NSGminer v0.9.4: The Fastest NeoScrypt GPU Miner

[hapo]

Hey,

You've developed a great product!
And I hope you could help me with a small problem of running it on my ubuntu 16.04 with GeForce GTX 550 Ti on board. Thanks in advance!
I've already tryed to reinstall drivers but it didn't help. It doesn't mine at all so I have "0.0 KH/s".

I'm using NeoScrypt GPU Miner v0.9.3 that I compiled from source.

The verbose output:

user@user-P5K-SE:~/nsgminer$ ./nsgminer --neoscrypt -I 10  -o stratum+tcp://s1.theblocksfactory.com:3332 -O hapo.1:x -D -T --verbose
[2017-11-24 23:33:37] Started nsgminer 0.9.3                   
[23:33:37] OpenCL platform 0 vendor: NVIDIA Corporation                   
[23:33:37] OpenCL platform 0 name: NVIDIA CUDA                   
[23:33:37] OpenCL platform 0 version: OpenCL 1.2 CUDA 9.0.229                   
[23:33:37] Platform 0 devices: 1                   
[23:33:37]    0   GeForce GTX 550 Ti
                   
[23:33:37] OCL 0: Set temperature config: target=89 cutoff=95                   
[23:33:37] Probing for an alive pool                   
[23:33:37] Testing pool stratum+tcp://s1.theblocksfactory.com:3332                   
[23:33:37] Popping work to stage thread                   
[23:33:37] HTTP request failed: Protocol "stratum+tcp" not supported or disabled in libcurl                   
[23:33:37] Failed to connect in json_rpc_call                   
[23:33:37] initiate_stratum with sockbuf=(nil)                   
[23:33:37] Pool 0 difficulty set to 8.000000                   
[23:33:38] Pool 0 difficulty set to 8.000000                   
[23:33:38] Stratum authorisation success for pool 0                   
[23:33:38] Pool 0 stratum+tcp://s1.theblocksfactory.com:3332 active                   
[23:33:38] Pushing ping to thread 0                   
[23:33:38] Init GPU thread 0 GPU 0 virtual GPU 0                   
[23:33:38] OpenCL platform vendor: NVIDIA Corporation                   
[23:33:38] OpenCL platform name: NVIDIA CUDA                   
[23:33:38] OpenCL platform version: OpenCL 1.2 CUDA 9.0.229                   
[23:33:38] List of devices:                   
[23:33:38]    0   GeForce GTX 550 Ti                   
[23:33:38] Selected 0: GeForce GTX 550 Ti                   
[23:33:38] Activated NVIDIA management for GPU 0                   
[23:33:38] Preferred vector width reported 1                   
[23:33:38] Max work group size reported 1024                   
[23:33:38] Max mem alloc size is 252411904                   
[23:33:38] Selecting the default NeoScrypt kernel                   
[23:33:38] GPU 0: max. intensity is 10                   
[23:33:38] No binary found, generating from source                   
[23:33:38] Setting work size to 256                   
[23:33:38] cl_amd_media_ops not found, will not set BITALIGN                   
[23:33:38] CompilerOptions: -D WORKSIZE=256 -D GOFFSET                   
[23:33:38] Error -9999: Building Program (clBuildProgram)                   
[23:33:38] ptxas application ptx input, line 199; error   : Instruction 'shf.r' requires .target sm_32 or higher
ptxas application ptx input, line 205; error   : Instruction 'shf.r' requires .target sm_32 or higher
ptxas application ptx input, line 211; error   : Instruction 'shf.r' requires .target sm_32 or higher
ptxas application ptx input, line 217; error   : Instruction 'shf.r' requires .target sm_32 or higher
ptxas application ptx input, line 223; error   : Instruction 'shf.r' requires .target sm_32 or higher
ptxas application ptx input, line 229; error   : Instruction 'shf.r' requires .target sm_32 or higher
ptxas application ptx input, line 235; error   : Instruction 'shf.r' requires .target sm_32 or higher
ptxas application ptx input, line 241; error   : Instruction 'shf.r' requires .target sm_32 or higher
ptxas application ptx input, line 247; error   : Instruction 'shf.r' requires .target sm_32 or higher
ptxas application ptx input, line 313; error   : Instruction 'shf.r' requires .target sm_32 or higher
ptxas application ptx input, line 318; error   : Instruction 'shf.r' requires .target sm_32 or higher
ptxas application ptx input, line 323; error   : Instruction 'shf.r' requires .target sm_32 or higher
ptxas application ptx input, line 328; error   : Instruction 'shf.r' requires .target sm_32 or higher
ptxas application ptx input, line 333; error   : Instruction 'shf.r' requires .target sm_32 or higher
ptxas application ptx input, line 338; error   : Instruction 'shf.r' requires .target sm_32 or higher
ptxas application ptx input, line 343; error   : Instruction 'shf.r' requires .target sm_32 or higher
ptxas application ptx input, line 348; error   : Instruction 'shf.r' requires .target sm_32 or higher
ptxas application ptx input, line 353; error   : Instruction 'shf.r' requires .target sm_32 or higher
ptxas application ptx input, line 358; error   : Instruction 'shf.r' requires .target sm_32 or higher
ptxas application ptx input, line 363; error   : Instruction 'shf.r' requires .target sm_32 or higher
ptxas application ptx input, line 368; error   : Instruction 'shf.r' requires .target sm_32 or higher
ptxas application ptx input, line 373; error   : Instruction 'shf.r' requires .target sm_32 or higher
ptxas application ptx input, line 378; error   : Instruction 'shf.r' requires .target sm_32 or higher
ptxas application ptx input, line 383; error   : Instruction 'shf.r' requires .target sm_32 or higher
ptxas application ptx input, line 388; error   : Instruction 'shf.r' requires .target sm_32 or higher
ptxas application ptx input, line 3813; error   : Instruction 'shf.r' requires .target sm_32 or higher
ptxas application ptx input, line 3816; error   : Instruction 'shf.r' requires .target sm_32 or higher
ptxas application ptx input, line 3838; error   : Instruction 'shf.r' requires .target sm_32 or higher
ptxas application ptx input, line 3849; error   : Instruction 'shf.r' requires .target sm_32 or higher
ptxas application ptx input, line 3860; error   : Instruction 'shf.r' requires .target sm_32 or higher
ptxas application ptx input, line 3871; error   : Instruction 'shf.r' requires .target sm_32 or higher
ptxas application ptx input, line 3882; error   : Instruction 'shf.r' requires .target sm_32 or higher
ptxas application ptx input, line 3893; error   : Instruction 'shf.r' requires .target sm_32 or higher
ptxas application ptx input, line 3904; error   : Instruction 'shf.r' requires .target sm_32 or higher
ptxas application ptx input, line 3916; error   : Instruction 'shf.r' requires .target sm_32 or higher
ptxas application ptx input, line 3933; error   : Instruction 'shf.r' requires .target sm_32 or higher
ptxas application ptx input, line 3950; error   : Instruction 'shf.r' requires .target sm_32 or higher
ptxas application ptx input, line 3964; error   : Instruction 'shf.r' requires .target sm_32 or higher
ptxas application ptx input, line 3975; error   : Instruction 'shf.r' requires .target sm_32 or higher
ptxas application ptx input, line 3986; error   : Instruction 'shf.r' requires .target sm_32 or higher
                   
