU mining, whether built in to a coin or using minerd aka cpuminer or whatever, use "nice coind" or "nice minerd", it runs at lower priority so does not seem to impact GPU mining nor FPGA mining nor ASIC mining yet still gets to use most of the cores most of the time.

-MarkM-


For debug purpose, I've shut down my GPU mining and tried all cores.  I still only get about 1200PPS in linux vm but I'm getting 6000 in Windows.

I was chitchatting with someone else and he said his Linux PPS got halved when he switched to this client too.

Which Linux distribution are you using? I suspect some distributions might not be supplying a properly optimized version of GMP. I may eventually start releasing my own Linux binaries.

Ubuntu 12.04
member
Activity: 77
Merit: 10
Now I'm looking into this "nice" command which is difficult to google by the way.

Maybe you should look into the "man" command first. But that's probably difficult to google as well.  Roll Eyes
hero member
Activity: 552
Merit: 500
Excellent!! found first block in 1 hour time with this windows build..... Grin

What build? link?
sr. member
Activity: 301
Merit: 250
I understand that.  It kills my GPU mining when I use all cores.  But I think you missed something in my post.  I'm getting 5k PPS on 7 cores in the windows client but only 1k PPS in the Linux one.

Look up the "nice" command.

Any CPU mining, whether built in to a coin or using minerd aka cpuminer or whatever, use "nice coind" or "nice minerd", it runs at lower priority so does not seem to impact GPU mining nor FPGA mining nor ASIC mining yet still gets to use most of the cores most of the time.

-MarkM-


For debug purpose, I've shut down my GPU mining and tried all cores.  I still only get about 1200PPS in linux vm but I'm getting 6000 in Windows.

I was chitchatting with someone else and he said his Linux PPS got halved when he switched to this client too.

Which Linux distribution are you using? I suspect some distributions might not be supplying a properly optimized version of GMP. I may eventually start releasing my own Linux binaries.
member
Activity: 63
Merit: 10
The client seems to crash for me after few minutes of mining

running hp3-win64
on Win 7 x64 I7 930



any ideas?
newbie
Activity: 37
Merit: 0
I understand that.  It kills my GPU mining when I use all cores.  But I think you missed something in my post.  I'm getting 5k PPS on 7 cores in the windows client but only 1k PPS in the Linux one.

Look up the "nice" command.

Any CPU mining, whether built in to a coin or using minerd aka cpuminer or whatever, use "nice coind" or "nice minerd", it runs at lower priority so does not seem to impact GPU mining nor FPGA mining nor ASIC mining yet still gets to use most of the cores most of the time.

-MarkM-


I fixed my PPS in linux by restarting the linux VM.

Now I'm looking into this "nice" command which is difficult to google by the way.

I set:
Code:
sudo nice -n -20 primecoind setgenerate true -1

And I'm still taking a 200khps hit on my GPU mining.  It did improve by 50khps over the old way.
legendary
Activity: 1484
Merit: 1000
Excellent!! found first block in 1 hour time with this windows build..... Grin
hero member
Activity: 552
Merit: 500
Any updated windows builds ?
hero member
Activity: 840
Merit: 1000
Jeez it looks like a lot has happened when I was away. So, do you just set the sieve size in the primecoin.conf file? Also what is the optimal sieve size?
legendary
Activity: 2940
Merit: 1090
I think the block finder has a huge advantage over the rest of the network because they can start work on the next block immediately, and it takes time for the newly found block to propagate through the network. Others cannot start work on the next block until they receive it.

Perhaps this helps explain why primecoin blocks tend to come in spurts?

Except that that would result in back to back blocks, not spurts that have gaps between blocks the one node found, other nodes having found blocks inbetween.

-MarkM-
hero member
Activity: 742
Merit: 500
ok

so it seems the more blocks one instance finds, the more it can find i *know* this is maths - wrong

but seriously as soon as my instances find one block, then next just seem to roll in.

[about 350 XPM mined so far in 48 hours so I am pretty small scale]

I think the block finder has a huge advantage over the rest of the network because they can start work on the next block immediately, and it takes time for the newly found block to propagate through the network. Others cannot start work on the next block until they receive it.

Perhaps this helps explain why primecoin blocks tend to come in spurts?
newbie
Activity: 37
Merit: 0
I understand that.  It kills my GPU mining when I use all cores.  But I think you missed something in my post.  I'm getting 5k PPS on 7 cores in the windows client but only 1k PPS in the Linux one.

Look up the "nice" command.

Any CPU mining, whether built in to a coin or using minerd aka cpuminer or whatever, use "nice coind" or "nice minerd", it runs at lower priority so does not seem to impact GPU mining nor FPGA mining nor ASIC mining yet still gets to use most of the cores most of the time.

-MarkM-


For debug purpose, I've shut down my GPU mining and tried all cores.  I still only get about 1200PPS in linux vm but I'm getting 6000 in Windows.

I was chitchatting with someone else and he said his Linux PPS got halved when he switched to this client too.
legendary
Activity: 1484
Merit: 1000

Yes you're successfully mining if primespersec is > 0.  How long to find a block?  This is solo mining where luck plays a huge part.  If the diff were to stay the same (which it is not), I would estimate you will get 1 block every 24 hours.  That's just an extremely rough guide based on how many blocks I'm currently getting with my PPS.  You could get one in 5 minutes or nothing for 3 days though!

