NSGminer v0.9.4: The Fastest NeoScrypt GPU Miner - page 13.

Buguito

newbie

Activity: 4

Merit: 0

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!

totallyG

newbie

Activity: 34

Merit: 0

@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.

hapo

newbie

Activity: 2

Merit: 0

Quote from: ghostlander on November 25, 2017, 05:10:46 PM

Quote from: hapo on November 24, 2017, 06:54:16 PM

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

Quote

[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)

ghostlander

legendary

Activity: 1242

Merit: 1020

No surrender, no retreat, no regret.

Quote from: hapo on November 24, 2017, 06:54:16 PM

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

Quote

[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

newbie

Activity: 2

Merit: 0

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:

Quote

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":

Quote

[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

halker2010

sr. member

Activity: 532

Merit: 250

The harder your life is the more meaning it has.

right now what's the best oc setting for good temp in gtx 1070 and 1080 ?

ghostlander

legendary

Activity: 1242

Merit: 1020

No surrender, no retreat, no regret.

Quote from: rednoW on November 23, 2017, 05:57:13 PM

Quote from: ghostlander on November 23, 2017, 04:12:44 PM

Quote from: rednoW on November 23, 2017, 01:41:28 PM

Quote from: ghostlander on November 23, 2017, 01:06:49 PM

NSGminer v0.9.3 is out, the download links are in the OP. There are bug fixes and performance improvements for all major GPU architectures. GetBlockTemplate is now working in addition to GetWork and Stratum.

Great news! Did you test it on modern AMD gpus and recent blockchain drivers? Maybe Vega? What speed should we expect?

I was getting 515KH/s on a non-overclocked RX480 with the amdgpu-pro v17.40 drivers and +5% with v16.40. These drivers are a disaster, but there's no choice on Linux as fglrx is gone. Maybe the performance is better on Windows.

Thanks. But 515kk/s isn't good. rx480 is a competitor of gtx1066 or gtx1070 and they can do 800-1400 kh/s neoscrypt

My GTX 1070 does 750KH/s using the same kernel.

rednoW

legendary

Activity: 1510

Merit: 1003

Quote from: ghostlander on November 23, 2017, 04:12:44 PM

Quote from: rednoW on November 23, 2017, 01:41:28 PM

Quote from: ghostlander on November 23, 2017, 01:06:49 PM

NSGminer v0.9.3 is out, the download links are in the OP. There are bug fixes and performance improvements for all major GPU architectures. GetBlockTemplate is now working in addition to GetWork and Stratum.

Great news! Did you test it on modern AMD gpus and recent blockchain drivers? Maybe Vega? What speed should we expect?

I was getting 515KH/s on a non-overclocked RX480 with the amdgpu-pro v17.40 drivers and +5% with v16.40. These drivers are a disaster, but there's no choice on Linux as fglrx is gone. Maybe the performance is better on Windows.

Thanks. But 515kk/s isn't good. rx480 is a competitor of gtx1066 or gtx1070 and they can do 800-1400 kh/s neoscrypt

ghostlander

legendary

Activity: 1242

Merit: 1020

No surrender, no retreat, no regret.

Quote from: rednoW on November 23, 2017, 01:41:28 PM

Quote from: ghostlander on November 23, 2017, 01:06:49 PM

NSGminer v0.9.3 is out, the download links are in the OP. There are bug fixes and performance improvements for all major GPU architectures. GetBlockTemplate is now working in addition to GetWork and Stratum.

Great news! Did you test it on modern AMD gpus and recent blockchain drivers? Maybe Vega? What speed should we expect?

I was getting 515KH/s on a non-overclocked RX480 with the amdgpu-pro v17.40 drivers and +5% with v16.40. These drivers are a disaster, but there's no choice on Linux as fglrx is gone. Maybe the performance is better on Windows.

rednoW

legendary

Activity: 1510

Merit: 1003

Quote from: ghostlander on November 23, 2017, 01:06:49 PM

NSGminer v0.9.3 is out, the download links are in the OP. There are bug fixes and performance improvements for all major GPU architectures. GetBlockTemplate is now working in addition to GetWork and Stratum.

Great news! Did you test it on modern AMD gpus and recent blockchain drivers? Maybe Vega? What speed should we expect?

ghostlander

legendary

Activity: 1242

Merit: 1020

No surrender, no retreat, no regret.

