Topic: [ANN] ccminer 2.3 - opensource - GPL (tpruvot) - page 135. (Read 500113 times)

ethminer gave me a MAJOR headache trying to compile - so i gave up after three attempts with issues on fedroa 23 x64 ... i havent mined one eth ... decred was my thing - and now im back to different algos ( currently with one gb 750ti oc lp card that in my opwn system under fedora 23 x64 and lower intensity than i normally mine ... dont want to lock up my workstation ) ...

alexis78 - the miner is in his git with his optimizations here - https://github.com/alexis78/ccminer/ ... its the earlier fork of tpruvots version with alexis78 optimizations - so its reasonably easy to compile ... all you have to do is clone and compile and run ... test between the versions ... sp hasnt sent me the latest of his miner yet - so i cannot test that against anything - which is the main reason i cant give any figures or results of MY tests ...

but as with any of these sorts of tests - they will always conflict with results as different people use different miners with different settings with different operating systems ...

i have tested both alexis and sp miners ( though only against the earlier versions of the sp miner - decred only as i havent tested vanilla at all ... yet ) and the results vary between versions also ... so i cant really comment much more until i get the latest sp miner and do a thorough test ... apart from that - i need to get my test miner running - and my friend is loaning me his beast that i built him for a few weeks - so that will show where we are at with the 980ti g1 and extreme cards on my environment ...

when i get my miners back up and running - ill be able to do better testing ... although i will be selling off the majority if the gb 750ti oc lp cards - so testing for that will become obsolete for me in the next few weeks ...

ill only have the gb 980ti extreme cards left ...

#crysx

yes ethminer is different, but that version is running only eth and its clone, i'm curious about allexis version, i can't find it in this section where it is? fpr the sp version i'm a bit sceptical about the optimization, the improvements aren't as good as he claim it to be

Thanks guys.

Lot have changed since last time. i will look for those options.
i managed to start mine Vcash @ suprnova, but i don't remember how to but right options into command line. i get around 2500mh/s. is it good for this card or should i look for some tweaks?
I'm using windows 10 pro x64.

Also ethereum seems promising too, but i don't know how to get most out of my card with that, but still managed to mine that too @ suprnova.

it sounds like you are saying that the ccminer-tpruvot can mine eth ... it cant ...

depending on your setup and os and what it is you want to mine - that should determine what miner you require ...

ccminer-tpruvot ( this miner ) has the widest range of algorithms that can be used with some of the algos optimized by alexis78 ( amongst others ) which is great to mine ( like blake2s ) ...

ccminer-spmod ( the 'other' miner ) has some of the most optimized algorithms compared to this miner ... some algos are better optimized in that miner - some better here - some better in ccminer-alexis78 ( until epsylon3 merges the kernels into his miner - if he does it at all ) ...

ccminer-alexis78 has some of the algos highly optimized for THOSE specific algos ( decred / blake2s / lyra2rev2 ) ...

so you need to make a choice as to what you would like to mine - and which miner you would mine with for the algo you wish to mine with ...

eth is a completely different story altogether - and NONE of these miners mine ethereum ... you need a specialized miner for that ...

you also need to decide whether you are going to compile yourself - or use the compiled miners ...

hope this helps some more - and doesnt confuse you ...

#crysx

this is the right place, this versio can mine everything, it just missed something in the documentation, like lyra2v2, 1 970 is good if pointed at ethereum right now, try dwarf pool

Hi.

I'm trying to start mine again and it's been at least year since last time.
I got gtx 970 and water block on the way and the question is: what should i consider to mine and where i can find best cudaminer to do the job.
Thanks.

ps. i know it might not be profitable...

and sorry if this is the wrong thread to ask such things.

This miner does not seem to support client.reconnect within the stratum protocol.
I'm disappointed. client.reconnect support is required for users withing to point their rigs to my service MiningRigrentals.com

That was two years ago, so I doubt it's the same problem. Maybe it's the same error message. It means the software
isn't compatible with your GPU. Check here for your GPU and compare with the SM versions supported in the version
of ccminer you are using.

https://developer.nvidia.com/cuda-gpus#collapse4

I am having the same problem.  Has anyone found a solution to this?

Yeah. And I have stolen this idea from puddinpop's rpcminer from 2011  

I never mined bitcoin, I came later :-)

Multiple nonces per thread has been in the sp-modded kernals since bitcoin.


https://github.com/sp-hash/ccminer/blob/windows/cuda_bitcoin.cu

from 2014.

I didn't notice that and I'll have a look ASAP.
Surely it's something the standard ccminer can improve: the old versions even had bugs, in that they might hash the same nonce range two times (not the current tpruvot's, AFAIK).
It may be useful for fast algos like blake, so you can use the nonce range better ;-)

I didnt know of that. Anyway, no bad feelings.
I'm very curious to know what led you do this thing 

Since you saw my kernel, what do you think about the way i divide the scanning ranges of threads?

Isn't it a more accurate way to experiment with parameters
since you can use non powers (or even multiples) of 2 in intensity, blocks, nonces_per_thread and actually dont hash nonces > maxNonce while maintaing 100% threads of the block for the kernel's life span? (except on the last iterations)

No Sir, the kernel I posted is all my own work. I saw your kernel AFTER finishing mine. Believe it or not, the multiple nonces per round was a novel idea to me, which I had cause of insomnia :-)
And I also have other speedups which I may opensource in the future.

@pallas

Hi and nice to meet you here too .
Again, nice work on the decred algo.
Though i have a small objection.
You didn't credit the logic which led you get the boost and since the code is on an open source world, credits are all that matters.

If you want some inspiration here is a 1.5% fee binary which reports locally a ~+6% boost from the latest decred commit on 970

http://s000.tinyupload.com/index.php?file_id=07209038321464518069
A little hint: The devil hides in kernel and midstate computation

I'll opensource it in the near future, it's just the fact that too much work was put into it

the donation address is in my signature ;-)

Sorry I didn't answer sooner was on the way out the door.

Just running long poll for now, it is about a 10% improvement for me with cuda 7.5.

Feel bad not mining on your yiimp, but I need a constant block rate, it's been financing a hobby of mine hehe.

They will probably announce them in June at Computex.

No consumer models announced yet 
And so the wait continues