Aww thanks that was clear...I will wait and see now when it will find a block.
legendary
Activity: 2632
Merit: 1023
ok

so it seems the more blocks one instance finds, the more it can find i *know* this is maths - wrong

but seriously as soon as my instances find one block, then next just seem to roll in.

[about 350 XPM mined so far in 48 hours so I am pretty small scale]
legendary
Activity: 2940
Merit: 1090
I understand that.  It kills my GPU mining when I use all cores.  But I think you missed something in my post.  I'm getting 5k PPS on 7 cores in the windows client but only 1k PPS in the Linux one.

Look up the "nice" command.

Any CPU mining, whether built in to a coin or using minerd aka cpuminer or whatever, use "nice coind" or "nice minerd", it runs at lower priority so does not seem to impact GPU mining nor FPGA mining nor ASIC mining yet still gets to use most of the cores most of the time.

-MarkM-
newbie
Activity: 37
Merit: 0
Prior to this release I was getting about 1400 PPS in the Windows client and about 1800-2000 PPS in my ubuntu VM.

I downloaded this Windows one and get about 5k PPS on 7 cores on my FX-8350 @ 4.7ghz.

I tried the the linux one and I'm only getting ~996 PPS?

I like using the linux one because it always seems to be 10c cooler and prior to this I was getting higher PPS in linux.

Steps I did:

Code:
apt-get install make
apt-get install git
git clone https://github.com/mikaelh2/primecoin
apt-get install libgmp-dev
sudo apt-get update
sudo apt-get install python-software-properties screen python-rrdtool python-pygame python-scipy python-twisted python-twisted-web python-imaging build-essential libglib2.0-dev libglibmm-2.4-dev python-dev autoconf automake ncurses-dev sysstat gcc-mingw32 libmysql++-dev cloog-ppl build-essential automake gcc libevent-dev libmemcached-dev libcurl4-openssl-dev zlib1g-dev libjansson-dev curl memcached libtool unzip freeglut3-dev libxi-dev libxmu-dev build-essential freeglut3-dev libxi-dev libxmu-dev mpich2 libdb-dev libminiupnpc-dev libboost-all-dev libdb++-dev git uthash* libgmp3-dev

cd ~/primecoin/src
make -f makefile.unix
mv primecoind /usr/local/bin

primecoind setgenerate true 7

I'm a linux noob so go easy on me.

primecoind setgenerate true -1 is better as it optimizes for your cpu

I understand that.  It kills my GPU mining when I use all cores.  But I think you missed something in my post.  I'm getting 5k PPS on 7 cores in the windows client but only 1k PPS in the Linux one.
hero member
Activity: 820
Merit: 1000


{
"blocks" : 48294,
"currentblocksize" : 1000,
"currentblocktx" : 0,
"errors" : "",
"generate" : true,
"genproclimit" : -1,
"primespersec" : 3195,
"pooledtx" : 0,
"testnet" : false
}

Thats getmininginfo details.. so this means my qt client is mining primecoin? and if its mining, how long time it will take to find a block on current pps? Thanks.  Wink
Yes you're successfully mining if primespersec is > 0.  How long to find a block?  This is solo mining where luck plays a huge part.  If the diff were to stay the same (which it is not), I would estimate you will get 1 block every 24 hours.  That's just an extremely rough guide based on how many blocks I'm currently getting with my PPS.  You could get one in 5 minutes or nothing for 3 days though!
sr. member
Activity: 359
Merit: 250
Prior to this release I was getting about 1400 PPS in the Windows client and about 1800-2000 PPS in my ubuntu VM.

I downloaded this Windows one and get about 5k PPS on 7 cores on my FX-8350 @ 4.7ghz.

I tried the the linux one and I'm only getting ~996 PPS?

I like using the linux one because it always seems to be 10c cooler and prior to this I was getting higher PPS in linux.

Steps I did:

Code:
apt-get install make
apt-get install git
git clone https://github.com/mikaelh2/primecoin
apt-get install libgmp-dev
sudo apt-get update
sudo apt-get install python-software-properties screen python-rrdtool python-pygame python-scipy python-twisted python-twisted-web python-imaging build-essential libglib2.0-dev libglibmm-2.4-dev python-dev autoconf automake ncurses-dev sysstat gcc-mingw32 libmysql++-dev cloog-ppl build-essential automake gcc libevent-dev libmemcached-dev libcurl4-openssl-dev zlib1g-dev libjansson-dev curl memcached libtool unzip freeglut3-dev libxi-dev libxmu-dev build-essential freeglut3-dev libxi-dev libxmu-dev mpich2 libdb-dev libminiupnpc-dev libboost-all-dev libdb++-dev git uthash* libgmp3-dev

cd ~/primecoin/src
make -f makefile.unix
mv primecoind /usr/local/bin

primecoind setgenerate true 7

I'm a linux noob so go easy on me.

primecoind setgenerate true -1 is better as it optimizes for your cpu
Jump to: