[ANN] ccminer 2.3 - opensource - GPL (tpruvot)

[Epsylon3]

Welcome to the discussion thread for my ccminer fork.

It contains most algorithms, and was adapted to be usable on both linux and windows.
My releases are generally tuned with the Geforce GTX 750Ti, which was, and still is a good reference for tests (Linux + Windows) and has less hashrate drop in activity (unlike the 9xx and pascal series).

For a comprehensive list of the algorithms supported as well as details about the various configuration parameters, please check the ReadMe file.

Direct linksWindows Binaries : https://github.com/tpruvot/ccminer/releases
Source Code : https://github.com/tpruvot/ccminer
Linux, read first : https://github.com/tpruvot/ccminer/wiki/Compatibility
Sample Command Line : ccminer -a bitcore -o stratum+tcp://yiimp.eu:3556 -u 1Hpjgz8i4aRdnmLXm9Eri8fFgje2dA9BaQ -p c=BTX

[Epsylon3]

Version 1.4.3 released, with "penta" Pentablake
Qubit and Doom were also imported..

Color is now enabled by default (without the "removed" -C param)

https://github.com/tpruvot/ccminer/releases

[gaalx]

are there any plans to create for AMD GPU?

[royalpro]

I am trying to use this with Xubuntu 14.04.1, kernel 3.13.0-37, nvidia driver 340.46, cuda 6.5.  I have other versions of ccminer compiled and working.  I can compile this without error.  The problem comes when I am trying to execute ccminer.  It gives me this error message: "Cuda error in file 'blake32.cu' in line 273 : invalid device symbol."  I get the same type of 'invalid device symbol' for other algorithms too.  Could someone please help me figure out why I ma getting this error.

[Epsylon3]

gallx: cuda is a nvidia framework I also own a R9 285 but i dont plan to use opencl in this project.

royalpro: This cuda error could be wrong, that means the cuda has detected an error before, generally if there is no more memory on the card.

Which card are you using ? default on my project use shaders model 5.0 only, you need to change it if you dont use a 750Ti in the Makefile.am and run ./build.sh

[royalpro]

Epsylon3, I think you might be correct with your advise.  I thought in bed before sleeping and realized these are for the 750 ti and not my 780 gtx.  I am compiling a new one with the advice your gave me.  Thanks you.

Edit: You were spot on and it is working just fine now.  Donations sent.

[royalpro]

This might be another easy question that I don't have the answer to yet, but how do I know to use either the blake or blakecoin algorithm?  They both seem to give me accepted shares, but one is half the hash rate of the other.  Thanks for any help again.

[Epsylon3]

This might be another easy question that I don't have the answer to yet, but how do I know to use either the blake or blakecoin algorithm?  They both seem to give me accepted shares, but one is half the hash rate of the other.  Thanks for any help again.

oh, thx hmm ? on which pool ?

[SS2006]

I've been getting the best performance out of my GTX970 for weeks using this release. Thank you. I hope there is more improvement to come!
6100 KH/S after OC currently for X11 (that's all i mine)

PS do i have to install 'CUDA 5.2' or something, like that other guy was saying install 'CUDA toolkit 6.5' but it did nothing for me

[Epsylon3]

it's also CUDA 6.5, there was a newer test sdk variant for the GTX 9x0 : cuda_6.5.19_windows_general_64 : https://developer.nvidia.com/cuda-downloads-geforce-gtx9xx

[SS2006]

thanks boss. is it worth installing? does it improve performance in anyway

also do you see any room for improvement for x11 etc mining in ur upcoming release, or have we hit ceiling

I was reading some good info here: https://bitcointalk.org/index.php?topic=824764.0

[royalpro]

oh, thx hmm ? on which pool ?

I am mining on la1.blakecoin.com.  I started out using blakecoin algorithm and thought I would see what happens when I used the blake algorithm.  To my surprise I was getting accepted shares but with half the hash rate.  I have been running with the blakecoin now for a while, but  I swithed it back to blake now to see if I was just wrong before.  So far I have one accepted (booo) with a message that says 'h is not zero' in red.  As of the finish writing of this I still just have the one bad accepted share with blake algorithm.

Also accepts are few and far between on both of them.  So I get a pool reported hash rate that is low or 0 for a while then it spikes up when I get an accept.

[Epsylon3]

The "blake" algo is only for SFR and NEOS (its the real blake256 algo), blakecoin changed it to speed up the hashrate

[Epsylon3]

v1.4.5 released

[SS2006]

v1.4.5 released

will this yield any improvement in x11 over your old release?

[Epsylon3]

1%, almost nothing

[BlueDragon747]

The "blake" algo is only for SFR and NEOS (its the real blake256 algo), blakecoin changed it to speed up the hashrate

The original submission of Blake-256 was 10 rounds but in the final they increased the rounds to 14 to try and gain more support from the judges but in the end NIST wanted the sponge function to get total immunity from length extension attacks rather than just resistance like all the other algo's in the final of the SHA-3 competition

Blake is based on ChaCha which is often used in an 8 round configuration hence why I chose to use that over the other extra security margin versions as that is all the extra rounds do!

Here is the proof that 8 round Blake-256 has a minimum security margin of 2²⁰⁰ for best attack by a world leading expert academic team in cryptography
http://eprint.iacr.org/2013/852.pdf

for brute force attack there is still no better than 2²⁵⁶ and as the process of mining is like a partial preimage attack seems pointless to waste time, resources and efficient using the final submission 14 round version as it did not help win over the NIST judges  

all variants are real due to a feature of the Blake algo "simple speed/confidence trade-off with the tunable number of rounds"

reducing the rounds also makes the algo smaller in hardware for FPGA so you can build higher Fmax bitstreams and would also make an Asic smaller thus cheaper, faster, and use less watt/hash for same nanometer process

*also note that the rounds in Blake2 are also reduced "from 14 to 10 for BLAKE2s (successor of BLAKE-256)"

Thanks for porting the Blakecoin version    

[smolen]

reducing the rounds also makes the algo smaller in hardware for FPGA so you can build higher Fmax bitstreams and would also make an Asic smaller thus cheaper, faster, and use less watt/hash for same nanometer process

I really like this approach - instead of inventing ASIC-proof PoW make the transition to ASIC mining as smooth as possible.

[Epsylon3]

v1.4.6 released with S3 algo

[grendel25]

Not working for me.  At first it would give a red error message, something about not detecting any NVIDIA gpu.  So I set the ccminer to run as administrator and it doesn't give the same error.  Instead, it now just sits there and says press any key to continue and when I press a key the window closes? 

I gotta reload drivers?  pfft... tooooo lazy lol.  I'll wait til another release and see if it works then maybe.  hahaha