CCminer(SP-MOD) Modded GPU kernels.

[sp_]

- merged with klaust fork
- improved blake,keccak and skein

The 28th public beta(windows executable) is available here: (10-jan-2015)

http://www.filedropper.com/release28

The sourcecode is available here:

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

[flipclip]

I believe quark consumes roughly (less then a watt difference) the same as x11.  I haven't noticed any power fluctuations from my kill-a-watt from recent version updates.

x14 can be profitable but it is pretty short lived (as you noticed).

Yeah I don't think so.  Quark uses about 50 more watts on a 6 card rig.  When mine switches to quark it's around 425w and x11 is 375w.

That's not my experience. Are you changing clocks or something?

My 6x 750ti rig uses about 400-405 watts on x11 or Quark. I just switched it back and forth to double-check. There is barely a change in the wattage. Cards are all running at max stable overclock.  
I'm getting 36,360 kh/s on Quark with v27 btw

Nope. Same watt meter as I've had on my control rig since the beginning.  Maybe since they're evga FTW cards they're different, but all past miners and algos they've been consistent and only used slightly more than my gigabyte Ti's.  H81 btc board, gold psu, i3 cpu.

With v28 I just switched between quark and x11, no watt change.  

2 750 Ti's v28= Quark ~11,160 kh/s , x11 ~5,750 kh/s

[sp_]

my numbers. Rigs are running at standard clocks mining quark

testrig 1: 6 750ti(non powered)

release 24
31500

release 28
32100

testrig 2: 4 cards. (1 gtx980, 1gtx970, 2 750ti.)

release 24
38500

release 28
39500

[sp_]

If you compiling yourself on 970 and 980 cards, remember to add compute_52 and sm_52 in the project file. Default on github is compute 5.0 only (750ti)

[go6ooo1212]

I expect couple of testing samples of GTX960 next week. Would be interesting what could they do

[sp_]

I expect couple of testing samples of GTX960 next week. Would be interesting what could they do

Thats early.. Could you send me one?

[scryptr]

- merged with klaust fork
- improved blake,keccak and skein

The 28th public beta(windows executable) is available here: (10-jan-2015)

http://www.filedropper.com/release28

The sourcecode is available here:

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

SP_ v28 beta-

Sorry to report this, but on both of my rigs I got the following error when attempting to compile v28:

"cuda_myriadgroestl.o:(.bss+0x0): multiple definition of `d_resultNonce'
Algo256/cuda_fugue256.o:(.bss+0x0): first defined here
collect2: error: ld returned 1 exit status
make[2]: *** [ccminer] Error 1
make[2]: Leaving directory `/home/borqd/spminer'
make[1]: *** [all-recursive] Error 1
make[1]: Leaving directory `/home/borqd/spminer'
make: *** [all] Error 2"

There was no executable "ccminer" left in the directory.  I am still running v26, which appears very stable.  It went over 4000 accepts on each rig with less than 1% rejects.       --scryptr

[flipclip]

- merged with klaust fork
- improved blake,keccak and skein

The 28th public beta(windows executable) is available here: (10-jan-2015)

http://www.filedropper.com/release28

The sourcecode is available here:

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

SP_ v28 beta-

Sorry to report this, but on both of my rigs I got the following error when attempting to compile v28:

"cuda_myriadgroestl.o:(.bss+0x0): multiple definition of `d_resultNonce'
Algo256/cuda_fugue256.o:(.bss+0x0): first defined here
collect2: error: ld returned 1 exit status
make[2]: *** [ccminer] Error 1
make[2]: Leaving directory `/home/borqd/spminer'
make[1]: *** [all-recursive] Error 1
make[1]: Leaving directory `/home/borqd/spminer'
make: *** [all] Error 2"

There was no executable "ccminer" left in the directory.  I am still running v26, which appears very stable.  It went over 4000 accepts on each rig with less than 1% rejects.       --scryptr

To fix this edit cuda_myriadgroestl.cu (in ccminer's root folder).  
On line eleven add: extern
the line should read:  extern uint2 *d_resultNonce[8];

I made that edit, compiled without issue, and have been hashing away 2 hours without issue.

(This line was changed to its current iteration of lacking "extern" 1/5/14 via commit: 462dd9bc99a255e4b593f82b1237ce152d55cb59)

[Schleicher]

SP_ v28 beta-

Sorry to report this, but on both of my rigs I got the following error when attempting to compile v28:

"cuda_myriadgroestl.o:(.bss+0x0): multiple definition of `d_resultNonce'
Algo256/cuda_fugue256.o:(.bss+0x0): first defined here
collect2: error: ld returned 1 exit status
make[2]: *** [ccminer] Error 1
make[2]: Leaving directory `/home/borqd/spminer'
make[1]: *** [all-recursive] Error 1
make[1]: Leaving directory `/home/borqd/spminer'
make: *** [all] Error 2"

There was no executable "ccminer" left in the directory.  I am still running v26, which appears very stable.  It went over 4000 accepts on each rig with less than 1% rejects.       --scryptr

To fix this edit cuda_myriadgroestl.cu (in ccminer's root folder).  
On line eleven add: extern
the line should read:  extern uint2 *d_resultNonce[8];

I made that edit, compiled without issue, and have been hashing away 2 hours without issue.

