CCminer(SP-MOD) Modded GPU kernels.

[rednoW]

Sorry, i'm not strong neither in gcc nor in osx ... maybe this thread will help http://stackoverflow.com/questions/14898003/enable-openmp-support-in-xcode-4-6

[pallas]

A closed case will need very powerful fans, drawing a lot of power, decreasing global efficiency.

[bathrobehero]

A closed case will need very powerful fans, drawing a lot of power, decreasing global efficiency.

Yep. Closed case > small space for airflow > needs more pressure.

8 chassis cooling fans that are 24W each

[michaelbippom]

I successfully compiled this build on OSX 10.10 and hashing away on a GTX 970 if anyone is interested in this build!

[Fuzzbawls]

I successfully compiled this build on OSX 10.10 and hashing away on a GTX 970 if anyone is interested in this build!

Yeah i'd like to take a look

[sp_]

Submitted an optimalization in neoscrypt.

7% faster on the 750ti  (+11KHASH)
4% faster on the 970 (+20KHASH)

[fenomenhaa]

Submitted an optimalization in neoscrypt.

7% faster on the 750ti  (+11KHASH)
4% faster on the 970 (+20KHASH)

Is there any tutarial to create a miner with latest commit? I knew that i need cuda,visial basic and cc miner .I downloaded all but i don2t know how to do ))

[djm34]

Submitted an optimalization in neoscrypt.

7% faster on the 750ti  (+11KHASH)
4% faster on the 970 (+20KHASH)

Is there any tutarial to create a miner with latest commit? I knew that i need cuda,visial basic and cc miner .I downloaded all but i don2t know how to do ))

good luck with visual basic

[pallas]

Submitted an optimalization in neoscrypt.

7% faster on the 750ti  (+11KHASH)
4% faster on the 970 (+20KHASH)

Works fine, thanks! :-)

[sp_]

-neoscrypt 7% faster on the 750ti and 4% faster on the gtx 970
-fixed crash in neoscrypt
-Stop mining when disconnecting (fixed by pallas)
-tiny speedup quark


1.5.57(sp-MOD) is available here: (25-07-2015)

https://github.com/sp-hash/ccminer/releases/

The sourcecode is available here:

https://github.com/sp-hash/ccminer

[sp_]

Submitted an optimalization in neoscrypt.

7% faster on the 750ti  (+11KHASH)
4% faster on the 970 (+20KHASH)

Is there any tutarial to create a miner with latest commit? I knew that i need cuda,visial basic and cc miner .I downloaded all but i don2t know how to do ))

1. Download and install visualstudio 2013 express (free)
2. install cuda 6.5
3. download the sourcecode from git.
4. Open the project file (.sln)
5. Select release in the combobox on the top.
6. Press play

[Epsylon3]

I successfully compiled this build on OSX 10.10 and hashing away on a GTX 970 if anyone is interested in this build!

Congrats, and tx for the beer biggest one i received this year...

[fenomenhaa]

-neoscrypt 7% faster on the 750ti and 4% faster on the gtx 970
-fixed crash in neoscrypt
-Stop mining when disconnecting (fixed by pallas)
-tiny speedup quark


1.5.57(sp-MOD) is available here: (25-07-2015)

https://github.com/sp-hash/ccminer/releases/

The sourcecode is available here:

https://github.com/sp-hash/ccminer

This realase quick test, 750Ti +%3,with 970 -%6 decrease ...

[fenomenhaa]

Submitted an optimalization in neoscrypt.

7% faster on the 750ti  (+11KHASH)
4% faster on the 970 (+20KHASH)

Is there any tutarial to create a miner with latest commit? I knew that i need cuda,visial basic and cc miner .I downloaded all but i don2t know how to do ))

1. Download and install visualstudio 2013 express (free)
2. install cuda 6.5
3. download the sourcecode from git.
4. Open the project file (.sln)
5. Select release in the combobox on the top.
6. Press play

Didn't understand  stage 5 ) what is combobox and where is relase ? )) Hope it sounds not sooo fool ))

[sp_]

This realase quick test, 750Ti +%3,with 970 -%6 decrease ...

my zotac 970 is trottling I need to finetune it on the gigabyte later. Note that if you run with a -i parameter this needs to be finetuned. the kernal is only running with 32 threads per block.(was 128 before) so you can increase the intensity abit.

