CCminer(SP-MOD) Modded NVIDIA Maxwell / Pascal kernels. - page 1240.

sp_

legendary

Activity: 2912

Merit: 1087

Team Black developer

Quote from: Epsylon3 on October 18, 2014, 06:47:50 PM

Quote from: SS2006 on October 18, 2014, 06:43:53 PM

Do you think its because the GTX 970 uses compute 5.2?
Even when I installed the CUDA 6.5 it told me 'you have no compatible card for this software'

Yep, there is a difference when compiled with 5.2, try my 5.2 binary here : https://github.com/tpruvot/ccminer/releases

Note: 1.4.5 is almost ready with a new curl, i detected some hangs problems with the 1.4.4 (mainly in solo mining)

Code:

$ ccminersp.exe --help
*** ccMiner for nVidia GPUs by Christian Buchner and Christian H. ***
This is the forked version 1.4.5 (tpruvot@github)

_sp: Thanks to take all the credits, my version was already faster than djm34 one on x11, what did you change (remember its GPL code) ?

But if you plan to release changes on github, i totally agree

I will publish the sourcecode, but it breaks compabillity with old cards, so I need some more work. I don't take all the credit. Start the miner and you willl see you name:)

Happy hashing

Dotcommie

sr. member

Activity: 476

Merit: 250

Quote from: sp_ on October 18, 2014, 03:41:44 PM

Yeah. The x13 performance is not so good. The next build will be bether. Try jackpointcoin

Have you investigated whatever is going on with jackpotcoin? I emailed you what was going on when I tried mining it. It seemed like a nice improvement, but the pools only saw about 1/8 of the reported hash in the cmd prompt.

Thanks for your development efforts, I will definitely be messing around with this tomorrow night and Monday and will see how my rigs perform. I am just using old nvminer on jackpot currently since the gf and her family are taking up all my time this weekend :-/

**Sorry, I didn't refresh before posting, I see someone else had the same issue. I'll get back to you if I find any more out tomorrow.

djm34

legendary

Activity: 1400

Merit: 1050

Quote from: Epsylon3 on October 18, 2014, 06:47:50 PM

Quote from: SS2006 on October 18, 2014, 06:43:53 PM

Do you think its because the GTX 970 uses compute 5.2?
Even when I installed the CUDA 6.5 it told me 'you have no compatible card for this software'

Yep, there is a difference when compiled with 5.2, try my 5.2 binary here : https://github.com/tpruvot/ccminer/releases

Note: 1.4.5 is almost ready with a new curl, i detected some hangs problems with the 1.4.4 (mainly in solo mining)

Code:

$ ccminersp.exe --help
*** ccMiner for nVidia GPUs by Christian Buchner and Christian H. ***
This is the forked version 1.4.5 (tpruvot@github)

_sp: Thanks to take all the credits, my version was already faster than djm34 one on x11, what did you change (remember its GPL code) ?

But if you plan to release changes on github, i totally agree

there is no djm34 version of x11 (at least not yet...).
I let you fight over which one is faster, it is funny to watch Grin

edit: I don't think there is any relationship between your version and the one of _sp who has chosen is own path
... all this for 200kh/s Roll Eyes

Epsylon3

legendary

Activity: 1484

Merit: 1082

ccminer/cpuminer developer

i dont see any improvements on X11 against my version :

Yours :

Code:

==30576== Profiling application: ccminersp.exe -a x11 --benchmark
==30576== Profiling result:
Time(%) Time Calls Avg Min Max Name
20.95% 14.5236s 296 49.066ms 48.325ms 52.197ms x11_echo512_gpu_hash_64(int, unsi
19.27% 13.3595s 297 44.982ms 44.554ms 50.817ms quark_groestl512_gpu_hash_64_quad
11.74% 8.13434s 296 27.481ms 26.976ms 30.969ms x11_shavite512_gpu_hash_64(int, u
10.63% 7.36593s 296 24.885ms 24.556ms 27.590ms x11_simd512_gpu_expand_64(int, un
6.99% 4.84835s 296 16.380ms 16.199ms 18.494ms x11_cubehash512_gpu_hash_64(int,
5.29% 3.66619s 297 12.344ms 12.201ms 13.939ms quark_jh512_gpu_hash_64(int, unsi
5.02% 3.47754s 296 11.748ms 11.322ms 13.248ms x11_luffa512_gpu_hash_64(int, uns
4.24% 2.93967s 296 9.9313ms 9.4073ms 10.335ms x11_simd512_gpu_compress2_64(int,
3.51% 2.43131s 296 8.2139ms 8.0358ms 8.5024ms x11_simd512_gpu_compress1_64(int,
2.87% 1.99201s 297 6.7071ms 6.3100ms 7.6564ms quark_bmw512_gpu_hash_64(int, uns
2.69% 1.86559s 297 6.2814ms 6.1864ms 7.1012ms quark_skein512_gpu_hash_64(int, u
2.66% 1.84484s 297 6.2116ms 6.1330ms 7.0152ms quark_keccak512_gpu_hash_64(int,
2.38% 1.65098s 297 5.5589ms 5.4953ms 6.2752ms quark_blake512_gpu_hash_80(int, u
1.47% 1.01983s 296 3.4454ms 3.3832ms 3.6530ms x11_simd512_gpu_final_64(int, uns
0.28% 195.46ms 295 662.57us 652.23us 699.94us cuda_check_gpu_hash_64(int, unsig
0.00% 875.13us 295 2.9660us 2.4000us 12.479us [CUDA memset]
0.00% 627.25us 295 2.1260us 1.3430us 9.0870us [CUDA memcpy DtoH]
0.00% 63.389us 82 773ns 512ns 1.2480us [CUDA memcpy HtoD]

and mine :

