hsrminer - Nvidia mining software for various algos by palgin&alexkap

[marecek666]

sorry devs, but this neoscrypt miner does not work with my MSI GTX1060 6GB+
this is the result:
https://imgur.com/iofNlxE
I have my nvidia driver installed ok, other neoscrypt miners work fine at 800-830 kH/s with this card.
see:
https://imgur.com/p3PILkS

What driver version do you have? hsrminer for Neoscrypt is CUDA 9.1, so minimum driver version if I'm right should be 388.3x

sorry, I have older one. Going to reinstall to the latest and will try.

Thank you for your participation!

so after the latest driver installation it is mining. But I have more questions:
why -help does not work ?
why -i does nothing ?
my rate is around 745kH/s and does not matter what -i I set
and tell me why should I use your miner which gives me 745kH/s per card - 1% fee if there is other miner which gives about 830kHs with 0 fee in the same environment? Do you have any optimalization for neoscrypt which will be in the next ver and will be better than others?
ccminer alexis 1.0 c7.5 with older drivers gives 830kH/s
ccminer klausT8.17 c9.1 with latest drivers gives 820kH/s

[1714683328]

1714683328

[1714683328]

1714683328

[1714683328]

1714683328

[fraggyb_]

Nice! 

I get around 770 kH/s with this on my Strix 1060 GB on neoscrypt, and around 750 kH/s with ccminer using the same o/c settings. 

[Ryuh]

OUT OF MEMORY ERROR SOLUTION:

increase windows memory pagefile!!

I increased from 16gb to 32gb the windows pagefile, and it worked.

I've just wanted to post the same  
My development machine has 128 Gb physical RAM installed, so I didn't observe this issue, my rigs are 5x1060 3Gb's, so they don't suffer from this issue too.

Little explanation: there's thing in Nvidia architecture called Unified Memory and this one works quite differently starting form Pascal and needs prefetching , in other words, code is accessible from both CPU and GPU, memory pages and tables are created on access. Kernel specific makes it need to have full amount of memory (even by using virtual pagefile) installed on the system to cover full amount of memory allocated on GPU's. So, if it's 6x1080? each allocates 6260+, amount of physical RAM + pagefile should be more than 6260*6 >= 37560 Mb.

I'll add this info to the first page, thank you  natan069 !!!

I can confirm this, miner is working now without problems. I put 45000 mb swap file and started all 8 cards in one instance.
Will continue observing it.

Great job palgin!

[palgin]

so after the latest driver installation it is mining. But I have more questions:
why -help does not work ?
why -i does nothing ?
my rate is around 745kH/s and does not matter what -i I set
and tell me why should I use your miner which gives me 745kH/s per card - 1% fee if there is other miner which gives about 830kHs with 0 fee in the same environment? Do you have any optimalization for neoscrypt which will be in the next ver and will be better than others?

Help is working, if you want to see help using short option, it should be -h , if you use full option name, it should be --help. And -i setting is also working, look at previous posts where people were tuning their hardware.

Maybe you do something wrong? If you post your batfile it'll be more clear.

Anyway, miner quotes clearly: "Product contains devfee 1%. If you don't agree, don't use this product" . If it doesn't suite your hardware or you feel discomfort using it - don't use it, simple

P.S.: main purpose of this for me was to learn how to handle GDDR5X properly, in which I think I succeeded, 1060's are little addition and can behave differently.

[marecek666]

so after the latest driver installation it is mining. But I have more questions:
why -help does not work ?
why -i does nothing ?
my rate is around 745kH/s and does not matter what -i I set
and tell me why should I use your miner which gives me 745kH/s per card - 1% fee if there is other miner which gives about 830kHs with 0 fee in the same environment? Do you have any optimalization for neoscrypt which will be in the next ver and will be better than others?

Help is working, if you want to see help using short option, it should be -h , if you use full option name, it should be --help. And -i setting is also working, look at previous posts where people were tuning their hardware.

Maybe you do something wrong? If you post your batfile it'll be more clear.

Anyway, miner quotes clearly: "Product contains devfee 1%. If you don't agree, don't use this product" . If it doesn't suite your hardware or you feel discomfort using it - don't use it, simple

P.S.: main purpose of this for me was to learn how to handle GDDR5X properly, in which I think I succeeded, 1060's are little addition and can behave differently.

ok thanks for the explanation
so I see, -i works fine but I am used to use this settings from other ccminers:
-i xx
and it is:
-ixx
in your miner

anyway, great job if it is able to handle gddr5x. My MSI GTX1060 6G + gives 755kH/s with -i adjusted. Higher it can not handle for my GPU.

[palgin]

ok thanks for the explanation
so I see, -i works fine but I am used to use this settings from other ccminers:
-i xx
and it is:
-ixx
in your miner

anyway, great job if it is able to handle gddr5x. My MSI GTX1060 6G + gives 755kH/s with -i adjusted. Higher it can not handle for my GPU.

Just curious, -i xx doesn't work on your machine?
That's strange, because after option specification everything you enter considereded as an argument, watching for "," and "-", skipping spaces.
For example, string -i 17 and string -i17 are treated the same, even string -i                17 will be treated the same.

Is your card flashed with any other signed BIOS? Because on my rigs even 1060 3Gb's are giving 740-770 kh without OC.

[Chetty]

Hi Dev! Which minimum version of the driver should be installed for the correct operation of the miner?

[natan069]

OUT OF MEMORY ERROR SOLUTION:

increase windows memory pagefile!!

I increased from 16gb to 32gb the windows pagefile, and it worked.

I've just wanted to post the same  
My development machine has 128 Gb physical RAM installed, so I didn't observe this issue, my rigs are 5x1060 3Gb's, so they don't suffer from this issue too.

Little explanation: there's thing in Nvidia architecture called Unified Memory and this one works quite differently starting form Pascal and needs prefetching , in other words, code is accessible from both CPU and GPU, memory pages and tables are created on access. Kernel specific makes it need to have full amount of memory (even by using virtual pagefile) installed on the system to cover full amount of memory allocated on GPU's. So, if it's 6x1080? each allocates 6260+, amount of physical RAM + pagefile should be more than 6260*6 >= 37560 Mb.

I'll add this info to the first page, thank you  natan069 !!!

thanks for the explanation 

keep the good work

[palgin]

Hi Dev! Which minimum version of the driver should be installed for the correct operation of the miner?

Hi, 388.3x should be enough, it's CUDA 9.1 . It has no performance difference from CUDA 9.0, just switched to be able to fetch all incoming features of Volta architecture.

[abudfv2008]

1060/6Gb 920 (Alexis78 910Mh)
1080: 1530Mh 189W (KlausT 1180 same OC)


Actually - now neoscrypt is for 1080 also. Almost 2 times performance of 1060 which it should be.

But the problem is with running on many GPU rig (6 in that case). Receive cuda_errors, while each GPU one by one runs smoothly.

P.S. More info in miner window would be great. For example the intensity used by default.

[pallas]

OUT OF MEMORY ERROR SOLUTION:

increase windows memory pagefile!!

I increased from 16gb to 32gb the windows pagefile, and it worked.

I've just wanted to post the same  
My development machine has 128 Gb physical RAM installed, so I didn't observe this issue, my rigs are 5x1060 3Gb's, so they don't suffer from this issue too.

Little explanation: there's thing in Nvidia architecture called Unified Memory and this one works quite differently starting form Pascal and needs prefetching , in other words, code is accessible from both CPU and GPU, memory pages and tables are created on access. Kernel specific makes it need to have full amount of memory (even by using virtual pagefile) installed on the system to cover full amount of memory allocated on GPU's. So, if it's 6x1080? each allocates 6260+, amount of physical RAM + pagefile should be more than 6260*6 >= 37560 Mb.

I'll add this info to the first page, thank you  natan069 !!!

Just wanted to say that this is not needed on linux.
I'm mining with 5x1080ti (using almost all of the gpu ram), 4gb system ram and 2gb swap.
Looking forward to the Linux version.

[palgin]

1060/6Gb 920 (Alexis78 910Mh)
1080: 1530Mh 189W (KlausT 1180 same OC)

But the problem is with running on many GPU rig (6 in that case). Receive cuda_errors, while each GPU one by one runs smoothly.

