Topic: [ANN] cudaMiner & ccMiner CUDA based mining applications [Windows/Linux/MacOSX] - page 543. (Read 3426976 times)
May 01, 2014, 01:54:42 PM
Just realised that's the wrong compute version.... that might explain it. Is there a 3.0 exe?
May 01, 2014, 01:53:09 PM
Bigjme or others, are you perhaps able to post a compiled Windows exe of v0.7 please? I use Linux for my main miner but wanted to test on my other PC. I have 2x GTX670 and just tried the 0.6.1 version you posted (compute 3.5, not maxwell) and it crashes with heavy and gives crazy hash rates for groestl :/
May 01, 2014, 01:47:53 PM
Chris84, I secretly hoped you were going to complain that my collection of tray icons is not long enough.
May 01, 2014, 01:43:14 PM
Chris84, I secretly hoped you were going to complain that my collection of tray icons is not long enough.
Christian, you need good antivirus with internet security.. 
May 01, 2014, 01:18:23 PM
Chris84, I secretly hoped you were going to complain that my collection of tray icons is not long enough.
May 01, 2014, 12:49:45 PM
Anyone know the ETA of x11 algo ? Been reading this thread for quite sometime. Saw the previous eta was to be end of April. Any idea what would be the new ETA ?
we're down to optimizing the SIMD hash, which takes a ridiculous 66% of the total runtime.
Current hash rate on GTX 780 is 2.4 MHash/s. Want to reach 5 MH.
Christian
here's a screenshot showing the execution timeline for X11 hashing. Ignore the gaps between the hashes, these are likely caused by the code profiling operation.
The relative lengths of the execution times are the interesting parts. Biggest offenders are currently the SIMD and Echo hashes. The measurement was taken on
a GTX 780.
https://i.imgur.com/DK1vGEZ.png
And ECHO is the next problem-child
May 01, 2014, 11:47:57 AM
I'm trying to compile ccminer-0.7 in Ubuntu but I'm getting the following error after running make
I no idea how to fix!
Code:
nvcc -g -O2 -I . -Xptxas "-abi=no -v" -gencode=arch=compute_20,code=\"sm_20,compute_20\" -gencode=arch=compute_30,code=\"sm_30,compute_30\" -gencode=arch=compute_35,code=\"sm_35,compute_35\" --maxrregcount=80 --ptxas-options=-v -I./compat/jansson -o heavy.o -c heavy.cu
/bin/bash: nvcc: command not found
make[2]: *** [heavy.o] Error 127
make[2]: Leaving directory `/home/miner/ccminer-0.7'
make[1]: *** [all-recursive] Error 1
make[1]: Leaving directory `/home/miner/ccminer-0.7'
make: *** [all] Error 2
/bin/bash: nvcc: command not found
make[2]: *** [heavy.o] Error 127
make[2]: Leaving directory `/home/miner/ccminer-0.7'
make[1]: *** [all-recursive] Error 1
make[1]: Leaving directory `/home/miner/ccminer-0.7'
make: *** [all] Error 2
I no idea how to fix!
enter:
Code:
export PATH=$PATH:/usr/local/cuda/bin
after that, run make again.
Thanks, had to install Cuda toolkit as per Christian's advice, and then run that command. Working fine now.
glad to hear that, have fun mining
May 01, 2014, 11:45:57 AM
I'm trying to compile ccminer-0.7 in Ubuntu but I'm getting the following error after running make
I no idea how to fix!
Code:
nvcc -g -O2 -I . -Xptxas "-abi=no -v" -gencode=arch=compute_20,code=\"sm_20,compute_20\" -gencode=arch=compute_30,code=\"sm_30,compute_30\" -gencode=arch=compute_35,code=\"sm_35,compute_35\" --maxrregcount=80 --ptxas-options=-v -I./compat/jansson -o heavy.o -c heavy.cu
/bin/bash: nvcc: command not found
make[2]: *** [heavy.o] Error 127
make[2]: Leaving directory `/home/miner/ccminer-0.7'
make[1]: *** [all-recursive] Error 1
make[1]: Leaving directory `/home/miner/ccminer-0.7'
make: *** [all] Error 2
/bin/bash: nvcc: command not found
make[2]: *** [heavy.o] Error 127
make[2]: Leaving directory `/home/miner/ccminer-0.7'
make[1]: *** [all-recursive] Error 1
make[1]: Leaving directory `/home/miner/ccminer-0.7'
make: *** [all] Error 2
I no idea how to fix!
enter:
Code:
export PATH=$PATH:/usr/local/cuda/bin
after that, run make again.
Thanks, had to install Cuda toolkit as per Christian's advice, and then run that command. Working fine now.
May 01, 2014, 11:39:58 AM
for anyone mining jackpot, allcrypt.com seems broken 
May 01, 2014, 11:38:46 AM
hmm..is the simd part such more complex compared to e.g. groestl? or is this just overhead which you are trying to optimize?
it uses a metric ton of state per thread, which exceeds the number of registers greatly and causes huge spills into memory.
The spills slow us down an order of magnitude. It will be easy to work around, but will take me a couple of evenings.
thank you for the previous reply. I guess it will be ready in about 2 weeks time ? I think many people on this thread are very interested to know.
also, what would be the target hashrate for maxwell card , specifically gtx 750ti . I am guessing at least 1.5 mh.