[23:33:38] Failed to init GPU thread 0, disabling device 0                   
[23:33:38] Restarting the GPU from the menu will not fix this.                   
[23:33:38] Try to restart the miner.                   
[23:33:38] 1 GPU mining threads started                   
[23:33:38] API not running - API will not be available                   
[23:33:38] Creating extra submit work thread

In normal mode it writes me after lines with "error   : Instruction 'shf.r' requires .target sm_32 or higher":

[23:38:13] Failed to init GPU thread 0, disabling device 0                   
[23:38:13] Restarting the GPU from the menu will not fix this.                   
[23:38:13] Try to restart the miner.                   
[23:38:14] The network difficulty has been set to 5095

[ghostlander]

In normal mode it writes me after lines with "error   : Instruction 'shf.r' requires .target sm_32 or higher":

[23:38:13] Failed to init GPU thread 0, disabling device 0                    
[23:38:13] Restarting the GPU from the menu will not fix this.                    
[23:38:13] Try to restart the miner.                    
[23:38:14] The network difficulty has been set to 5095

shf.r is unsupported by GK107 and older NVIDIA GPUs. There is no way to work around this automatically within the kernel. You need to open it in a text editor and comment out the shf.r PTX asm function.

Code:

#if (cl_amd_media_ops)
#define AMD 1
#elif (cl_nv_pragma_unroll)
#define NVIDIA 1
/* Available for Fermi and newer, use the macro below for Tesla */
//uint amd_bitalign(uint src0, uint src1, uint src2) {
//    uint dst;
//    asm("shf.r.wrap.b32 %0, %2, %1, %3;" : "=r"(dst) : "r"(src0), "r"(src1), "r"(src2));
//    return(dst);
//}
//#else
#define amd_bitalign(src0, src1, src2) ((uint)(((((ulong)src0) << 32) | (ulong)src1) >> (src2 & 0x1F)))
#endif

[hapo]

In normal mode it writes me after lines with "error   : Instruction 'shf.r' requires .target sm_32 or higher":

[23:38:13] Failed to init GPU thread 0, disabling device 0                    
[23:38:13] Restarting the GPU from the menu will not fix this.                    
[23:38:13] Try to restart the miner.                    
[23:38:14] The network difficulty has been set to 5095

shf.r is unsupported by GK107 and older NVIDIA GPUs. There is no way to work around this automatically within the kernel. You need to open it in a text editor and comment out the shf.r PTX asm function.

Code:

#if (cl_amd_media_ops)
#define AMD 1
#elif (cl_nv_pragma_unroll)
#define NVIDIA 1
/* Available for Fermi and newer, use the macro below for Tesla */
//uint amd_bitalign(uint src0, uint src1, uint src2) {
//    uint dst;
//    asm("shf.r.wrap.b32 %0, %2, %1, %3;" : "=r"(dst) : "r"(src0), "r"(src1), "r"(src2));
//    return(dst);
//}
//#else
#define amd_bitalign(src0, src1, src2) ((uint)(((((ulong)src0) << 32) | (ulong)src1) >> (src2 & 0x1F)))
#endif

Thank you so much! Works fine now.
p.s. just added

Code:

-w 128

to prevent this error:

Code:

[15:19:48] Error -9999: Building Program (clBuildProgram)                    
[15:19:48] ptxas error   : Entry function 'search' uses too much shared data (0x10040 bytes, 0xc000 max)

[totallyG]

@ghostlander

Any chance we will see this miner for Intel GPU that support Open CL 1.2?
Old AMD 1.2 cards is allready supported, and code would'nt be that different?
It would be the one algo where the Intel GPUs would shine?
Have I understood correctly if I state that NeoScrypt is not that dependent on fast memory? (compared to Cryptonight, Ethhash and Equihash)

Intel HD4400 have 128 cores and 16 EU (execution units similar to AMD CU but just with 8 cores each) for instance.

[Buguito]

I'm trying to use the NSGMiner with Nicehash for NeoScrypt mining but i'm unable to connect to any of their pools. Here's the debug output:

[2017-11-27 17:13:00] Started nsgminer 0.9.3
[17:13:00] Loaded configuration file neo.conf
[17:13:00] DISPLAY not set, setting :0 just in case
[17:13:00] OpenCL platform 0 vendor: NVIDIA Corporation
[17:13:00] OpenCL platform 0 name: NVIDIA CUDA
[17:13:00] OpenCL platform 0 version: OpenCL 1.2 CUDA 9.0.159
[17:13:00] Platform 0 devices: 1
[17:13:00]      0       GeForce GTX 1070

[17:13:00] OCL 0: Set temperature config: target=89 cutoff=95
[17:13:00] Probing for an alive pool
[17:13:00] Testing pool stratum+tcp://neoscrypt.br.nicehash.com:3341
[17:13:00] Popping work to stage thread
[17:13:00] JSON protocol request:
{"id": 0, "method": "getblocktemplate", "params": [{"capabilities": ["coinbasetxn", "workid", "longpoll", "coinbase/append", "time/increment", "version/force", "versio$
[17:13:00] Pool 0: Rebuilt URL to: stratum+tcp://neoscrypt.br.nicehash.com:3341/
[17:13:00] Pool 0: Protocol "stratum+tcp" not supported or disabled in libcurl
[17:13:00] Pool 0: Closing connection -1
[17:13:00] HTTP request failed: Protocol "stratum+tcp" not supported or disabled in libcurl
[17:13:00] Failed to connect in json_rpc_call
[17:13:00] initiate_stratum with sockbuf=(nil)
[17:13:00] Pool 0: Rebuilt URL to: http://neoscrypt.br.nicehash.com:3341/
[17:13:01] Pool 0:   Trying 169.57.150.34...
[17:13:01] Pool 0: TCP_NODELAY set
[17:13:02] Pool 0: Connected to neoscrypt.br.nicehash.com (169.57.150.34) port 3341 (#0)
[17:13:02] Pool 0: Curl_http_done: called premature == 0
[17:13:02] Pool 0: Connection #0 to host neoscrypt.br.nicehash.com left intact
[17:13:02] Pool 0: SEND: {"id": 0, "method": "mining.subscribe", "params": []}
[17:13:03] Pool 0: RECV: {"id":0,"error":null,"result":[["mining.notify","f46edce63f791f9f1d1edc6226d04fa7"],"1800073600",3]}
[17:13:03] Pool 0 confirmed mining.subscribe with extranonce1 1800073600 extran2size 3
[17:13:03] Pool 0: SEND: {"id": "auth", "method": "mining.authorize", "params": ["39ztvGnzwtBaDr3xkLz34D1BZTTSQHiwQH.GTX1070", "x"]}
[17:13:05] Socket closed waiting in recv_line
[17:13:05] Unable to get work from pool 0 stratum+tcp://neoscrypt.br.nicehash.com:3341

Any help would be really appreciated!

[Hangterisaan]

800 kh for a 1070? Not really the fastest then, my 1070's do 1300/1400 on ccminer.

[jnator]

800 kh for a 1070? Not really the fastest then, my 1070's do 1300/1400 on ccminer.

What version of ccminer are you using and what clock configuration?
I'm not capable to obtain more than 740kh 1050/1100kh in my 1070's

[ghostlander]

@ghostlander

Any chance we will see this miner for Intel GPU that support Open CL 1.2?
Old AMD 1.2 cards is allready supported, and code would'nt be that different?
It would be the one algo where the Intel GPUs would shine?
Have I understood correctly if I state that NeoScrypt is not that dependent on fast memory? (compared to Cryptonight, Ethhash and Equihash)

Intel HD4400 have 128 cores and 16 EU (execution units similar to AMD CU but just with 8 cores each) for instance.

There is a chance that it works with Intel already. Minor changes may be required for performance tuning.

[ghostlander]

800 kh for a 1070? Not really the fastest then, my 1070's do 1300/1400 on ccminer.

If you expect OpenCL to match CUDA in performance, that's unlikely to happen any time soon. Although there are more optimisations to come.

[totallyG]

@ghostlander

Any chance we will see this miner for Intel GPU that support Open CL 1.2?
Old AMD 1.2 cards is allready supported, and code would'nt be that different?
It would be the one algo where the Intel GPUs would shine?
Have I understood correctly if I state that NeoScrypt is not that dependent on fast memory? (compared to Cryptonight, Ethhash and Equihash)

Intel HD4400 have 128 cores and 16 EU (execution units similar to AMD CU but just with 8 cores each) for instance.

There is a chance that it works with Intel already. Minor changes may be required for performance tuning.

Tried to start it just now. It just list the devices and write output that AMD ADL and nVidia something is not available. Any dependancy to AMD or nVidia should be removed from the general neoscrypt kernel.

[ghostlander]

@ghostlander

Any chance we will see this miner for Intel GPU that support Open CL 1.2?
Old AMD 1.2 cards is allready supported, and code would'nt be that different?
It would be the one algo where the Intel GPUs would shine?
Have I understood correctly if I state that NeoScrypt is not that dependent on fast memory? (compared to Cryptonight, Ethhash and Equihash)

Intel HD4400 have 128 cores and 16 EU (execution units similar to AMD CU but just with 8 cores each) for instance.

There is a chance that it works with Intel already. Minor changes may be required for performance tuning.

Tried to start it just now. It just list the devices and write output that AMD ADL and nVidia something is not available. Any dependancy to AMD or nVidia should be removed from the general neoscrypt kernel.

Of course ADL and NVML are not available on Intel, but that's not an error. The generic kernel is supposed to be compatible with most OpenCL devices.

[sousartgames]

Nice!!!

[sousartgames]

how much I can miner whith a Rx 360?

[totallyG]

@ghostlander

Any chance we will see this miner for Intel GPU that support Open CL 1.2?
Old AMD 1.2 cards is allready supported, and code would'nt be that different?
It would be the one algo where the Intel GPUs would shine?
Have I understood correctly if I state that NeoScrypt is not that dependent on fast memory? (compared to Cryptonight, Ethhash and Equihash)

Intel HD4400 have 128 cores and 16 EU (execution units similar to AMD CU but just with 8 cores each) for instance.

There is a chance that it works with Intel already. Minor changes may be required for performance tuning.

Tried to start it just now. It just list the devices and write output that AMD ADL and nVidia something is not available. Any dependancy to AMD or nVidia should be removed from the general neoscrypt kernel.

Of course ADL and NVML are not available on Intel, but that's not an error. The generic kernel is supposed to be compatible with most OpenCL devices.

Maybee it is not in the kernel. I can have a look in the code and see if I find something.

[dj--alex]

NSGMiner not working with GPU on Linux.  

