Topic: [XPM] Primecoin Built-in Miner Sieve Performance Issue - page 26. (Read 69141 times)
I am running release 5 on my mac air i5 windows system. There are 8 active connections, but there are only 2 blocks synchronized after half an hour, so what's the matter?
This is my build, I used -O2 -march=native, I'm running a FX 8350, it gives me ~1250 PPS on that. I placed all the dlls in the rar. Please do your virus check and everything, I compiled that on a VM but still take precautions running a file made by another person.
https://www.dropbox.com/s/6hycwi0i4dzwu1i/Primecoin.rar
https://www.dropbox.com/s/6hycwi0i4dzwu1i/Primecoin.rar
So this seems to work better for Older AMD processors? My Girlfriend has a Turion X2 laptop laying around. I'll give it a try (in VMWARE) and let everyone know
I don't think it will work there, if you want I can make you a build for that ASAP.
This is my build, I used -O2 -march=native, I'm running a FX 8350, it gives me ~1250 PPS on that. I placed all the dlls in the rar. Please do your virus check and everything, I compiled that on a VM but still take precautions running a file made by another person.
https://www.dropbox.com/s/6hycwi0i4dzwu1i/Primecoin.rar
https://www.dropbox.com/s/6hycwi0i4dzwu1i/Primecoin.rar
So this seems to work better for Older AMD processors? My Girlfriend has a Turion X2 laptop laying around. I'll give it a try (in VMWARE) and let everyone know
all shows around the same primespersec. thats wtih the release8 (releases before crashed).
Now lets see if it changes the things (5 servers havent found for 14 hours a single block).
Now lets see if it changes the things (5 servers havent found for 14 hours a single block).
Cool, glad to hear this version is working better for some folks anyways.
waiting to see if anything will be better (meaning finding actual blocks). If yes, then I will definately drop some xpm's for your great work!
Just pushed another commit on github.
https://github.com/primecoin/primecoin/commit/eda38d62ad571be02095ce44469c0cc26a2b3675
sieve round limit further reduced to 1 second.
added mike's suggestion.
Give it a try and let me know how it goes.
This nearly tripled my PPS from the original release.https://github.com/primecoin/primecoin/commit/eda38d62ad571be02095ce44469c0cc26a2b3675
sieve round limit further reduced to 1 second.
added mike's suggestion.
Give it a try and let me know how it goes.
Now getting 2000-2400 PPS on an i7-980X @ 4 GHz. 7 blocks found so far (of which 1 was orphaned).
Compiled with -O3 -march=native.
Do you mean:
Quote
make -f makefile.unix -O3 -march=native
Code:
make CXXFLAGS="-O3 -march=native" -j6 -f makefile.unix
These extra flags speed things up? As I did not use any flag when compiling
I know my Phenom II x4 955 is quite old by today's standards, but is it supposed to be getting just ~120(not constant) with three cores and even less with 4(I have used all the optimizations provided in the threads)? People reporting over 1000 on a single CPU machine is just ridiculous.
It isn't normal. With the original build released by sunny king I got 10 pps on 8 cores of my FX 8350, and 80 pps on 4 cores (using 1 core of each module). With the recent commits and the build I made, I get 600 pps on 4 cores (1 per module) and 1200 pps on the 8 cores. And I'm a noob at compiling, so it could be made better.
all shows around the same primespersec. thats wtih the release8 (releases before crashed).
Now lets see if it changes the things (5 servers havent found for 14 hours a single block).
Now lets see if it changes the things (5 servers havent found for 14 hours a single block).
Cool, glad to hear this version is working better for some folks anyways.
double xeons 6 cores 3.07 showing this
18:02:44

{
"blocks" : 13970,
"currentblocksize" : 1000,
"currentblocktx" : 0,
"errors" : "",
"generate" : true,
"genproclimit" : -1,
"primespersec" : 3235,
"pooledtx" : 0,
"testnet" : false
}
all shows around the same primespersec. thats wtih the release8 (releases before crashed).
Now lets see if it changes the things (5 servers havent found for 14 hours a single block).
18:02:44

{
"blocks" : 13970,
"currentblocksize" : 1000,
"currentblocktx" : 0,
"errors" : "",
"generate" : true,
"genproclimit" : -1,
"primespersec" : 3235,
"pooledtx" : 0,
"testnet" : false
}
all shows around the same primespersec. thats wtih the release8 (releases before crashed).
Now lets see if it changes the things (5 servers havent found for 14 hours a single block).
legendary
Activity: 1862
Merit: 1011
Reverse engineer from time to time
I know my Phenom II x4 955 is quite old by today's standards, but is it supposed to be getting just ~120(not constant) with three cores and even less with 4(I have used all the optimizations provided in the threads)? People reporting over 1000 on a single CPU machine is just ridiculous.
This is my build, I used -O2 -march=native, I'm running a FX 8350, it gives me ~1250 PPS on that. I placed all the dlls in the rar. Please do your virus check and everything, I compiled that on a VM but still take precautions running a file made by another person.
https://www.dropbox.com/s/6hycwi0i4dzwu1i/Primecoin.rar
https://www.dropbox.com/s/6hycwi0i4dzwu1i/Primecoin.rar
Could be the 32-bit windows, as I am running x64. Maybe a -m32 switch is needed during the compile to force it into 32-bit mode.
right. my i5 and zeon boxes are 64bit win7
Here's one more you can try. Last one from me tonight.