May 01, 2014, 11:33:37 AM
About the compilation inUbuntu 13.10 x64 I made this
Install these if you don't libcurl4-openssl-dev
libssl-dev, libcurl-dev,
./autogen.sh
./configure -with-cuda=/usr/local/cuda-5.5
make
if fails
make clean
./configure -with-cuda=/usr/local/cuda-5.5
make
It works for me I have 3 days trying to compile in different ways versiob 0.7 I have less hashrate in myrgrs
About x11 lot of x11 coins now I hope at least have same performance that 270 with a 660ti or 670.
I think we should compile a guide to linux very specific or a scrypt I don't know.
Install these if you don't libcurl4-openssl-dev
libssl-dev, libcurl-dev,
./autogen.sh
./configure -with-cuda=/usr/local/cuda-5.5
make
if fails
make clean
./configure -with-cuda=/usr/local/cuda-5.5
make
It works for me I have 3 days trying to compile in different ways versiob 0.7 I have less hashrate in myrgrs
About x11 lot of x11 coins now I hope at least have same performance that 270 with a 660ti or 670.
I think we should compile a guide to linux very specific or a scrypt I don't know.
May 01, 2014, 11:17:50 AM
hmm..is the simd part such more complex compared to e.g. groestl? or is this just overhead which you are trying to optimize?
it uses a metric ton of state per thread, which exceeds the number of registers greatly and causes huge spills into memory.
The spills slow us down an order of magnitude. It will be easy to work around, but will take me a couple of evenings.
May 01, 2014, 11:07:13 AM
hmm..is the simd part such more complex compared to e.g. groestl? or is this just overhead which you are trying to optimize?
May 01, 2014, 11:04:14 AM
yeh i see what you mean, x11 is very long. but i guess for now i would rather having something, and an optimized version later then nothing, if you get what i mean?
May 01, 2014, 11:00:09 AM
Anyone know the ETA of x11 algo ? Been reading this thread for quite sometime. Saw the previous eta was to be end of April. Any idea what would be the new ETA ?
we're down to optimizing the SIMD hash, which takes a ridiculous 66% of the total runtime.
Current hash rate on GTX 780 is 2.4 MHash/s. Want to reach 5 MH.
Christian
here's a screenshot showing the execution timeline for X11 hashing. Ignore the gaps between the hashes, these are likely caused by the code profiling operation.
The relative lengths of the execution times are the interesting parts. Biggest offenders are currently the SIMD and Echo hashes. The measurement was taken on
a GTX 780.
May 01, 2014, 10:50:55 AM
I've also received a report by email that some 750 TIs just drop their hash rate randomly, but sometimes they recover. It's spooky.
Christian
Christian
I am hashing on Jackpot right now with the -q flag off to just watch the numbers for a bit... will let you know if I see anything like that.
Is ~3 Mhash/s for each 750 ti about right after the switch?
And with the cards dropping speed, has anyone tried it without the -d option? I am not using it since all 6 cards are on the same instance right now, but thought we should be thorough in testing.
that's what I get per 750Ti (stock clocks) too, so it seems about right :>
May 01, 2014, 10:48:51 AM
I am hashing on Jackpot right now with the -q flag off to just watch the numbers for a bit... will let you know if I see anything like that.
Is ~3 Mhash/s for each 750 ti about right after the switch?
And with the cards dropping speed, has anyone tried it without the -d option? I am not using it since all 6 cards are on the same instance right now, but thought we should be thorough in testing.
Is ~3 Mhash/s for each 750 ti about right after the switch?
And with the cards dropping speed, has anyone tried it without the -d option? I am not using it since all 6 cards are on the same instance right now, but thought we should be thorough in testing.
you get random hashrate drops but its while the diff adjusts, i got it at the start, now its 3MH/s with the occasional drop
May 01, 2014, 10:47:28 AM
I've also received a report by email that some 750 TIs just drop their hash rate randomly, but sometimes they recover. It's spooky.
Christian
Christian
I am hashing on Jackpot right now with the -q flag off to just watch the numbers for a bit... will let you know if I see anything like that.
Is ~3 Mhash/s for each 750 ti about right after the switch?
And with the cards dropping speed, has anyone tried it without the -d option? I am not using it since all 6 cards are on the same instance right now, but thought we should be thorough in testing.
May 01, 2014, 10:46:59 AM
didn't even notice a new version of ccminer was out since yesterday evening... and I read regularly what get posted in the thread...
Just some comparisons:
R9-290x: 3.2MHash
gtx780ti: 7.9MHash
However, the program crash very easily (the groestl effect... I had the problem in the previous version)
actually I can't run more than 3~4 minutes without crashing (the driver don't crash... well it just did)
ok, I was using the wrong version (self compiled...)Just some comparisons:
R9-290x: 3.2MHash
gtx780ti: 7.9MHash
However, the program crash very easily (the groestl effect... I had the problem in the previous version)
actually I can't run more than 3~4 minutes without crashing (the driver don't crash... well it just did)
so I turned off the "femri" flag and the shared one (must have forgotten a few, is there a way to change their value globally ?)
I am getting now 8.6MHash/s on jackpot
May 01, 2014, 10:23:40 AM
Anyone notice or have this problem. Just me on 2 rigs? :S
Thanks!
Alanx
Thanks!
Alanx
I've also received a report by email that some 750 TIs just drop their hash rate randomly, but sometimes they recover. It's spooky.
Christian
Jump to: