Nexus - Pure SHA3 + CPU/GPU + nPoS + 15 Active Innovations + More to Come

[MisO69]

Tested on one of my 280x rigs 6 GPU, Windows 8.1, Celeron processor.

Installed the vcredis_x86 2013.

CPU pinned at 100%. With 1 thread or two, same results. I'ma try tweaking some of the other parameters such as intensity.

[ncunt]

Is there a link for the windows gpu miner?

here my dear:
https://bitcointalk.org/index.php?topic=657601.msg9287828#msg9287828

[djm34]

Tested on one of my 280x rigs 6 GPU, Windows 8.1, Celeron processor.

Installed the vcredis_x86 2013.

CPU pinned at 100%. With 1 thread or two, same results. I'ma try tweaking some of the other parameters such as intensity.

which config file are you tweaking the one in the same directory as the exe or the one in Release/Ressource/Config (because it is the second one you need to tweak. the first one is a something I forgot to clean...)
May-be I should release a new version without that one for those who really don't want to read what I wrote.  

[hero18688]

Thanks , Ill check out
EDIT: Stabilized it at 145 MH/s with 5x280X Vapor-X, but did not find any block.
I have a rig of Nvidia which mining with 110MH/s and hits blocks ...

It should work, in principle, I will check again (to make sure I didn't mess with before releasing) if the nonce doesn't skip any value...

What's the conf of 750ti for your previous released miner,I got only around 19mh/s for each 750ti.

I was getting 28mh/s with a mild oc.
Did you compile it yourself or is this a binaries you found here ?
If you compile it yourself, make sure you use compute_50 (however the driver should take care of that...)

No,I just using the windows build I found here.I have oc to 1350mhz.

[djm34]

Thanks , Ill check out
EDIT: Stabilized it at 145 MH/s with 5x280X Vapor-X, but did not find any block.
I have a rig of Nvidia which mining with 110MH/s and hits blocks ...

It should work, in principle, I will check again (to make sure I didn't mess with before releasing) if the nonce doesn't skip any value...

What's the conf of 750ti for your previous released miner,I got only around 19mh/s for each 750ti.

I was getting 28mh/s with a mild oc.
Did you compile it yourself or is this a binaries you found here ?
If you compile it yourself, make sure you use compute_50 (however the driver should take care of that...)

No,I just using the windows build I found here.I have oc to 1350mhz.

I will have an update tomorrow together with a new keccak kernel compatible with compute 5_2, will see if it's boost the performance on 50 too...
Also the possibility to chose cards.
and also a kernel working for compute_30 (but don't expect miracle at that level the performance are, there is no other word, terrible...)

[hero18688]

Thanks , Ill check out
EDIT: Stabilized it at 145 MH/s with 5x280X Vapor-X, but did not find any block.
I have a rig of Nvidia which mining with 110MH/s and hits blocks ...

It should work, in principle, I will check again (to make sure I didn't mess with before releasing) if the nonce doesn't skip any value...

What's the conf of 750ti for your previous released miner,I got only around 19mh/s for each 750ti.

I was getting 28mh/s with a mild oc.
Did you compile it yourself or is this a binaries you found here ?
If you compile it yourself, make sure you use compute_50 (however the driver should take care of that...)

No,I just using the windows build I found here.I have oc to 1350mhz.

I will have an update tomorrow together with a new keccak kernel compatible with compute 5_2, will see if it's boost the performance on 50 too...
Also the possibility to chose cards.
and also a kernel working for compute_30 (but don't expect miracle at that level the performance are, there is no other word, terrible...)

Is this the total hashrate of my 4*750ti or just for each one?

[go6ooo1212]

Its the total hashrate of all 4 cards.

[djm34]

Its the total hashrate of all 4 cards.

yes and the very big number is the bug... (the problem with those c++ class is that at some point you always end up a non initialized variable... even though, I reset it all the time...)

[hoze]

Its the total hashrate of all 4 cards.

yes and the very big number is the bug... (the problem with those c++ class is that at some point you always end up a non initialized variable... even though, I reset it all the time...)