This one has -mtune=i686 -m32, so hopefully it will cross compile to 32-bit OSes. Still runs on my 64-bit one.https://www.dropbox.com/s/uvgn1bcso6sppd0/release7_tuned_for_i686_32-bit.zip
Were tested all versions of updated client. APPCRASH on all versions of windows (W7 embedded standart x32, w7 home basic x32, w7 proffessional x64, w8 proffesional x64) with 2 core processors celeron, sempron, athlon x2 and 6core phenom x6. Works only on quad core: i5 and athlon X4 works well.
I'm running in a Windows VM and maxing out 6 cores 12 threads at 4.7ghz 1950 PPS. Not sure why you only have 4 cores.
Found 6 blocks in 8 hours
pps now higher. hope to found a block with this
EDIT: It opens up, starts downloading the blocks and then it crashes.. ;__;
Geeze... just can't win. Works fine on MY computer.
Here's another binary I've made which targets simply x86-64 processors. This SHOULD work over a wider range of processors than my earlier releases. It may not work on 32-bit OSes (not sure).
https://www.dropbox.com/s/q7w4fosqjyqydff/release8_x86-64.zip
https://www.dropbox.com/s/q7w4fosqjyqydff/release8_x86-64.zip
i will give it a try
Here's another binary I've made which targets simply x86-64 processors. This SHOULD work over a wider range of processors than my earlier releases. It may not work on 32-bit OSes (not sure).
https://www.dropbox.com/s/q7w4fosqjyqydff/release8_x86-64.zip
Thanks! That works. I'll try mining and see the speed I get.https://www.dropbox.com/s/q7w4fosqjyqydff/release8_x86-64.zip
EDIT: It opens up, starts downloading the blocks and then it crashes.. ;__;
Here's another binary I've made which targets simply x86-64 processors. This SHOULD work over a wider range of processors than my earlier releases. It may not work on 32-bit OSes (not sure).
https://www.dropbox.com/s/q7w4fosqjyqydff/release8_x86-64.zip
https://www.dropbox.com/s/q7w4fosqjyqydff/release8_x86-64.zip
Just pushed another commit on github.
https://github.com/primecoin/primecoin/commit/eda38d62ad571be02095ce44469c0cc26a2b3675
sieve round limit further reduced to 1 second.
added mike's suggestion.
Give it a try and let me know how it goes.
This nearly tripled my PPS from the original release.https://github.com/primecoin/primecoin/commit/eda38d62ad571be02095ce44469c0cc26a2b3675
sieve round limit further reduced to 1 second.
added mike's suggestion.
Give it a try and let me know how it goes.
Now getting 2000-2400 PPS on an i7-980X @ 4 GHz. 7 blocks found so far (of which 1 was orphaned).
Compiled with -O3 -march=native.
Do you mean:
Quote
make -f makefile.unix -O3 -march=native
Code:
make CXXFLAGS="-O3 -march=native" -j6 -f makefile.unix
These extra flags speed things up? As I did not use any flag when compiling
Yes, they should. -O3 is a higher level of optimization than the default -O2, so runtime performance should be better but compile will take longer. -march=native optimizes the instructions specifically for your processor.
Just pushed another commit on github.
https://github.com/primecoin/primecoin/commit/eda38d62ad571be02095ce44469c0cc26a2b3675
sieve round limit further reduced to 1 second.
added mike's suggestion.
Give it a try and let me know how it goes.
This nearly tripled my PPS from the original release.https://github.com/primecoin/primecoin/commit/eda38d62ad571be02095ce44469c0cc26a2b3675
sieve round limit further reduced to 1 second.
added mike's suggestion.
Give it a try and let me know how it goes.
Now getting 2000-2400 PPS on an i7-980X @ 4 GHz. 7 blocks found so far (of which 1 was orphaned).
Compiled with -O3 -march=native.
Do you mean:
Quote
make -f makefile.unix -O3 -march=native
Code:
make CXXFLAGS="-O3 -march=native" -j6 -f makefile.unix
These extra flags speed things up? As I did not use any flag when compiling
Just tested the 3 new windows releases from anonppcoin, shinkicker and zalfrin. Zalfrins is the fastest but shinkickers didn't crash on exit.
Heh... that's a new development. It also appears to be broken again, not pulling in new blocks as they come. Last time that was a UPNP problem.
All right, here's a fixed version. Includes all dlls, shouldn't be crashing on exit, and should be following the blockchain properly.
https://www.dropbox.com/s/6dwcc5p1koj86vz/release5.zip
Win 8 x64 AMD Phenom X4 965
Yes, that version was specifically targeted at a sandy bridge i5 processor architecture, so it's unlikely to work on other processors like your AMD. release6 MIGHT work for you, but there hasn't been a whole lot of luck with my builds if you aren't using a similar processor.
Just tested the 3 new windows releases from anonppcoin, shinkicker and zalfrin. Zalfrins is the fastest but shinkickers didn't crash on exit.
Heh... that's a new development. It also appears to be broken again, not pulling in new blocks as they come. Last time that was a UPNP problem.
All right, here's a fixed version. Includes all dlls, shouldn't be crashing on exit, and should be following the blockchain properly.
https://www.dropbox.com/s/6dwcc5p1koj86vz/release5.zip
Win 8 x64 AMD Phenom X4 965
Yes, that version was specifically targeted at a sandy bridge i5 processor architecture, so it's unlikely to work on other processors like your AMD. release6 MIGHT work for you, but there hasn't been a whole lot of luck with my builds if you aren't using a similar processor.
Jump to: