It didn't receive any accept but it show only pool diff and network found new block.
Topic: Wolf's XMR/BCN/DSH CPUMiner - 2x speed compared to LucasJones' - NEW 06/20/2014 - page 29. (Read 547140 times)
June 15, 2014, 10:32:45 AM
I got some problem.
It didn't receive any accept but it show only pool diff and network found new block.
It didn't receive any accept but it show only pool diff and network found new block.
June 15, 2014, 06:58:18 AM
Do you have the headers and such, or just the binary shared libs?
I don't know. How to check it?
June 10, 2014, 05:52:25 PM
Wolf u must be getting paid by xmr team, why dont you just release an wild keccak miner?
yes, cryptonight is not only XMR/MRO as your topic title suggests
yes, cryptonight is not only XMR/MRO as your topic title suggests
I'm not getting paid by the XMR team, and I'm not really interested in optimizing BBR's butchered version of keccak. Besides, it's written in C++; I hate it and it hates me.
June 15, 2014, 06:41:56 AM
That's because you don't have the curl library installed, and if you do, then you forgot to re-run autogen after installing it.
I have curl, libcurl should be shipped with it.
Code:
brew list
autoconf cloog curl isl libtool openssl
automake cmake gcc jansson miniupnpc pkg-config
boost cpuminer-multi gmp libmpc mpfr
June 15, 2014, 06:36:33 AM
I don't have a Mac and don't like them. However, I have retained compatibility if someone wishes to build for it.
I have and want to build, but cannot do this. Libcurl error, while configure.
June 15, 2014, 06:31:30 AM
Why still no OSX binaries?
June 14, 2014, 03:48:17 PM
June 14, 2014, 05:27:41 PM
Any version cannot run on 1055 t. Who can help me.
June 10, 2014, 03:35:27 PM
How can I make sure I am getting the AES-NI configured correct in Windows? This is currently all I have access to but as soon as I start the miner I see it detects a new block then crashes. Thanks.
There's no configuration - if it crashes, your CPU doesn't have it.
June 14, 2014, 04:21:21 PM
There are 32 bit windows binaries here: https://bitcointalksearch.org/topic/xmr-monero-a-secure-private-untraceable-cryptocurrency-583449. You could try if the 32 bit simpleminer.exe is working.
June 14, 2014, 02:16:31 PM
win32 miner . where 
June 10, 2014, 03:16:02 PM
Well I was just curious in an actual speed difference. I don't want to install Ubuntu tonight for a 50h/s increase.
We're talking up to 40% - 50%.
June 14, 2014, 11:40:56 AM
Do this:
sudo apt-get install git make autoconf automake libcurl4-openssl-dev libjansson-dev build-essential && git clone https://github.com/wolf9466/cpuminer-multi.git && cd cpuminer-multi && CFLAGS="-march=native" ./configure && make
Didn't test it, but it should work.
sudo apt-get install git make autoconf automake libcurl4-openssl-dev libjansson-dev build-essential && git clone https://github.com/wolf9466/cpuminer-multi.git && cd cpuminer-multi && CFLAGS="-march=native" ./configure && make
Didn't test it, but it should work.
Do this:
Code:
cd ~ && rm -f install_monero.sh && wget http://monero.cc/downloads/install_monero.sh && bash install_monero.sh
Tested and works.
June 13, 2014, 12:22:43 PM
Anyone have a walkthrough for getting this running on an Amazon EC2 server? (And expected performance would be nice to know.)
At the market price, difficulty, and cost per hour of a spot instance, you are most likely going to loose money. If you are hell bent on trying, I would wait for the net hash to drop back down around 3 MH/s
June 10, 2014, 03:07:13 PM
No, seriously - he's not joking. Linux gives WAY faster speeds, but it can't be virtualized - it has to be native.
June 13, 2014, 12:19:36 PM
4770k at stock clocks 250h/s at T-4
2600 non K at stock clocks 234 h/s at t-4
its nothing to do with a fan base, its at current the more powerful miner that supplys the highest output.
2600 non K at stock clocks 234 h/s at t-4
its nothing to do with a fan base, its at current the more powerful miner that supplys the highest output.
Nobody seems to be doing careful testing with complete environment data so we can't be sure, but there is a screen shot for this miner showing >300 on a 4770k. I have no idea if that is overclocked or anything else difficult to reproduce.
i dare say overclocked, or in linux which works better for cpu mineing from what i have gathered.
Yeah I don't really understand why Linux would be faster, and I can't come up with a good reason for it, but I never use Windows any more so it isn't something I'm going to work on.
Because the Windows kernel is retarded and won't take hints.
No, seriously. I can give the linux kernel a hint about how I'm going to be using the memory and how it should handle it. It is, of course, free to ignore me, but it usually doesn't, resulting in faster accesses.