I have 6 x 750ti and only 4 GPU working. Do U know maybe what is wrong? It's interesting that GPU 1,2,3 and 6 work and GPU 4,5 not . Total hashrate is ~68Mhs

[go6ooo1212]

Its the total hashrate of all 4 cards.

yes and the very big number is the bug... (the problem with those c++ class is that at some point you always end up a non initialized variable... even though, I reset it all the time...)

I have 6 x 750ti and only 4 GPU working. Do U know maybe what is wrong? It's interesting that GPU 1,2,3 and 6 work and GPU 4,5 not . Total hashrate is ~68Mhs

give as more info:
miner version , command line , driver ver..

[hoze]

Its the total hashrate of all 4 cards.

yes and the very big number is the bug... (the problem with those c++ class is that at some point you always end up a non initialized variable... even though, I reset it all the time...)

I have 6 x 750ti and only 4 GPU working. Do U know maybe what is wrong? It's interesting that GPU 1,2,3 and 6 work and GPU 4,5 not . Total hashrate is ~68Mhs

give as more info:
miner version , command line , driver ver..

windows 8.1 , skminer v0.1, skminer.exe 127.0.0.1 9325 6 , driver 344.48

[djm34]

reviewing a bit the numbers, my 750ti is doing 19.5MH/s 22.5MH/s and not 28MH/s (not sure where I got these actually...).
This number is compatible with kernel time measurement (266ms 231ms for 5.24M threads) and makes more sense compared to the 780ti (41MH/s 43MH/s) or the 980 (53MH/s 63MH/s)

[MisO69]

Tested on one of my 280x rigs 6 GPU, Windows 8.1, Celeron processor.

Installed the vcredis_x86 2013.

CPU pinned at 100%. With 1 thread or two, same results. I'ma try tweaking some of the other parameters such as intensity.

which config file are you tweaking the one in the same directory as the exe or the one in Release/Ressource/Config (because it is the second one you need to tweak. the first one is a something I forgot to clean...)
May-be I should release a new version without that one for those who really don't want to read what I wrote.  

Damn, thats what I get for being lazy. I'll try again. Just assumed it was the one in the same folder as the miner.exe

[MisO69]

Can't get miner to work no matter how low I go on the settings. It does identify the cards when the miner starts up but no usage on the actual GPUs. Also the CPU is still pinned at 99% usage. I am using the 13.1 drivers so maybe that is the problem?

280x 6 GPUs on windows 8.1 using the microsoft drivers for the GPUs.

platform name Advanced Micro Devices, Inc. index 0
threadconcurrency 600
threadconcurrency 600
threadconcurrency 600
threadconcurrency 600
threadconcurrency 600
threadconcurrency 600
ADL Overdrive 5 Was Found!
ADL Overdrive 6 Was Found!
ADL index 0, id 12388992 - BIOS partno.: 113-2104XTLAFR-V45, version: 015.040.00
0.001, date: 2013/10/17 06:22
ADL index 1, id 12388992 - BIOS partno.: 113-2104XTLAFR-V45, version: 015.040.00
0.001, date: 2013/10/17 06:22
ADL index 2, id 12388992 - BIOS partno.: 113-2104XTLAFR-V45, version: 015.040.00
0.001, date: 2013/10/17 06:22
ADL index 3, id 12388992 - BIOS partno.: 113-2104XTLAFR-V45, version: 015.040.00
0.001, date: 2013/10/17 06:22
ADL index 4, id 12388992 - BIOS partno.: 113-2104XTLAFR-V45, version: 015.040.00
0.001, date: 2013/10/17 06:22
ADL index 5, id 15115712 - BIOS partno.: 113-2104XTLAFR-V44, version: 015.039.00
0.001, date: 2013/09/02 23:08
vecgpudata size 6

                ********* Coinshield sk-1024 miner for amd GPUs by djm34 *******
**
                                     original design by bitslapper 2014

                                           Copyright 2014 djm34

                donation address CSD: 2S2pCpRXyb8Lpre52U3Xjq2MguSdaea5YGjVTsJqgZ
BfL2S24ag

Connected to 127.0.0.1:9325...
[MASTER] Coinshield Network | New Block 35040
[METERS] 0 Hashes | 0.000000 KHash/s | Height = 35040
[METERS] 0 Hashes | 0.000000 KHash/s | Height = 35040

[djm34]

Can't get miner to work no matter how low I go on the settings. It does identify the cards when the miner starts up but no usage on the actual GPUs. Also the CPU is still pinned at 99% usage. I am using the 13.1 drivers so maybe that is the problem?

