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

[SparkyU]

What parameters would be best to lower gpu load?  I browsed the readme but wasn't sure which would be the correct ones.  Each coin and applicable miner seems to have different settings.  I typically have a low settings bat I like to use when I'm on my pc so it's still useable for watching movies or whatever and then a full blast bat when I'm away.

[1713543672]

1713543672

[1713543672]

1713543672

[CodyF86]

I posted this on the decred forum because I've been lurking over there fiddling around, but wanted to post it here if anyone had any use for it. I know you can just use telnet also, but I was bored and wanted to mess around.

https://github.com/codyf86/PyCCMiner

I wrote a simple API interface in Python 3. I know everyone just cringed when they read Python 3, but you can easily convert it to work on Python 2, but come-on just install Python 3 already it's been 10 years nearly...as everyone who uses linux cringes. =)

Use PyCCMiner.py to be able to input console commands to send. Type help for a list of options.

Can also use PyCCMiner.py -C <command> to constantly send one command that refreshes.

Right now it just connects to port 4068 under local host (which you can easily change) and does a GET request.

If you have an android device can install Python for android and run it from console can be pretty nifty. Slowly doing more to learn more python so will update it when I can.

[Epsylon3]

What parameters would be best to lower gpu load?  I browsed the readme but wasn't sure which would be the correct ones.  Each coin and applicable miner seems to have different settings.  I typically have a low settings bat I like to use when I'm on my pc so it's still useable for watching movies or whatever and then a full blast bat when I'm away.

-i (range is 10-30 for blake, no memory used)

[SparkyU]

-i (range is 10-30 for blake, no memory used)

Ah yes, that does the trick.  Thank you much.

One more noob question.  Why is the decred miner using strictly all the gpu and not any of its memory?  The eth cuda miner uses a significant portion of the gpu memory and doesn't seem to make a video card run as hot under a similar load.

[Epsylon3]

blake is pure compute, and only a single algo (no result to pass to another one like X11).

eth is a "memohash" algo, a lot of work data is "required" to hash faster (More than 1 GB)... too big imo
Your card will mostly wait the data read...

[tbearhere]

ccminer-1.7.3-tpruvot-decred is only showing 25% of my hash at suprnova stratum port.  Is it his pool?
cmd window shows correctly 100% accepts.
EDIT: Must be his pool...showing 400% hashrate now.
Going back to getwork protocol.

[tbearhere]

tpruvot...... where is your page to download ccminer-1.7.3-tpruvot?  thx

[sp_]

What's wrong with decred and yiimp? 72.1 gh/s    should give a block every 4 hours, but it has been over 24 hours since the last block was found.

everybody is using release 1.7.3

Version   Count   Donators   Extranonce   Percent   Hashrate*   Reject
ccminer/1.7.3   31   0   31   100%   73 gh/s   0%
Total   31   0   31      73 gh/s   0%

http://yiimp.ccminer.org/site/block?id=1574

[Epsylon3]

nothing wrong, i double checked the wallet one hour ago, diff is too high i guess :/

[restless]

IMHO 1.7.3 is >10% slower.

PS: see 2 posts below, I was comparing different versions (32bit vs 64bit)

[chrysophylax]

IMHO 1.7.3 is >10% slower.

with what? ...

algos? ... performance mining? ... compiling? ... what? ...

you make a statement like that - but dont clarify what you are saying OR what you are comparing it to ...

my car is 30% faster ... there you go ... ...

#crysx

[restless]

Well, I was comparing 1.7.2 x64 for Cuda 7.5 vs 1.7.3 x86 for Cuda 6.5

for me x64 Cuda 7.5 builds give ~ 440MH/S at decred with 750Ti on Win8.1
x86 Cuda 6.5 versions give ~ 400MH/s

[pallas]

it is known cuda 7.5 gives a slight advantage on blake-based algos.

[chrysophylax]

Well, I was comparing 1.7.2 x64 for Cuda 7.5 vs 1.7.3 x86 for Cuda 6.5

for me x64 Cuda 7.5 builds give ~ 440MH/S at decred with 750Ti on Win8.1
x86 Cuda 6.5 versions give ~ 400MH/s

kool ...

i find the same also - though my builds are in fedora 23 x64 cuda 7.5 ...

with most algos now - there is no need to go back to compiling with cuda 6.5 ...

in a few months - nvidia will upgrade and release the next cuda version when they release the pascal chipset ...

this should make cuda 6.5 obsolete soon after ...

#crysx

[Epsylon3]

Indeed, CUDA 8.0 is coming soon (already referenced in 363.91 beta driver in the OpencCL version suffix)

Fine tuning will be made on it... If the perf is good enough compared to the 7.x

[sp_]

nothing wrong, i double checked the wallet one hour ago, diff is too high i guess :/

The diff is the same as it was yesterday. The last block mined @ 258581.70751273. With the current speed yiimp should hit a block every 4 hours. Could be extremely bad luck, but I doubt it..
Another thing. Try to run a single 750ti on the pool. The diff seems to be stuck at 0.75 and the reported hashrate is 1/5th of the actual rate. if you run 3 or more cards The rates are normal.

[Epsylon3]

if you are not happy with the default diff, use d=<value> in your password

it was made to allow a lower diff to the cpu miners...

else at the last block found, there was 100GH/s

[sp_]

Something has happened with your pool. All my workers are still connected and they are hashing, but the diff has dropped to 0.75 and the hashrate reported on the pool has been cut to 1/3 - 1/4.
I think there is a bug in the vardiff / stratum implementation.. When I was doing 15GHASH, the diff was around 16-32 on each of my workers. I am mining with (1.7.3). The miner is showing the correct hashrate and shares are submitted and accepted.

[Epsylon3]

i just changed the default diff to 1, instead of 0.75, which is more normal... it was set to a default of 0.5 before...

i never modified the vardiff system in yiimp... i never tried to understand it neither for the moment, there was more important to do.

PS: there is no way to tweak user hashrate in yiimp... and for me the pools doing that are wrong...

[sp_]

All my workers are back to normal speeds now. You can also see that the decreed hashrate has doubled on the poolside with the same number of workers. (60GHASH)
I guess restarting the pool helped.