[sp_]

Didn't understand  stage 5 ) what is combobox and where is relase ? )) Hope it sounds not sooo fool ))

just open up visual studio. there will be a white box in the top that says "debug" you have to change it to release.

[fenomenhaa]

This realase quick test, 750Ti +%3,with 970 -%6 decrease ...

my zotac 970 is trottling I need to finetune it on the gigabyte later. Note that if you run with a -i parameter this needs to be finetuned. the kernal is only running with 32 threads per block.(was 128 before) so you can increase the intensity abit.

İ edited tdp to 230W on my 970 's according to gpuz 970 's core clock is stable and there is no throttling.i didn't use -i parameter ,default settings...And quark is better on 1.5.56 all on my rigs.Is there  way to know deafult intensity on miner?

[scryptr]

This realase quick test, 750Ti +%3,with 970 -%6 decrease ...

my zotac 970 is trottling I need to finetune it on the gigabyte later. Note that if you run with a -i parameter this needs to be finetuned. the kernal is only running with 32 threads per block.(was 128 before) so you can increase the intensity abit.

İ edited tdp to 230W on my 970 's according to gpuz 970 's core clock is stable and there is no throttling.i didn't use -i parameter ,default settings...And quark is better on 1.5.56 all on my rigs.Is there  way to know deafult intensity on miner?

FEATURE REQUEST--

When CCminer is started at default intensity, with no "-i" parameter set, it would be very useful to the user if the default intensity is displayed on the console, i.e.  "... 24, XXXXXXX threads, (DEFAULT)" along with the other standard startup readings.  The author(s) would not have to maintain a chart of current default values per algo for user reference.

I will donate again if this feature is implemented.  The intensity setting is a user-adjustable parameter that seems to have real effect at times, and can be set for the user's personal equipment.  That being said, the programmer should set the default value conservatively, to work well with many different rigs.       --scryptr

[Epsylon3]

When CCminer is started at default intensity, with no "-i" parameter set, it would be very useful to the user if the default intensity is displayed on the console, i.e.  "... 24, XXXXXXX threads, (DEFAULT)" along with the other standard startup readings.  The author(s) would not have to maintain a chart of current default values per algo for user reference.

I will donate again if this feature is implemented.  The intensity setting is a user-adjustable parameter that seems to have real effect at times, and can be set for the user's personal equipment.  That being said, the programmer should set the default value conservatively, to work well with many different rigs.       --scryptr

i agree, most of the time, it's the difference between our forks... I let it a bit less than sp (or djm) which often set the max possible.... to be more compatible.
The problem is they dont tune the intensity but the number of cuda threads directly... and djm34 doesnt use at all the device_intensity() function made for that...

But im also in late on some algos... still dont have picked djm lyra2 changes...

[djm34]

When CCminer is started at default intensity, with no "-i" parameter set, it would be very useful to the user if the default intensity is displayed on the console, i.e.  "... 24, XXXXXXX threads, (DEFAULT)" along with the other standard startup readings.  The author(s) would not have to maintain a chart of current default values per algo for user reference.

I will donate again if this feature is implemented.  The intensity setting is a user-adjustable parameter that seems to have real effect at times, and can be set for the user's personal equipment.  That being said, the programmer should set the default value conservatively, to work well with many different rigs.       --scryptr

i agree, most of the time, it's the difference between our forks... I let it a bit less than sp (or djm) which often set the max possible.... to be more compatible.
The problem is they dont tune the intensity but the number of cuda threads directly... and djm34 doesnt use at all the device_intensity() function made for that...

intensity is bullcrap..  
sometime democracy is a bad thing... well intensity is one of those time.
first fractional intensity  (I mean which isn't a multiple of the thread per block) is a total non sense...
some people here tends to see 50kh/s difference to algo running at several 10's of MH/s, if you use basic math background
you will know that it is most likely not significant at all (and running several time, will probably brings different results)

So no, no way, I encourage implementing a feature which goes against kernel tuning (reason I also remove another  line of code, I see from time to time )
and favors the expression of non significant result.