280x 6 GPUs on windows 8.1 using the microsoft drivers for the GPUs.

platform name Advanced Micro Devices, Inc. index 0
threadconcurrency 600
threadconcurrency 600
threadconcurrency 600
threadconcurrency 600
threadconcurrency 600
threadconcurrency 600
ADL Overdrive 5 Was Found!
ADL Overdrive 6 Was Found!
ADL index 0, id 12388992 - BIOS partno.: 113-2104XTLAFR-V45, version: 015.040.00
0.001, date: 2013/10/17 06:22
ADL index 1, id 12388992 - BIOS partno.: 113-2104XTLAFR-V45, version: 015.040.00
0.001, date: 2013/10/17 06:22
ADL index 2, id 12388992 - BIOS partno.: 113-2104XTLAFR-V45, version: 015.040.00
0.001, date: 2013/10/17 06:22
ADL index 3, id 12388992 - BIOS partno.: 113-2104XTLAFR-V45, version: 015.040.00
0.001, date: 2013/10/17 06:22
ADL index 4, id 12388992 - BIOS partno.: 113-2104XTLAFR-V45, version: 015.040.00
0.001, date: 2013/10/17 06:22
ADL index 5, id 15115712 - BIOS partno.: 113-2104XTLAFR-V44, version: 015.039.00
0.001, date: 2013/09/02 23:08
vecgpudata size 6

                ********* Coinshield sk-1024 miner for amd GPUs by djm34 *******
**
                                     original design by bitslapper 2014

                                           Copyright 2014 djm34

                donation address CSD: 2S2pCpRXyb8Lpre52U3Xjq2MguSdaea5YGjVTsJqgZ
BfL2S24ag

Connected to 127.0.0.1:9325...
[MASTER] Coinshield Network | New Block 35040
[METERS] 0 Hashes | 0.000000 KHash/s | Height = 35040
[METERS] 0 Hashes | 0.000000 KHash/s | Height = 35040

probably the driver (honestly I don't know for sure), and how many core has your cpu ?
I don't think it can support more card than there are cpu thread

[djm34]

new sources and binaries for nvidia miner: http://ge.tt/4dUHeF32/v/0?c

main change:
* support added for compute_30 (the performance are bad... but it works... not enough register for skein 1024 leading to large spill store
* kernel working now with compute 5_2
* optimization of skein for compute >= 3_5
gtx750: 22.5MH/s
gtx780ti: 43MH/s
gtx 980: 63MH/s
* possibility to select the card to be used together with new format for IP:PORT selection

example Release/skminer 127.0.0.1:9325 gtx980#1,gtx980#2,gtxWhatever

(in the example, the program will run on two gtx980 and ignore the gtxWhatever as it won't find it on your computer (so make sure you put names of card which exist)... alternatively 980#1,980#2 will also work (as in ccminer).

If no second argument is given, it will use every card found on the system.
There is may-be a limitation, as I think, the number of card should be smaller or equal to the number of cpu core/thread (need to be verified though... and I don't have enough pci port for that...)

* new display format, I used supercomputing one.
* added compilation for linux (it is again supercomputing one...) please let me know if it works.


current hashrate on my system with this version: 193MH/s (2 gtx980, 1 gtx750 and 1 gtx780ti), cpu usage 26% (which is high compared to ccminer)

[skunk]

new sources and binaries for nvidia miner: http://ge.tt/4dUHeF32/v/0?c

main change:
* support added for compute_30 (the performance are bad... but it works... not enough register for skein 1024 leading to large spill store
* kernel working now with compute 5_2
* optimization of skein for compute >= 3_5
gtx750: 22.5MH/s
gtx780ti: 43MH/s
gtx 980: 63MH/s
* possibility to select the card to be used together with new format for IP:PORT selection

example Release/skminer 127.0.0.1:9325 gtx980#1,gtx980#2,gtxWhatever

(in the example, the program will run on two gtx980 and ignore the gtxWhatever as it won't find it on your computer (so make sure you put names of card which exist)... alternatively 980#1,980#2 will also work (as in ccminer).

If no second argument is given, it will use every card found on the system.
There is may-be a limitation, as I think, the number of card should be smaller or equal to the number of cpu core/thread (need to be verified though... and I don't have enough pci port for that...)

* new display format, I used supercomputing one.
* added compilation for linux (it is again supercomputing one...) please let me know if it works.


current hashrate on my system with this version: 193MH/s (2 gtx980, 1 gtx750 and 1 gtx780ti), cpu usage 26% (which is high compared to ccminer)

hi djm34,
after symlinking every hash/Keccak*.c with hash/Keccak*.cpp, i get this error:

Code:

g++ -DHAVE_CONFIG_H -I.  -msse2  -fopenmp -fno-strict-aliasing  -DSCRYPT_KECCAK512 -DSCRYPT_CHACHA -DSCRYPT_CHOOSE_COMPILETIME -DBOOST_THREAD_USE_LIB -DBOOST_SPIRIT_THREADSAFE   -g -O2 -MT hash/skminer-Connection.o -MD -MP -MF hash/.deps/skminer-Connection.Tpo -c -o hash/skminer-Connection.o `test -f 'hash/Connection.cpp' || echo './'`hash/Connection.cpp
In file included from hash/DDOS_Filter.h:5:0,
                 from hash/Connection.cpp:2:
hash/DDOS_Score.h:24:33: error: ‘>>’ should be ‘> >’ within a nested template argument list
   std::vector<std::pair<bool,int>> SCORE;
                                 ^
hash/DDOS_Score.h:45:39: error: ‘>>’ should be ‘> >’ within a nested template argument list
   const std::vector<std::pair<bool,int>> GetScores()  const { return this->SCORE;  }

thank you

[djm34]

new sources and binaries for nvidia miner: http://ge.tt/4dUHeF32/v/0?c

main change:
* support added for compute_30 (the performance are bad... but it works... not enough register for skein 1024 leading to large spill store
* kernel working now with compute 5_2
* optimization of skein for compute >= 3_5
gtx750: 22.5MH/s
gtx780ti: 43MH/s
gtx 980: 63MH/s
* possibility to select the card to be used together with new format for IP:PORT selection

example Release/skminer 127.0.0.1:9325 gtx980#1,gtx980#2,gtxWhatever

(in the example, the program will run on two gtx980 and ignore the gtxWhatever as it won't find it on your computer (so make sure you put names of card which exist)... alternatively 980#1,980#2 will also work (as in ccminer).

If no second argument is given, it will use every card found on the system.
There is may-be a limitation, as I think, the number of card should be smaller or equal to the number of cpu core/thread (need to be verified though... and I don't have enough pci port for that...)

* new display format, I used supercomputing one.
* added compilation for linux (it is again supercomputing one...) please let me know if it works.


current hashrate on my system with this version: 193MH/s (2 gtx980, 1 gtx750 and 1 gtx780ti), cpu usage 26% (which is high compared to ccminer)

hi djm34,
after symlinking every hash/Keccak*.c with hash/Keccak*.cpp, i get this error:

Code:

g++ -DHAVE_CONFIG_H -I.  -msse2  -fopenmp -fno-strict-aliasing  -DSCRYPT_KECCAK512 -DSCRYPT_CHACHA -DSCRYPT_CHOOSE_COMPILETIME -DBOOST_THREAD_USE_LIB -DBOOST_SPIRIT_THREADSAFE   -g -O2 -MT hash/skminer-Connection.o -MD -MP -MF hash/.deps/skminer-Connection.Tpo -c -o hash/skminer-Connection.o `test -f 'hash/Connection.cpp' || echo './'`hash/Connection.cpp
In file included from hash/DDOS_Filter.h:5:0,
                 from hash/Connection.cpp:2:
hash/DDOS_Score.h:24:33: error: ‘>>’ should be ‘> >’ within a nested template argument list
   std::vector<std::pair<bool,int>> SCORE;
                                 ^
hash/DDOS_Score.h:45:39: error: ‘>>’ should be ‘> >’ within a nested template argument list
   const std::vector<std::pair<bool,int>> GetScores()  const { return this->SCORE;  }

thank you

well just do as the compiler says: replace >> by > > (kinda picky linux compiler  ) line 33 and 39 of DDOS_SCore (I must admit I forgot to check if anything had been edited at that level... I'll need to instal some linux distro on VM to compile

