NSGminer v0.9.4: The Fastest NeoScrypt GPU Miner - page 22.

sp_

legendary

Activity: 2912

Merit: 1087

Team Black developer

Quote from: ghostlander on February 13, 2016, 02:55:26 PM

GTX 750 Ti @ 1400 = 285KH/s with CUDA 6.5 or 280KH/s with CUDA 7.5

Although I've somehow broken Stratum in the process, so it's solo mining through Getwork now.

These are nice results. Have you open sourced the code or compiled a binary? want to see how you did it.

ghostlander

legendary

Activity: 1239

Merit: 1020

No surrender, no retreat, no regret.

NeoScrypt CPUminer v2.4.3 is ready.

It incorporates the latest NeoScrypt reference code improved over the last half a year. There are minor performance improvements for Linux and Windows, major ones for MacOS X which gets the SSE2 4-way mining fixed.

ghostlander

legendary

Activity: 1239

Merit: 1020

No surrender, no retreat, no regret.

Quote from: ?? on ??

mining easier than before

Many miners with Radeons switched for ETH, though FTC is also very profitable now with increased price and decreased difficulty. Reminds me the old good times.

ghostlander

legendary

Activity: 1239

Merit: 1020

No surrender, no retreat, no regret.

Quote from: barry8212 on March 02, 2016, 05:17:37 PM

i get this error

ubuntu@ip-172-31-7-223:~/nsgminer$ ./autogen.sh
Getting submodules...
Running autoreconf -ifv...
autoreconf: Entering directory `.'
autoreconf: configure.ac: not using Gettext
autoreconf: running: aclocal --force -I m4
autoreconf: configure.ac: tracing
autoreconf: configure.ac: adding subdirectory libjansson to autoreconf
autoreconf: Entering directory `libjansson'
autoreconf: running: aclocal --force
autoreconf: configure.ac: not using Libtool
autoreconf: running: /usr/bin/autoconf --force
autoreconf: running: /usr/bin/autoheader --force
autoreconf: running: automake --add-missing --copy --force-missing
Makefile.am:1: error: Libtool library used but 'LIBTOOL' is undefined
Makefile.am:1: The usual way to define 'LIBTOOL' is to add 'LT_INIT'
Makefile.am:1: to 'configure.ac' and run 'aclocal' and 'autoconf' again.
Makefile.am:1: If 'LT_INIT' is in 'configure.ac', make sure
Makefile.am:1: its definition is in aclocal's search path.
autoreconf: automake failed with exit status: 1

sudo apt-get install libtool

barry8212

newbie

Activity: 5

Merit: 0

Quote from: ghostlander on March 02, 2016, 02:54:50 PM

Quote from: Buratinos on February 29, 2016, 01:48:19 PM

Hi guys, I have a hd 7750 in Neoscrypt produces 111 kh / s, it is normal ?. It was established Catalyst 14.6

That's expected for 512 shader units. About 20% performance of R9 280X.

Quote from: barry8212 on March 01, 2016, 08:45:54 PM

Hi,

Can someone please help me to setup on ubuntu 14.04.

not sure what to type as code. please post step by step code. i appreciate the help. I managed to install on my pc with R9 280x and currently hashrate is 515kh/s

thanks

barry

git clone https://github.com/ghostlander/nsgminer.git
cd nsgminer
./autogen.sh
make
sudo make install

i get this error

ubuntu@ip-172-31-7-223:~/nsgminer$ ./autogen.sh
Getting submodules...
Running autoreconf -ifv...
autoreconf: Entering directory `.'
autoreconf: configure.ac: not using Gettext
autoreconf: running: aclocal --force -I m4
autoreconf: configure.ac: tracing
autoreconf: configure.ac: adding subdirectory libjansson to autoreconf
autoreconf: Entering directory `libjansson'
autoreconf: running: aclocal --force
autoreconf: configure.ac: not using Libtool
autoreconf: running: /usr/bin/autoconf --force
autoreconf: running: /usr/bin/autoheader --force
autoreconf: running: automake --add-missing --copy --force-missing
Makefile.am:1: error: Libtool library used but 'LIBTOOL' is undefined
Makefile.am:1: The usual way to define 'LIBTOOL' is to add 'LT_INIT'
Makefile.am:1: to 'configure.ac' and run 'aclocal' and 'autoconf' again.
Makefile.am:1: If 'LT_INIT' is in 'configure.ac', make sure
Makefile.am:1: its definition is in aclocal's search path.
autoreconf: automake failed with exit status: 1

barry8212

newbie

Activity: 5

Merit: 0

Quote from: ghostlander on March 02, 2016, 03:48:12 PM

Quote from: barry8212 on March 02, 2016, 03:37:17 PM

Hi,

Thanks for the code. Will check it out later tonight. Can you please assist me in downloading the correct driver version for MSI R9-280X Tahiti needed for the best performance on Windows 10 64bit.
What is the best feathercoin pool to mine on?