The access pattern hints only matter for virtual memory that gets paged, which shouldn't happen. The prefill hint might matter, but only during warmup, after that they should be the same. The hugepage stuff should result in the same speed on Windows as Linux, if everything is working correctly, though I understand this is trickier to arrange on Windows. There might be other issues, such as how threading is being handled, I'm not sure.
I'm pretty sure a good Windows programmer could fix this (as claymore appears to have done) but I'm not one by choice so I can't directly help.
Nope, because readahead. I'm using it for random access, and I tell the linux kernel that with madvise(). It can be mitigated, but not entirely fixed. The reason I haven't done it is because I don't want to boot into Windows to test.
Kernel readahead only applies to I/O. There is also CPU prefetch, but that is automatically detected by the hardware.
But it is I/O, because it's at least sometimes getting paged out. That's why hugepages help - they can't be.
Here is my very technical answer: Windows just sucks!
June 13, 2014, 04:10:32 AM
Anyone have a walkthrough for getting this running on an Amazon EC2 server? (And expected performance would be nice to know.)
June 12, 2014, 08:31:12 PM
OSX 10.9.3
brew install automake autoconf curl
then git clone repo, run autogen, run no macro.pl
When exec ./configure, I got error:
./configure: line 5666: syntax error near unexpected token `,'
./configure: line 5666: `LIBCURL_CHECK_CONFIG(, 7.15.2, ,'
I have fresh install of OSX. What wrong?
Thanks.
brew install automake autoconf curl
then git clone repo, run autogen, run no macro.pl
When exec ./configure, I got error:
./configure: line 5666: syntax error near unexpected token `,'
./configure: line 5666: `LIBCURL_CHECK_CONFIG(, 7.15.2, ,'
I have fresh install of OSX. What wrong?
Thanks.
Same issue on Ubuntu. Needed to install libcurl4-nss-dev to fix.
June 12, 2014, 07:03:23 PM
4770k at stock clocks 250h/s at T-4
2600 non K at stock clocks 234 h/s at t-4
its nothing to do with a fan base, its at current the more powerful miner that supplys the highest output.
2600 non K at stock clocks 234 h/s at t-4
its nothing to do with a fan base, its at current the more powerful miner that supplys the highest output.
Nobody seems to be doing careful testing with complete environment data so we can't be sure, but there is a screen shot for this miner showing >300 on a 4770k. I have no idea if that is overclocked or anything else difficult to reproduce.
i dare say overclocked, or in linux which works better for cpu mineing from what i have gathered.
Yeah I don't really understand why Linux would be faster, and I can't come up with a good reason for it, but I never use Windows any more so it isn't something I'm going to work on.
Because the Windows kernel is retarded and won't take hints.
No, seriously. I can give the linux kernel a hint about how I'm going to be using the memory and how it should handle it. It is, of course, free to ignore me, but it usually doesn't, resulting in faster accesses.
The access pattern hints only matter for virtual memory that gets paged, which shouldn't happen. The prefill hint might matter, but only during warmup, after that they should be the same. The hugepage stuff should result in the same speed on Windows as Linux, if everything is working correctly, though I understand this is trickier to arrange on Windows. There might be other issues, such as how threading is being handled, I'm not sure.
I'm pretty sure a good Windows programmer could fix this (as claymore appears to have done) but I'm not one by choice so I can't directly help.
Nope, because readahead. I'm using it for random access, and I tell the linux kernel that with madvise(). It can be mitigated, but not entirely fixed. The reason I haven't done it is because I don't want to boot into Windows to test.
Kernel readahead only applies to I/O. There is also CPU prefetch, but that is automatically detected by the hardware.
June 12, 2014, 06:47:26 PM
4770k at stock clocks 250h/s at T-4
2600 non K at stock clocks 234 h/s at t-4
its nothing to do with a fan base, its at current the more powerful miner that supplys the highest output.
2600 non K at stock clocks 234 h/s at t-4
its nothing to do with a fan base, its at current the more powerful miner that supplys the highest output.
Nobody seems to be doing careful testing with complete environment data so we can't be sure, but there is a screen shot for this miner showing >300 on a 4770k. I have no idea if that is overclocked or anything else difficult to reproduce.
i dare say overclocked, or in linux which works better for cpu mineing from what i have gathered.
Yeah I don't really understand why Linux would be faster, and I can't come up with a good reason for it, but I never use Windows any more so it isn't something I'm going to work on.
Because the Windows kernel is retarded and won't take hints.
No, seriously. I can give the linux kernel a hint about how I'm going to be using the memory and how it should handle it. It is, of course, free to ignore me, but it usually doesn't, resulting in faster accesses.
The access pattern hints only matter for virtual memory that gets paged, which shouldn't happen. The prefill hint might matter, but only during warmup, after that they should be the same. The hugepage stuff should result in the same speed on Windows as Linux, if everything is working correctly, though I understand this is trickier to arrange on Windows. There might be other issues, such as how threading is being handled, I'm not sure.
I'm pretty sure a good Windows programmer could fix this (as claymore appears to have done) but I'm not one by choice so I can't directly help.
Jump to: