Topic: CCminer(SP-MOD) Modded NVIDIA Maxwell / Pascal kernels. - page 844. (Read 2347659 times)

Submitted a tiny speedup in the quark algo. A teaser
+200KHASH on the gtx 970

Where are my donations?

I have another optimalization in the quark algo. If you want it opensource please donate.
The weekend is comming and I need beer

Cuda 7.5 seems to perform bether in most algos with the correct tuning. Chrx: did you try to compile the latest git? How is it performing on linux? x11/x13/qubit

much appreciated scryptr ...

i am streamlining the donation servers by culling a lot of long methods - and shortcutting - as well as cleaning a few scripts and writing new scripts for mining as well as the stratum proxies ... which lead me to the is amazing revelation that ccminer uses .conf files all along ( DUH! ) ...

im fiddling with making ccminer a centralized miner and a service rather a commandline executable - with .conf files being the only internal file required ... the .conf file is the bit im really stuck on before i can do the rest and test ...

this should make thefarm as efficient as possible ...

which also lends itself to allow the creation of a management console that could allow changes ( and creation / deletion ) of the .conf files using a web interface for the entire farm - instead of being changed individually ...

i have also looked at you script that you wrote a while back - and will be changing a few things in it to make it a little more uptodate and adjusted to what the farm is doing ... your help with this would be very much appreciated ... im actually perplexed why you havent worked on that script further - it seems to be logically sound ...

this script could also be the thing that activates / deactivates the miner on the various .conf files ... but i have yet to get my head around how that could work ...

tanx for the .conf layout though mate ... that is a great help ...

#crysx

Are You using release 74?
I'm using the same wallet but I get more non accepted shares then good ones. Overnight I got accepted 4/27. Not sure about the diff. It looks worst if I use it with factors 0.5 and 2.
I'll do another, longer run without it...

JSON.CONF--

The following is a simplified example:

=====cut line======

{
   "url" : "stratum+tcp://stratum.nicehash.com:3345",
   "user" : "BTCaddyhere.wrkr1",
   "pass" : "x",

   "algo" : "quark",
   "intensity" : "23.7",
      
   "api-bind" : "127.0.0.1:4068",
   "statsavg" : 20,
   "quiet" : false,
   "debug" : false,
   "protocol" : false
}

======cut line=====

Save it as "name.conf".  In Linux, use this command line to call it:

./ccminer -c name.conf

You can add parameters, like cpu mining and cpu priority, and I generally execute the command line in a BASH file with environment parameters.  I've given a number of examples for solo-mining in this thread.

--scryptr

hi all ...

i just wanted to know if anyone uses a configuration file for ccminer ...

if you do - what is the actual format of the .conf file and what structure should it take ( i know its in json format - but how is it done? ) ...

i have been using full commandline executions for as long as i can remember - and now im starting to automate the entire farm through cron / file / systemd ( and eventually through web interface ) ... this is taking me a long time - and then i realized i could use .conf files to set ccminer to run different commands at different times ...

anyway - if anyone could help - that would be great ... even with a couple of examples of the .conf files that could be written in different ways ...

it would be very much appreciated ...

tanx ...

#crysx

The diff parameter doesn't matter for solomining at all (sorry, missed that you said solo). Using wallet 0.8.7.1 I have no issues solomining (found a block while writing this). Also, you shouldn't be able to find shares (reject or not) anywhere near that frequently with that hashrate.
In fact, a 5 card 970 rig (ignoring the 980 for simplicity) have an average block finding frequency of 36 minutes with the current difficulty. It's like you're mining to another wallet that's not FTC. Or your wallet have sync issues.

Asking about solo mining... diff 2 seems even worst:

Now trying diff 0.5
And runing ccminer.exe -a neoscrypt -o IP:port -u x -p y --broken-neo-wallet -d 0,2,3,4 --diff 0.5

im about to compile and test on the test system - and see how it runs ...

fedora 22 x64 c7.5 - and stock clocks ... gigabyte 750ti oc lp ...

if i ever get to the office today ...

im following your instructions on your git genoil ...

trying to get it done on ssh now - but very limited with all the security i deal with on my network ...

edit - json compile breaks ...

#crysx

I've heard from several ETH miners using 750Ti that hashrate has already dropped significantly, like on Windows (because of TLB trashing). It's more like 4MH now...

I had to check again in old logs and i remembered it wrong. My cards where running around 10Mhs unclocked. They are factory oc:d a bit though. Never oc:d memory while i was mining ETH.

That's exactly why I asked because a couple months ago all the profit calc sites were way off, showing much more profit than it really was.


Depends on the pool. If there are rejects (low diff shares) then decrease the diff multiplier below 1.0 and if a pool shows low hashrates without rejects then increasing -f is a good place to start.

Maybe You forgot to OC  Memory should be overclocked too for ETH(especially 750ti). I was hashing at 21+MH with 2x750ti.

I should do some tests again... Never tried other maxwells in linux. Maybe I'll find some time around wednesday...

BTW what should be the --diff for FTC solo mining? Default settings in r.74 gets accepted 4/10 blocks.

With the current difficulty and price etherum is minable on the 750ti. (9 MHASH)

50watt *24 hours. 1,2Kwh

66% profit with a powercost of $0.1 per KWh.

I was mining ETH earlier with 750Ti:s on linux and the cards hashed at around 7-8Mhs back then.

http://etherscan.io/ether-mining-calculator

Check by yourself I allways got a correct calculation.
750ti was doing 8-11MH some time ago in linux.
Don't know if it degraded allready...

970 should be around 20MH.

I submitted another improved shavite and simd on compute 5.2. Now qubit and x13 is looking much bether.

but still abit slower than release 74. Can you compile and  test guys?

I have submitted more changes to x11,x13,x14,x15.

Closer to the cuda 6.5 hashrate in release 74. But still some more left.

With the right code, it will probably be better. It already is on some algo, will be on others.
From my experience, the 7.5 compiler is cleanier and more predictable: it generates less broken binaries than 6.5.
Since all was based on 6.5, it takes time to update, but it will be well worth.

https://github.com/Genoil/dagSimCL

Maxwell 5.2 cards are fine until the planned switch to POS. GCN1.0 cards already on the downslope. DAGpocalypse however is another thing. But as with these gloomy predictions, you'll never be sure whether or not it will actually happen