Code:

Time(%) Time Calls Avg Min Max Name
20.80% 6.36219s 146 43.577ms 42.978ms 44.897ms x11_echo512_gpu_hash_64(int, unsigned int, __int64*, unsigned int*)
18.86% 5.77071s 146 39.525ms 39.119ms 44.727ms quark_groestl512_gpu_hash_64_quad(int, unsigned int, unsigned int*, unsigned int*)
11.66% 3.56600s 146 24.425ms 24.071ms 27.458ms x11_shavite512_gpu_hash_64(int, unsigned int, __int64*, unsigned int*)
10.59% 3.24123s 146 22.200ms 21.875ms 24.802ms x11_simd512_gpu_expand_64(int, unsigned int, __int64*, unsigned int*, uint4*)
7.53% 2.30443s 146 15.784ms 15.575ms 17.894ms x11_cubehash512_gpu_hash_64(int, unsigned int, __int64*, unsigned int*)
5.30% 1.62276s 146 11.115ms 10.976ms 12.522ms quark_jh512_gpu_hash_64(int, unsigned int, __int64*, unsigned int*)
4.99% 1.52627s 146 10.454ms 10.235ms 11.892ms x11_luffa512_gpu_hash_64(int, unsigned int, __int64*, unsigned int*)
4.21% 1.28899s 146 8.8287ms 8.4681ms 9.1675ms x11_simd512_gpu_compress2_64(int, unsigned int, __int64*, unsigned int*, uint4*, int*)
3.48% 1.06443s 146 7.2906ms 6.9558ms 7.5275ms x11_simd512_gpu_compress1_64(int, unsigned int, __int64*, unsigned int*, uint4*, int*)
3.05% 934.36ms 147 6.3562ms 6.2786ms 7.1199ms quark_bmw512_gpu_hash_64(int, unsigned int, __int64*, unsigned int*)
2.68% 820.56ms 146 5.6203ms 5.5330ms 6.3415ms quark_skein512_gpu_hash_64(int, unsigned int, __int64*, unsigned int*)
2.64% 807.37ms 146 5.5299ms 5.4526ms 6.2354ms quark_keccak512_gpu_hash_64(int, unsigned int, __int64*, unsigned int*)
2.63% 804.03ms 147 5.4696ms 5.4003ms 6.1889ms quark_blake512_gpu_hash_80(int, unsigned int, void*)
1.46% 446.72ms 146 3.0597ms 3.0059ms 3.1561ms x11_simd512_gpu_final_64(int, unsigned int, __int64*, unsigned int*, uint4*, int*)
0.11% 33.535ms 145 231.28us 214.59us 265.21us cuda_check_gpu_hash_64(int, unsigned int, unsigned int*, unsigned int*, unsigned int*)
0.00% 428.88us 145 2.9570us 2.4000us 9.4720us [CUDA memset]
0.00% 320.11us 145 2.2070us 1.3440us 9.4390us [CUDA memcpy DtoH]
0.00% 77.406us 110 703ns

tbearhere

legendary

Activity: 3164

Merit: 1003

Quote from: sp_ on October 18, 2014, 03:41:44 PM

Yeah. The x13 performance is not so good. The next build will be bether. Try jackpointcoin

great wow

but hashatme has my worker difficulty set so high im throwing in 50mh/s and that pool is showing 10mh/s.

SS2006

sr. member

Activity: 285

Merit: 250

Actually thats the one I've been using for weeks ya

I thought this might even add more to it

So is 6200 my max or do u see further improvements in yours?

ive rread in other threads that having the miner keep the card in P0 state, and memory up to 3500 helps. I was able to replicate this iwth 'nvidia inspector' and saw some improvement, but it makes my computer hang too much

Epsylon3

legendary

Activity: 1484

Merit: 1082

ccminer/cpuminer developer

Quote from: SS2006 on October 18, 2014, 06:43:53 PM

Do you think its because the GTX 970 uses compute 5.2?
Even when I installed the CUDA 6.5 it told me 'you have no compatible card for this software'

Yep, there is a difference when compiled with 5.2, try my 5.2 binary here : https://github.com/tpruvot/ccminer/releases

Note: 1.4.5 is almost ready with a new curl, i detected some hangs problems with the 1.4.4 (mainly in solo mining)

Code:

$ ccminersp.exe --help
*** ccMiner for nVidia GPUs by Christian Buchner and Christian H. ***
This is the forked version 1.4.5 (tpruvot@github)

_sp: Thanks to take all the credits, my version was already faster than djm34 one on x11, what did you change (remember its GPL code) ?

But if you plan to release changes on github, i totally agree

SS2006

sr. member

Activity: 285

Merit: 250

Unfortunately there is no improvement on my EVGA GTX 970 Sad

Before OC I was getting 5700 KH/S and after OC around 6100 KH/S

with this new ccminer it is the same

Do you think its because the GTX 970 uses compute 5.2?
Even when I installed the CUDA 6.5 it told me 'you have no compatible card for this software'

Epsylon3

legendary

Activity: 1484

Merit: 1082

ccminer/cpuminer developer