Thanks again for the assistance.

Barry

Windows 10? Catalyst 15.7 probably. Although it shouldn't be of much difference there.

The best pool is the closest to your location with low fees and as low as possible downtime.

I have crimson 16.1.1 installed. attached is a screen shot for error. can you please assist? https://www.dropbox.com/s/giv1zx6hfpwth1c/Miner.jpg?dl=0 thanks

ghostlander

legendary

Activity: 1239

Merit: 1020

No surrender, no retreat, no regret.

Quote from: barry8212 on March 02, 2016, 03:37:17 PM

Hi,

Thanks for the code. Will check it out later tonight. Can you please assist me in downloading the correct driver version for MSI R9-280X Tahiti needed for the best performance on Windows 10 64bit.
What is the best feathercoin pool to mine on?

Thanks again for the assistance.

Barry

Windows 10? Catalyst 15.7 probably. Although it shouldn't be of much difference there.

The best pool is the closest to your location with low fees and as low as possible downtime.

barry8212

newbie

Activity: 5

Merit: 0

Hi,

Thanks for the code. Will check it out later tonight. Can you please assist me in downloading the correct driver version for MSI R9-280X Tahiti needed for the best performance on Windows 10 64bit.
What is the best feathercoin pool to mine on?

Thanks again for the assistance.

Barry

ghostlander

legendary

Activity: 1239

Merit: 1020

No surrender, no retreat, no regret.

Quote from: Buratinos on February 29, 2016, 01:48:19 PM

Hi guys, I have a hd 7750 in Neoscrypt produces 111 kh / s, it is normal ?. It was established Catalyst 14.6

That's expected for 512 shader units. About 20% performance of R9 280X.

Quote from: barry8212 on March 01, 2016, 08:45:54 PM

Hi,

Can someone please help me to setup on ubuntu 14.04.

not sure what to type as code. please post step by step code. i appreciate the help. I managed to install on my pc with R9 280x and currently hashrate is 515kh/s

thanks

barry

git clone https://github.com/ghostlander/nsgminer.git
cd nsgminer
./autogen.sh
make
sudo make install

barry8212

newbie

Activity: 5

Merit: 0

Hi,

Can someone please help me to setup on ubuntu 14.04.

not sure what to type as code. please post step by step code. i appreciate the help. I managed to install on my pc with R9 280x and currently hashrate is 515kh/s

thanks

barry

Buratinos

hero member

Activity: 952

Merit: 504

Hi guys, I have a hd 7750 in Neoscrypt produces 111 kh / s, it is normal ?. It was established Catalyst 14.6

fredeq

legendary

Activity: 1536

Merit: 1005

Impressive improvements, getting 470Khash on my crappy 15.11 drivers Cheesy

ghostlander

legendary

Activity: 1239

Merit: 1020

No surrender, no retreat, no regret.

Quote from: Klokan-NXT on February 23, 2016, 03:42:41 PM

You're right, if I added "# define VLIW 1" into `neoscrypt.cl', just after :

Code:


/* NeoScrypt(128, 2, 1) with Salsa20/20 and ChaCha20/20
 * Optimised for the AMD GCN, VLIW4 and VLIW5 architectures
 * v7, 20-Feb-2016 */

#define VLIW 1

I got this error :

Code:

[20:37:42] Probing for an alive pool
[20:37:43] The network difficulty has been set to 8307
[20:37:43] Stratum from pool 0 detected new block
[20:37:43] Error -11: Building Program (clBuildProgram)
[20:37:43] input.cl:100:26: warning: unknown OpenCL extension 'cl_amd_media_ops' - ignoring
input.cl:126:14: warning: implicit declaration of function 'amd_bitalign' is invalid in C99
unsupported call to function amd_bitalign in search
[20:37:43] Failed to init GPU thread 0, disabling device 0
[20:37:43] Restarting the GPU from the menu will not fix this.
[20:37:43] Try to restart the miner.

If I changed it on the same place to "#define OLD_VLIW 1", it's core dumped again :

Code:

[20:40:07] Probing for an alive pool
[20:40:07] The network difficulty has been set to 8307
[20:40:07] Stratum from pool 0 detected new block
[20:40:07] Pool 0 is sending mismatched block contents to us (0 is not 1-1)./nsgminer.sh: line 24: 23981 Segmentation fault      (core dumped) /usr/local/bin/nsgminer --neoscrypt -o stratum+tcp://....

Would be a problem with (enough of free) memory (HD7660D can allocate 256MB at maximum, if I'm not fallen), for example ?