Hi, natan069 has posted a solution:

OUT OF MEMORY ERROR SOLUTION:

increase windows memory pagefile!!

I increased from 16gb to 32gb the windows pagefile, and it worked.

And I've added more detailed instructions to the first post of this thread.

Please, try it.

[palgin]

Just wanted to say that this is not needed on linux.
I'm mining with 5x1080ti (using almost all of the gpu ram), 4gb system ram and 2gb swap.
Looking forward to the Linux version.

Yes, Linux version should be OK without those manipulations.
I'm still waiting for alexkap and Linux builds, but he seems to disappear completely, I'm starting to worry about him.
He's from country other than I live in and I can't even come and visit him.

If he won't show himself, I'll switch to Linux builds myself (will have to setup Linux developement machine first).

[starl1ng]

Just wanted to say that this is not needed on linux.
I'm mining with 5x1080ti (using almost all of the gpu ram), 4gb system ram and 2gb swap.
Looking forward to the Linux version.

Yes, Linux version should be OK without those manipulations.
I'm still waiting for alexkap and Linux builds, but he seems to disappear completely, I'm starting to worry about him.
He's from country other than I live in and I can't even come and visit him.

If he won't show himself, I'll switch to Linux builds myself (will have to setup Linux developement machine first).

Great job anyway! Looking forward to test your miner under CentOS Linux.

[marecek666]

ok thanks for the explanation
so I see, -i works fine but I am used to use this settings from other ccminers:
-i xx
and it is:
-ixx
in your miner

anyway, great job if it is able to handle gddr5x. My MSI GTX1060 6G + gives 755kH/s with -i adjusted. Higher it can not handle for my GPU.

Just curious, -i xx doesn't work on your machine?
That's strange, because after option specification everything you enter considereded as an argument, watching for "," and "-", skipping spaces.
For example, string -i 17 and string -i17 are treated the same, even string -i                17 will be treated the same.

Is your card flashed with any other signed BIOS? Because on my rigs even 1060 3Gb's are giving 740-770 kh without OC.

I have this card: MSI GTX 1060 GAMING X+ 6G with original BIOS

[Kompik]

Thanks for the great work!

What about x17 algo? Verge is on a roll and is a major coin to mine

Thought that X17 is ASIC's theme now (Bainkal etc.) ? I'll take a look, of course, but won't promise anything until closer look.

ASIC is X11 through X15, x17 is actually gpu only algo and is quite profitable for verge atm . I dont know how much it is optimised though.

[palgin]

Thanks for the great work!

What about x17 algo? Verge is on a roll and is a major coin to mine

Thought that X17 is ASIC's theme now (Bainkal etc.) ? I'll take a look, of course, but won't promise anything until closer look.

ASIC is X11 through X15, x17 is actually gpu only algo and is quite profitable for verge atm . I dont know how much it is optimised though.

Oh, now I see, I'll put it on my list.

[pallas]

Just wanted to say that this is not needed on linux.
I'm mining with 5x1080ti (using almost all of the gpu ram), 4gb system ram and 2gb swap.
Looking forward to the Linux version.

Yes, Linux version should be OK without those manipulations.
I'm still waiting for alexkap and Linux builds, but he seems to disappear completely, I'm starting to worry about him.
He's from country other than I live in and I can't even come and visit him.

If he won't show himself, I'll switch to Linux builds myself (will have to setup Linux developement machine first).

I'll be glad to help if I can

[pallas]

Thanks for the great work!

What about x17 algo? Verge is on a roll and is a major coin to mine

Thought that X17 is ASIC's theme now (Bainkal etc.) ? I'll take a look, of course, but won't promise anything until closer look.

ASIC is X11 through X15, x17 is actually gpu only algo and is quite profitable for verge atm . I dont know how much it is optimised though.

Alexis version is good, but the usual some-percent can be squeezed out of it.

[igotek]

Code:

 INFO : [23:25:00] : Unable to query number of CUDA devices! Is an nVidia driver installed?

EBWF and Claymore miners work without any problem. But your miner not working.
Why ?

GTX1060 3GB
Windows 10 x64