NSGminer v0.9.3 is out, the download links are in the OP. There are bug fixes and performance improvements for all major GPU architectures. GetBlockTemplate is now working in addition to GetWork and Stratum.

ghostlander

legendary

Activity: 1242

Merit: 1020

No surrender, no retreat, no regret.

I have uploaded two additional OpenCL kernels to GitHub for older AMD GPUs of the VLIW5 and VLIW4 architectures. One with minor improvements over v7, another with a parallel execution Salsa20/20 and ChaCha20/20. If anyone is still mining with these pre-GCN GPUs, use these kernels. I think VLIW4 is better with the sequential kernel while VLIW5 prefers the parallel one.

jayjay55

newbie

Activity: 21

Merit: 0

RX480 8G
i set
nsgminer --neoscrypt -g 1 -w 128 -I 16 -o stratum+tcp://xxxxx.port -u x -p x
only 440 kh/s
is this right? please help

Elder III

sr. member

Activity: 1246

Merit: 274

The miner hasn't been optimized for AMD 400/500 series yet, so adjusting settings with those GPUs doesn't make too much of a difference. Since a R9 series AMD GPU gets the same or higher hashrate as the RX series I have hope that there is more performance to be gained with AMD. Until the developer has time to work on it, or another developer takes time to try improving hashrate and stability for AMD GPUs I don't think it's going to be any better.

lankyman

sr. member

Activity: 382

Merit: 250

Quote from: George5 on November 05, 2017, 03:53:13 PM

Quote from: MarcusDe on November 05, 2017, 10:53:28 AM

Maybe don't use intensity 15?

Even if I don't set the intensity manually, it sets itself to 16 and the hashrate is about the same.

I tried gatelessgate instead of nsgminer and it gave me a bloody blue screen and then it was a real mission to set up my GPUs again, what a pain in the ball sack

George5

legendary

Activity: 1208

Merit: 1003

Quote from: MarcusDe on November 05, 2017, 10:53:28 AM

Maybe don't use intensity 15?

Even if I don't set the intensity manually, it sets itself to 16 and the hashrate is about the same.

lankyman

sr. member

Activity: 382

Merit: 250

Quote from: AikaPool on September 08, 2017, 03:54:24 PM

RX480/580 - only 440 kh/s with -I 16 ?
too low! Sad

I hate this algorithm because NSGMiner is rubbish or either i dont know how to use it.
I dont even get 100kh/s
Its ridiculous, i dont know why Neoscrypt is horrible to mine with AMD...
I have never used Nvidia so no idea how to compare.

MarcusDe

hero member

Activity: 796

Merit: 505

Quote from: George5 on November 04, 2017, 12:37:13 PM

Hey guys! I have a question about mining NeoScrypt.

Win10 x64, Asus GTX1080Ti Trix OC (388.13 drivers), I7-7700K 4.5GHz, 32Gb DDR4 2400MHz, SSD

CAn't make my 1080Ti mine at a h/rate more than 600-650Khash/s. Already tried like 5 diff ccminers. Doesn't matter if I overclock or run it at stock clocks.

Example of ccminer bat file:

Quote

ccminerAlexis78.exe --algo=neoscrypt -o http://127.0.0.1:9554 -u User -p 123 -i 15

Any Ideas on how to make it mine at usual 1500-1600Khash/s?

Maybe don't use intensity 15?

George5

legendary

Activity: 1208

Merit: 1003

Hey guys! I have a question about mining NeoScrypt.

Win10 x64, Asus GTX1080Ti Trix OC (388.13 drivers), I7-7700K 4.5GHz, 32Gb DDR4 2400MHz, SSD

CAn't make my 1080Ti mine at a h/rate more than 600-650Khash/s. Already tried like 5 diff ccminers. Doesn't matter if I overclock or run it at stock clocks.

Example of ccminer bat file:

Quote

ccminerAlexis78.exe --algo=neoscrypt -o http://127.0.0.1:9554 -u User -p 123 -i 15

Any Ideas on how to make it mine at usual 1500-1600Khash/s?

ghostlander

legendary

Activity: 1242

Merit: 1020

No surrender, no retreat, no regret.

Quote from: btc00 on October 25, 2017, 10:34:28 AM

any new version heave ?

Pull it from GitHub and compile yourself or wait until the next release.

Topic: NSGminer v0.9.4: The Fastest NeoScrypt GPU Miner - page 13. (Read 221777 times)