Topic: CCminer(SP-MOD) Modded NVIDIA Maxwell / Pascal kernels. - page 1207. (Read 2347601 times)

sgminer requires that you set the algorithm that it will be mining via -k x11mod ( as an example ) with other parameters to get it mining ...

BUT - as far as i know ( and guys in the know please help here ) sgminer is an opencl miner - which will work - but not well on nvidia ( cuda based ) gpu ...

i have both ... so with nvidia gpu - i use ccminer ... with amd gpu - i use sgminer ...

nvidia gpus are built with different chipsets - so different 'compute' instructions are required ... which means that different ccminer's are required for the type of gpu you have ...

as djm34 posted - you will require to specify the algorithm ( that the coin you are trying to mine ) is using AND get the compute 3.0 ccminer ... ie ccminer30 ( .exe if you are using windows )

if its say - granitecoin ( shameless plug - so sue me ) - then you will need to specify the kernel to be used which is in this case -a x11 ( the algorithm parameter for ccminer ) ... if it were feathercoin for instance - the kernel for the algorithm is -a neoscrypt ...

so you command line would look something like this ...

./ccminer -o stratum+tcp://pool.profitcoin.org:3893 -u Weblogin.WorkerName -p WorkerPassword -a x11 ( plus any other parameters you require after that - but stick to the basics first as a beginner ) ...

or ccminer30.exe -o stratum+tcp://pool.profitcoin.org:3893 -u Weblogin.WorkerName -p WorkerPassword -a x11 ...

if you use ccminer that is compiled for a different compute than yours - you will get wild hashrates and NO accepted shares ...

so ... two questions ...

what coin are you mining?
which ccminer are you using?

i hope this helps - at least a little ...

#crysx

Baristor-

The miner that is in your previous post's command line is sgminer:

"I keep getting the following : sgminer -o stratum+tcp://pool.profitcoin.org:3893 -u Weblogin.WorkerName -p WorkerPassword "

You did not specify a mining algorithm in the above command line.  Each class of coin will have its own mining algorithm, like scrypt for Litecoin and SHA256 for Bitcoin.  You need to read some tutorials.  Google "CryptoBadger", or the name of the coin you are trying to mine.

This is a ccminer thread, for nVidia GPU cards.  Sgminer mostly works well on AMD GPUs, with some exceptions.   Not all ccminer versions will work on a GTX 670 card, they have older chips.      --scryptr

what kindof algo are we talking about ? Its ccminer and windows what is x11 ?

So i tried with different pool now getting the foll

pool0.btcdig.com:3333 x11 bl ock 335183
GPU #0: GeForce GTX 670, 2112 kH/s
GPU #0: GeForce GTX 670, 2119 kH/s
GPU #0: GeForce GTX 670, 2058 kH/s
etc

pool 10.btcfdig.com:3333 x11 block 335184

bl a bla bla... but no shares nothing.. m I doing something wrong ? Sorry im new to mining and atm self mining on pc rig.

may-be using ccminer (as well as an algo would help)
now if it is ccminer and x11,you are probably using a version which doesn't support compute 3.0

I keep getting the following : sgminer -o stratum+tcp://pool.profitcoin.org:3893 -u Weblogin.WorkerName -p WorkerPassword


Following error

GPU #0: GEFORE GTX 670, 2104 kH/s
GPU  #0 result for 0185af4 does not validate on CPU !
GPU  #0 result for 0185af4 does not validate on CPU !
GPU  #0 result for 0185af4 does not validate on CPU !
GPU  #0 result for 0185af4 does not validate on CPU !
Stratum time is atleast 35s in the future
pool.profitcoin.org3893 x11 block 21897
GPU #0 gefore gtx 670 2219 kH/s

GPU Result for 00d42416 does not validate on CPU!
does not validate on CPU!
does not validate on CPU!
does not validate on CPU!
does not validate on CPU!
does not validate on CPU!
 WTF ?

[2014-12-21 10:42:21] accepted: 3954/3968 (99.65%), 16290 khash/s yay!!!
[2014-12-21 10:43:18] accepted: 3955/3969 (99.65%), 16290 khash/s yay!!!
[2014-12-21 10:43:53] accepted: 3956/3970 (99.65%), 16290 khash/s yay!!!


these are mine at 'stock' settings with -i 20.5 - it seems to be doing the best at the moment ...

i am unable to OC these cards under linux ( fedora 19 ) that i am aware of (without flashing firmware or the like ) ... unless someone can shed new light on that? ...

this is a 6 x gigabyte 70ti oc machine running fedora 19 with the latest nvidia drivers in the repo ( kmod-nvidia-3.14.23-100.fc19.x86_64-331.67-1.fc19.19.x86_64 ) and kernel ( kernel-3.14.23-100.fc19.x86_64 ) ...

the other machine ( 5 x gigabyte 70ti oc ) is doing approx the same hashes per gpu ...