[skunk]

new sources and binaries for nvidia miner: http://ge.tt/4dUHeF32/v/0?c

main change:
* support added for compute_30 (the performance are bad... but it works... not enough register for skein 1024 leading to large spill store
* kernel working now with compute 5_2
* optimization of skein for compute >= 3_5
gtx750: 22.5MH/s
gtx780ti: 43MH/s
gtx 980: 63MH/s
* possibility to select the card to be used together with new format for IP:PORT selection

example Release/skminer 127.0.0.1:9325 gtx980#1,gtx980#2,gtxWhatever

(in the example, the program will run on two gtx980 and ignore the gtxWhatever as it won't find it on your computer (so make sure you put names of card which exist)... alternatively 980#1,980#2 will also work (as in ccminer).

If no second argument is given, it will use every card found on the system.
There is may-be a limitation, as I think, the number of card should be smaller or equal to the number of cpu core/thread (need to be verified though... and I don't have enough pci port for that...)

* new display format, I used supercomputing one.
* added compilation for linux (it is again supercomputing one...) please let me know if it works.


current hashrate on my system with this version: 193MH/s (2 gtx980, 1 gtx750 and 1 gtx780ti), cpu usage 26% (which is high compared to ccminer)

hi djm34,
after symlinking every hash/Keccak*.c with hash/Keccak*.cpp, i get this error:

Code:

g++ -DHAVE_CONFIG_H -I.  -msse2  -fopenmp -fno-strict-aliasing  -DSCRYPT_KECCAK512 -DSCRYPT_CHACHA -DSCRYPT_CHOOSE_COMPILETIME -DBOOST_THREAD_USE_LIB -DBOOST_SPIRIT_THREADSAFE   -g -O2 -MT hash/skminer-Connection.o -MD -MP -MF hash/.deps/skminer-Connection.Tpo -c -o hash/skminer-Connection.o `test -f 'hash/Connection.cpp' || echo './'`hash/Connection.cpp
In file included from hash/DDOS_Filter.h:5:0,
                 from hash/Connection.cpp:2:
hash/DDOS_Score.h:24:33: error: ‘>>’ should be ‘> >’ within a nested template argument list
   std::vector<std::pair<bool,int>> SCORE;
                                 ^
hash/DDOS_Score.h:45:39: error: ‘>>’ should be ‘> >’ within a nested template argument list
   const std::vector<std::pair<bool,int>> GetScores()  const { return this->SCORE;  }

thank you

well just do as the compiler says: replace >> by > > (kinda picky linux compiler  ) line 33 and 39 of DDOS_SCore (I must admit I forgot to check if anything had been edited at that level... I'll need to instal some linux distro on VM to compile

yes, it was pretty obvious, sorry...
what about this?

Code:

hash/Keccak-compact64.cpp: In function ‘void KeccakF1600_StatePermute(void*)’:
hash/Keccak-compact64.cpp:189:11: error: invalid conversion from ‘void*’ to ‘tKeccakLane* {aka long long unsigned int*}’ [-fpermissive]
     state = argState;
           ^
Makefile:585: recipe for target 'hash/skminer-Keccak-compact64.o' failed

thank you

edit: solved by adding -fpermissive to compile flags, but now i really don't know what to do:

Code:

nvcc -g -O2 -I . -Xptxas "-v" -gencode=arch=compute_30,code=\"sm_30,compute_30\" --ptxas-options=-v  -o Algo/sk1024.o -c Algo/sk1024.cu
Algo/sk1024.cu(33): error: more than one instance of overloaded function "scanhash_sk1024" has "C" linkage

Algo/sk1024.cu(78): warning: variable "pskein" was declared but never referenced

Algo/sk1024.cu(85): warning: variable "pkeccak" was declared but never referenced

1 error detected in the compilation of "/tmp/tmpxft_000013e5_00000000-6_sk1024.cpp1.ii".

(fyi, i had also to symlink util.c to util.cpp)

[djm34]

for the error in Keccak-compact64.cpp
replace "state= argState" by "state = (tKeccakLane *)argState;" (line 189)

for the error in sk1024.cu remove "extern "C"" in front of "bool scanhash_sk1024(..." (line 33)