I tries 4 more times   - never working, even on CPU
I install Linux Kernel 4.13.2 final release but this not helped
I have 2x GTX1060

Please anybody upload normal COMPILED miner for test

#!/bin/bash/
#./minerd -a neoscrypt -o stratum+tcp://pool.minertopia.org:4233 -u TdBSh6BY2iU9uadNtDPXiXgMPMSaD4X5ac -p c=TZC
#./minerd -a neoscrypt -o stratum+tcp://eu1.altminer.net:4233 -u TdBSh6BY2iU9uadNtDPXiXgMPMSaD4X5ac -p c=TZC
#./minerd -k neoscrypt -g 1 -w 64 -I 15 -o  stratum+tcp://eu1.altminer.net:4233 -O TdBSh6BY2iU9uadNtDPXiXgMPMSaD4X5ac:x
./minerd -a neoscrypt -o  stratum+tcp://eu1.altminer.net:4233 -O TdBSh6BY2iU9uadNtDPXiXgMPMSaD4X5ac:x

ry `minerd --help' for more information.
user@PC1 /media/user/F/Work/Mining/nsgminer-trezarcoin-cryptopia-neoscrypt $ sh _2_start_trezar_cryptopia_neoscrypt.sh
NeoScrypt CPUminer v2.4.3
./minerd: invalid option -- 'g'
Try `minerd --help' for more information.
user@PC1 /media/user/F/Work/Mining/nsgminer-trezarcoin-cryptopia-neoscrypt $
user@PC1 /media/user/F/Work/Mining/nsgminer-trezarcoin-cryptopia-neoscrypt $ sh _2_start_trezar_cryptopia_neoscrypt.sh
NeoScrypt CPUminer v2.4.3
./minerd: invalid option -- 'w'
Try `minerd --help' for more information.
user@PC1 /media/user/F/Work/Mining/nsgminer-trezarcoin-cryptopia-neoscrypt $ sh _2_start_trezar_cryptopia_neoscrypt.sh
NeoScrypt CPUminer v2.4.3
./minerd: invalid option -- 'I'
Try `minerd --help' for more information.
user@PC1 /media/user/F/Work/Mining/nsgminer-trezarcoin-cryptopia-neoscrypt $
user@PC1 /media/user/F/Work/Mining/nsgminer-trezarcoin-cryptopia-neoscrypt $ sh _2_start_trezar_cryptopia_neoscrypt.sh
NeoScrypt CPUminer v2.4.3
Engines: INT SSE2 SSE2-4way (enabled: SSE2-4way)

[capt.khannaa@gmail.com]

I have 12 RX580 free now..
What would be the proffit.?

[silentsigma]

what miner and setting are you guys using for the 1070 to get neoscrypt pass 1000? Mine only does 800 . OC doesn't seem to do much

[Buguito]

I'm getting this on a GTX 1070 under Debian.

[07:30:42] Probing for an alive pool
[07:30:42] The network difficulty has been set to 8376950971
[07:30:42] Stratum from pool 0 detected new block
[07:30:43] Error -9999: Building Program (clBuildProgram)
[07:30:43] ptxas error   : Entry function 'search' uses too much shared data (0x10040 bytes, 0xc000 max)

[07:30:43] Failed to init GPU thread 0, disabling device 0
[07:30:43] Restarting the GPU from the menu will not fix this.
[07:30:43] Try to restart the miner.

[totallyG]

@ghostlander

[2017-11-30 13:34:34] OpenCL platform 0 vendor: Intel(R) Corporation
[13:34:34] OpenCL platform 0 name: Intel(R) OpenCL
[13:34:34] OpenCL platform 0 version: OpenCL 1.2
[13:34:34] Platform 0 devices: 1
[13:34:34]      0       Intel(R) HD Graphics 4000

[13:34:34] 1 OpenCL GPU device detected

[13:34:34] Unable to load the AMD Display Library
[13:34:34] Unable to load the NVIDIA Management Library

The device gets detected but never selected. It is not listed as expected suggesting:

ocl.c L: 465    if (numDevices > 0 ) { need to be if (numDevices >= 0 ) {

ocl.c L: 489 eventually if (gpu < numDevices) { needs to be set to if (gpu <= numDevices) {

Can you edit this and make a new build for windows? I have no machine set for building atm.

Another thing: Can intensity 4 be used for NeoScrypt? The intel gpus have 256kB L3 and can also use the CPUs L3 or eDRAM (if Iris), then it will need to use "slow" system RAM.

[Temporel]

Thanks - is NeoScrypt just something that will always work better on AMD due to inherent differences in the GPUs, or is it simply because no one has built an optimised miner for NVIDIA cards yet?

Neoscrypt work fine with 1060/1070. You can get:
1060 - 900+
1070 - 1100+
Actually 1060 make the best H/$ for neoscrypt among Nvidia GPU's.

what miner ?