Yes, it doesn't seem to support bitalign. You can allocate more memory for your APU in the BIOS, 512Mb or 1Gb. It should work with -I 12 for 256Mb.

Klokan-NXT

newbie

Activity: 3

Merit: 0

Quote from: ghostlander on February 23, 2016, 01:12:26 PM

HD7660D is a VLIW4 design which should be capable of some 60KH/s @ 800MHz. I'm almost sure the OSS driver doesn't expose any GPU specific data to an OpenCL compiler, so you have to set it up manually. Add "#define VLIW 1" somewhere in the kernel beginning. If there is no support for cl_amd_media_ops, change it to "#define OLD_VLIW 1"

You're right, if I added "# define VLIW 1" into `neoscrypt.cl', just after :

Code:


/* NeoScrypt(128, 2, 1) with Salsa20/20 and ChaCha20/20
 * Optimised for the AMD GCN, VLIW4 and VLIW5 architectures
 * v7, 20-Feb-2016 */

#define VLIW 1

I got this error :

Code:

[20:37:42] Probing for an alive pool
[20:37:43] The network difficulty has been set to 8307
[20:37:43] Stratum from pool 0 detected new block
[20:37:43] Error -11: Building Program (clBuildProgram)
[20:37:43] input.cl:100:26: warning: unknown OpenCL extension 'cl_amd_media_ops' - ignoring
input.cl:126:14: warning: implicit declaration of function 'amd_bitalign' is invalid in C99
unsupported call to function amd_bitalign in search
[20:37:43] Failed to init GPU thread 0, disabling device 0
[20:37:43] Restarting the GPU from the menu will not fix this.
[20:37:43] Try to restart the miner.

If I changed it on the same place to "#define OLD_VLIW 1", it's core dumped again :

Code:

[20:40:07] Probing for an alive pool
[20:40:07] The network difficulty has been set to 8307
[20:40:07] Stratum from pool 0 detected new block
[20:40:07] Pool 0 is sending mismatched block contents to us (0 is not 1-1)./nsgminer.sh: line 24: 23981 Segmentation fault      (core dumped) /usr/local/bin/nsgminer --neoscrypt -o stratum+tcp://....

Would be a problem with (enough of free) memory (HD7660D can allocate 256MB at maximum, if I'm not fallen), for example ?

ghostlander

legendary

Activity: 1239

Merit: 1020

No surrender, no retreat, no regret.

Quote from: KloNEM on February 23, 2016, 12:03:09 PM

Quote from: ghostlander on February 15, 2016, 12:27:30 PM

cl_amd_media_ops is for bitalign/bytealign mostly which are not used in v6 directly. The compiler is supposed to take care of this, but it doesn't do well in the drivers newer than 14.7. It won't be an issue in the next release.

Hi ghostlander,

thanks for amazing job! You're right, that isn't an issue in new 0.9.2 version anymore.

And also, I removed "static" from `neoscrypt.cl' source file (this string is three times there), and now the error "OpenCL does not support the 'static' storage class specifier" is gone. Great! But with --neoscrypt option, nsgminer exits very quickly - sometimes it lists current config and then "Segmentation fault (core dumped)", sometimes it gets this fault immediately. Unfortunately there is no dump file (even if I added --verbose option) for further analysis.

The same nsgminer, with option --scrypt, or without algo-option (it means sha256 algo) runs fine. Well, the hash-power is really ridiculous, but it's not just related to OSS drivers, it's also because I'm testing it on very weak IGP/APU (Radeon HD 7660D). If everything will be fine with --neoscrypt there, I'll try on more powerful/faster discrete AMD cards too...

Finally, within compilation of nsgminer from git repo, I didn't add AMD APP SDK; I guess it doesn't make a sense with OSS radeon.ko driver ... or would be there some performance impact also in this "specific OSS case" ?

Thanks for your help and your work on nsgminer again!

HD7660D is a VLIW4 design which should be capable of some 60KH/s @ 800MHz. I'm almost sure the OSS driver doesn't expose any GPU specific data to an OpenCL compiler, so you have to set it up manually. Add "#define VLIW 1" somewhere in the kernel beginning. If there is no support for cl_amd_media_ops, change it to "#define OLD_VLIW 1"

KloNEM

member

Activity: 181

Merit: 11

Quote from: ghostlander on February 15, 2016, 12:27:30 PM

cl_amd_media_ops is for bitalign/bytealign mostly which are not used in v6 directly. The compiler is supposed to take care of this, but it doesn't do well in the drivers newer than 14.7. It won't be an issue in the next release.

Hi ghostlander,

thanks for amazing job! You're right, that isn't an issue in new 0.9.2 version anymore.