(This line was changed to its current iteration of lacking "extern" 1/5/14 via commit: 462dd9bc99a255e4b593f82b1237ce152d55cb59)

Wierd, there's no error when compiling with Visual Studio.
Anyway, better use static instead of extern

[scryptr]

- merged with klaust fork
- improved blake,keccak and skein

The 28th public beta(windows executable) is available here: (10-jan-2015)

http://www.filedropper.com/release28

The sourcecode is available here:

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

SP_ v28 beta-

Sorry to report this, but on both of my rigs I got the following error when attempting to compile v28:

"cuda_myriadgroestl.o:(.bss+0x0): multiple definition of `d_resultNonce'
Algo256/cuda_fugue256.o:(.bss+0x0): first defined here
collect2: error: ld returned 1 exit status
make[2]: *** [ccminer] Error 1
make[2]: Leaving directory `/home/borqd/spminer'
make[1]: *** [all-recursive] Error 1
make[1]: Leaving directory `/home/borqd/spminer'
make: *** [all] Error 2"

There was no executable "ccminer" left in the directory.  I am still running v26, which appears very stable.  It went over 4000 accepts on each rig with less than 1% rejects.       --scryptr

To fix this edit cuda_myriadgroestl.cu (in ccminer's root folder).  
On line eleven add: extern
the line should read:  extern uint2 *d_resultNonce[8];

I made that edit, compiled without issue, and have been hashing away 2 hours without issue.

(This line was changed to its current iteration of lacking "extern" 1/5/14 via commit: 462dd9bc99a255e4b593f82b1237ce152d55cb59)

v28 TYPO--

It worked like a charm.  Running v28, each card is hashing about +50kh/s from v26 now, about +300kh/s on each rig.  My faster rig is running about 36,267kh/s.  This rig is over 36mh/s average for the first time.  One of my cards is at 6120kh/s.

--scryptr

[squadbox]

I am still running v26, which appears very stable.  It went over 4000 accepts on each rig with less than 1% rejects.       --scryptr

I agree, v26 was also very stable for me. 11540 +/- 50 khash/s, accepted: 11422/11476 (99.53%) for two EVGA 750Ti SCs.

I just built v28.  A little improvement for Quark, 11680 +/- 50 khash/s.

[sp_]

I submitted an optimized BMW. ptx assembly is down from 36000 lines to around 34000 lines.

[sp_]

my numbers. Rigs are running at standard clocks mining quark

testrig 2: 4 cards. (1 gtx980, 1gtx970, 2 750ti.)

release 28
39500

release 28.git (faster bmw)
39700

testrig 2: 4 cards. (1 gtx980, 1gtx970, 2 750ti.)

release 28
32100

release 28.git (faster bmw)
32500

[Grim]

@ sp

Any chance for an optimized Spreadcoin ccminer?

It seems the community is willing to pay for it.

[sp_]

@ sp

Any chance for an optimized Spreadcoin ccminer?

It seems the community is willing to pay for it.

I am working on it. 30-40% faster on 980 cards.

[flipclip]

my numbers. Rigs are running at standard clocks mining quark

testrig 2: 4 cards. (1 gtx980, 1gtx970, 2 750ti.)

release 28
39500

release 28.git (faster bmw)
39700

testrig 2: 4 cards. (1 gtx980, 1gtx970, 2 750ti.)

release 28
32100

release 28.git (faster bmw)
32500

Do you mean?  (This would correspond to your previously posted test rig profiles.)

testrig 1: 6 750ti(non powered)

release 28
32100

release 28.git (faster bmw)
32500

testrig 2: 4 cards. (1 gtx980, 1gtx970, 2 750ti.)

release 28
39500

release 28.git (faster bmw)
39700

My results:
2 750 Ti's
v28= Quark ~11,115 kh/s (my earlier report was higher, but it seems to have stabilized closer to this)
v28.git = Quark ~11,130 kh/s (it seems to start around ~11,175, but then slowly goes down to ~11,1130)

[sp_]

I the hashrate is dropping, check the temps on the cards. My ti's are downclocked automaticly if they reach 70degrees(c). With windows and firestorm software you can overvolt the cards. Then they will accept temps up to 80 degrees(c)

[flipclip]

I the hashrate is dropping, check the temps on the cards. My ti's are downclocked automaticly if they reach 70degrees(c). With windows and firestorm software you can overvolt the cards. Then they will accept temps up to 80 degrees(c)

My cards stay between 55c - 60c during the day (I assume at night they get a little cooler when the house heat doesn't come on).  Neither are overclocked (I run linux and never really delved into the overclocking aspect) by me (one is a SuperClocked EVGA 750Ti though).  It might just be because I am using the computer (albeit only firefox to write this post), b/c I noticed when I sat down it was ~11180 and now it is ~11145.  No change in temp, diff, pool, etc.

None the less, great work.

[flipclip]

I the hashrate is dropping, check the temps on the cards. My ti's are downclocked automaticly if they reach 70degrees(c). With windows and firestorm software you can overvolt the cards. Then they will accept temps up to 80 degrees(c)

My cards stay between 55c - 60c during the day (I assume at night they get a little cooler when the house heat doesn't come on).  Neither are overclocked (I run linux and never really delved into the overclocking aspect) by me (one is a SuperClocked EVGA 750Ti though).  It might just be because I am using the computer (albeit only firefox to write this post), b/c I noticed when I sat down it was ~11180 and now it is ~11145.  No change in temp, diff, pool, etc.

None the less, great work.

Yeah, its from using the computer lightly... my bad.

2 750 Ti's
v28= Quark ~11,160 kh/s
v28.git = Quark ~11,185 kh/s

[Handofmidas]

Here are my release28 stats for X11. GPU0 as seen in GPU-Z, GPU1 @ 1499 MHz.