[2014-12-21 10:52:04] GPU #4: GeForce GTX 750 Ti, 2741 kH/s
[2014-12-21 10:52:04] GPU #2: GeForce GTX 750 Ti, 2691 kH/s
[2014-12-21 10:52:04] GPU #1: GeForce GTX 750 Ti, 2633 kH/s
[2014-12-21 10:52:04] GPU #3: GeForce GTX 750 Ti, 2704 kH/s
[2014-12-21 10:52:04] GPU #0: GeForce GTX 750 Ti, 2691 kH/s
[2014-12-21 10:52:13] GPU #1: GeForce GTX 750 Ti, 2632 kH/s
[2014-12-21 10:52:13] accepted: 2917/2927 (99.66%), 13459 khash/s yay!!!


mining with ccminer compiled under fedora 20 using cuda 6.5 with x11 - version output below ...

[root@ace-cm013 ccminer]# ./ccminer --version

*** ccminer 1.5.1-git for nVidia GPUs by tpruvot@github ***
        Built with the nVidia CUDA SDK 6.5

  Based on pooler cpuminer 2.3.2
  CUDA support by Christian Buchner and Christian H.
  Include some of djm34 additions and sp optimisations

ccminer v1.5.1-git
libcurl/7.29.0 NSS/3.17.1 Basic ECC zlib/1.2.7 libidn/1.26 libssh2/1.4.3


is there any way of overclocking on a software level ( whether via parameters in ccminer or linux itself through nvidia drivers ) like that which can be done using sgminer? ...

tanx ...

#crysx

Thanks for this. So far I seem to get the most using 18.4 = 19,145 vs. 20.5 = 19,120 or no i param = 19,125 kh/s.

that's the problem, I never understood what exactly cgminer is doing with those numbers, I think it would rather be the raw intensity

not yet, but could be added as a new param, you see the result number when "-i" is set (decimals accepted)

is that the x-intensity on cgminer ?

considering the throughput is much smaller than x11 (or should be) the chance of finding several share in one gpu round is rather small
Is there a way to not use the intensity but just the throughtput (even if it is more user friendly, I find that a bit cryptic)

This happens if you mine with a high intensity. It was fixed by submitting two shares instead of one share. X11,x13,x14,x15 should be fine. Lyra2 is not, since the implementation only find one share.

it also happens on suprnova (less often than verters (hashharder) though)

the disconnection only happens on buggy pools which sometimes resend the same job (like Hashharder blake and hashlink lyra2)

Hashharder was banning the miner after some duplicated share submits, its why i made that first, hashlink seems to only give some rejects without bans.

I was about doing an optional flag for that, but the main reason is i reset the connection on duplicates to force ask a new job.

https://github.com/tpruvot/ccminer/commit/f08bd79a6eb544ca4e8224f707bb1a016b94dabb

as it was asked sometimes ago... it would be a lot easier if everything was moved to one thread (ie the original cuda/ccminer thread).
also I as I use sp release and didn't see any post regarding this issue on your thread, I assumed it could be a problem with sp commit...

by the way, would be good if you could point me to the solution to the problem as there is no obvious things linked to that in the commit comments, so I can update manually, because I don't want to recompile again everything as I need to compile for at least 3 compute versions...

why do you ask on sp thread about features i made, check my last commits

with ccminer-sp, I get rather incessant drop in gpu usage (meaning it goes right to 0 for a very short period then go back to 100% as well as incessant "reset connection".

all good mate ...

its all a matter of time and work ... we all do what we need to do ...

my farm is about to go down in the next couple of days as im doing an 'upgrade' if you want to call it that ...

ALL the amd cards are being pulled and sold - being replaced with nvidia ones ( Gigabyte 750ti OC low profile cards - at the moment ) ...

this means i can have about 3-4 times the amount of cards running with the SAME power draw ... which is quite high for me at the moment ...

so cuda/nvidia/ccminer - x11/neoscrypt/etc - optimization is something i am VERY interested in ...

the two machines that are running the nvidia cards ( one has 5 cards - the other has 6 ) are running very smooth with the -i 20.5 parameter ...

i really look forward to what you guys can come out with ... Wolf0 upped my farm to about 75% what it was with his optimizations running sgminer ...

im waiting in sheer anticipation as to what sort of leaps that can be taken using ccminer/cuda and the 750ti OC boards ...

#crysx

sorry forgot to answer...
well why not..., I will let you know when I have something...

im testing this right this minute ...

lets see how long the rig lasts ...

#crysx

i think its a good plan ... i do it as its easy ...

and not just to test - but also as a donation just for the great effort that these guys are doing ...

i wish i had all these guys' talent under my umbrella - what a team we could make ...

and i would pay well ... after all - what the hell are we mining for?



#crysx