And also, I removed "static" from `neoscrypt.cl' source file (this string is three times there), and now the error "OpenCL does not support the 'static' storage class specifier" is gone. Great! But with --neoscrypt option, nsgminer exits very quickly - sometimes it lists current config and then "Segmentation fault (core dumped)", sometimes it gets this fault immediately. Unfortunately there is no dump file (even if I added --verbose option) for further analysis.

The same nsgminer, with option --scrypt, or without algo-option (it means sha256 algo) runs fine. Well, the hash-power is really ridiculous, but it's not just related to OSS drivers, it's also because I'm testing it on very weak IGP/APU (Radeon HD 7660D). If everything will be fine with --neoscrypt there, I'll try on more powerful/faster discrete AMD cards too...

Finally, within compilation of nsgminer from git repo, I didn't add AMD APP SDK; I guess it doesn't make a sense with OSS radeon.ko driver ... or would be there some performance impact also in this "specific OSS case" ?

Thanks for your help and your work on nsgminer again!

ghostlander

legendary

Activity: 1239

Merit: 1020

No surrender, no retreat, no regret.

Quote from: sp_ on February 23, 2016, 09:10:19 AM

Quote from: ghostlander on February 22, 2016, 08:57:33 AM

10MH/s is much better, but still not enough to compete with these ASICs. There may be even faster ASICs by the time Pascal comes into production.

The (28nm) R9 Nano is already 1/2 the speed of the announced unreleased ASIC if you messure hash/Watt. With a modded kernal you can reach 20MHASH per card. for $2000 you can buy many cards..

Pascal is rumoured to be 12 times faster, and will be released in 2016. (16nm)

The announced asic is slow. It should be 2.3 GHASH / 700Watt. to beat the competition.

I'm sure this ASIC is of a very inexpensive design and production. Something like 150nm and multiproject wafers. Indeed it's overpriced. Any subsequent designs should be much more impressive.

sp_

legendary

Activity: 2912

Merit: 1087

Team Black developer

Quote from: ghostlander on February 22, 2016, 08:57:33 AM

10MH/s is much better, but still not enough to compete with these ASICs. There may be even faster ASICs by the time Pascal comes into production.

The (28nm) R9 Nano is already 1/2 the speed of the announced unreleased ASIC if you messure hash/Watt. With a modded kernal you can reach 20MHASH per card. for $2000 you can buy many cards..

Pascal is rumoured to be 12 times faster, and will be released in 2016. (16nm)

The announced asic is slow. It should be 2.3 GHASH / 700Watt. to beat the competition.

ghostlander

legendary

Activity: 1239

Merit: 1020

No surrender, no retreat, no regret.

Quote from: sp_ on February 22, 2016, 06:58:31 AM

Quote from: ghostlander on February 22, 2016, 06:32:40 AM

iBeLink DM384M Dash Miner

If what they advertise is true, 384MH/s for 715W and $2098 in cash, say good-bye to X11 GPU mining. A reference R9 280X outputs 2MH/s for maybe 150W. If they produce enough of these ASICs for themselves, they can do a 51% attack on any X11 coin including Dash.

The latest bins does 10MHASH on the reference 280x (nicehashminer)

DM384M Dash Miner: 0,537 Mhash /Watt
r9 280x: 0.066 Mhash /Watt (1/8 of the asic) (tahiti 2012)
750ti Maxwell: 0.075 Mhash /watt (1/7 of the asic) (maxwell 2014)
r9 NANO 0,141 MHASH /watt (1/4 of the asic)(fury 2015)

NVIDIA pascal(16nm) ... (1/2? of the asic, or perhaps the same speed.) (2016?)

10MH/s is much better, but still not enough to compete with these ASICs. There may be even faster ASICs by the time Pascal comes into production.

sp_

legendary

Activity: 2912

Merit: 1087

Team Black developer

Quote from: ghostlander on February 22, 2016, 06:32:40 AM

iBeLink DM384M Dash Miner

If what they advertise is true, 384MH/s for 715W and $2098 in cash, say good-bye to X11 GPU mining. A reference R9 280X outputs 2MH/s for maybe 150W. If they produce enough of these ASICs for themselves, they can do a 51% attack on any X11 coin including Dash.

The latest bins does 10MHASH on the reference 280x (nicehashminer)

DM384M Dash Miner: 0,537 Mhash /Watt
r9 280x: 0.066 Mhash /Watt (1/8 of the asic) (tahiti 2012)
750ti Maxwell: 0.075 Mhash /watt (1/7 of the asic) (maxwell 2014)
r9 NANO 0,141 MHASH /watt (1/4 of the asic)(fury 2015)

NVIDIA pascal(16nm) ... (1/2? of the asic, or perhaps the same speed.) (2016?)

Topic: NSGminer v0.9.4: The Fastest NeoScrypt GPU Miner - page 22. (Read 221582 times)