SRBMiner Cryptonight AMD GPU Miner V1.9.3 - native algo switching

[huntingthesnark]

What am I not doing please?

You didn't mention what you're mining. I think "Above target" means the wrong algo, there are 3 cryptonight algos on nicehash (regular, v7 and heavy), and you'd need to select the right one for the port you're aiming your miners at.

Yeah, you're right, it means NH isn't recognising the correct algo. Question is, how do you get the miner to tell NH? I've tried the obvious "normalv7" in config, and "nicehash" true in .bat, but no joy - immediate rejects... Mines XMR@nanopool quite happily on this setting...

[hamesh]

Hi,

Please give a config.txt file for AMD RX280X!
Thanks

I have four R9 280x cards that do between 475 to 550h/s depending on the ram type.  Here's my config:

Code:

"gpu_conf" :
[
{ "id" : 0, "intensity" : 27, "worksize" : 8, "threads" : 2, "adl_type" : 2},

You can try intensity 29 or 31.  Double the intensity if you drop to worksize 4.  But I find worksize 8 gives best hashrate.

[Ballscack]

Link must be broken, new one https://mega.nz/#!OaYRVJTI!Z5dsgQTfMxlVrn9fNdqdNvG36eIAx4smfpagUxMiM2U. Watchdog has done a failed reset. It didn't mine after that last part so...

this is v 1.5.8, please use latest as maybe the bug you are experiencing is fixed.

Doktor why with versión 1.4.2 nofee i get 67-72 kh/s stellite before new algo, and after with new algo 1.6.0 only 58-70 (miner show 70.5) same pools? A lot of less, not only 0.85%.

Tried 1.4.2 with Graft and works fine.
Tried 1.6.0 with Graft and down a lot the hashrate with same settings.

The miner show a bit more hashrate on 1.6.0, but in the pools show a lot less.

1.4.2 imagreedybastard -24 hours average hashrate 67-69kh/s
1.6.0                                  -24 hours average hashrate 60-62kh/s

care to share ur 1.4.2 srbminer for testing? wanted to message u buth it is disablet it seems on ur side to recive mails.

[livada]

What am I not doing please?

You didn't mention what you're mining. I think "Above target" means the wrong algo, there are 3 cryptonight algos on nicehash (regular, v7 and heavy), and you'd need to select the right one for the port you're aiming your miners at.

Yeah, you're right, it means NH isn't recognising the correct algo. Question is, how do you get the miner to tell NH? I've tried the obvious "normalv7" in config, and "nicehash" true in .bat, but no joy - immediate rejects... Mines XMR@nanopool quite happily on this setting...

NH vork fine with 6*vega56/64 + srbminer 1.6.0 and this option:

V7:
{"pool_use_tls" : false, "pool" : "cryptonightv7.eu.nicehash.com:3363", "wallet" : "adres", "password" : "x", "nicehash" : true, "max_difficulty" : 1700000},

heavy:
{"pool_use_tls" : false, "pool" : "cryptonightv7.eu.nicehash.com:3364", "wallet" : "adres", "password" : "x", "nicehash" : true, "max_difficulty" : 1700000},

and use  config file identical like nano or other algo

U use wrong port ...

"intensity" : 112 or 120 , "worksize" : 8 or 16, for v7
"intensity" : 56, "worksize" : 8 for heavy

[Kriss.Kross]

Iam mining Loki (CN-Heavy) on 6x HIS RX460 2GB, with a single thread intensity max value is 28 and for double thread intensity max value is 14. But the single thread hashrate more higher than a dual thread hashrate ... it is normal ?

Double threads only seems to be worthwhile (on cryptonight heavy) if you have 8 GB GPUs. At least that is how it has been in our testing for 470/570 and 480/580 GPUs (4Gb vs 8GB models).

So it's better to keep a single thread for my 2Gb cards ? Or maybe there is an option to boost my cards hashrate?

[frozen80]

Hello @doktor83,

I'm trying to use my CPU(Threadripper 1950x) also with SRBMiner v1.6.0 but i got an OpenCL error. I used the default config from the readme file and the --usecpuopencl arg on the command line. I've tried different parameters inside "cpu_conf" but it doesn't want to compile the kernel.

Code:

"cpu_conf" :
[
{ "intensity" : 1, "worksize" : 1, "threads" : 1}
]

The error is like this:
https://i.imgur.com/NASYeio.png

Thank you.

[wacko]

I'm trying to use my CPU(Threadripper 1950x) also with SRBMiner v1.6.0 but i got an OpenCL error.

This is a GPU miner.  The"cpu_conf" part, as I understand, is for the CPUs with inbuilt GPUs that are opencl-capable (like intel chips). And not for the actual cpu cores. Miners like xmrig or xmr-stak is what you need for your threadripper.

[huntingthesnark]

What am I not doing please?

You didn't mention what you're mining. I think "Above target" means the wrong algo, there are 3 cryptonight algos on nicehash (regular, v7 and heavy), and you'd need to select the right one for the port you're aiming your miners at.

Yeah, you're right, it means NH isn't recognising the correct algo. Question is, how do you get the miner to tell NH? I've tried the obvious "normalv7" in config, and "nicehash" true in .bat, but no joy - immediate rejects... Mines XMR@nanopool quite happily on this setting...

NH vork fine with 6*vega56/64 + srbminer 1.6.0 and this option:

V7:
{"pool_use_tls" : false, "pool" : "cryptonightv7.eu.nicehash.com:3363", "wallet" : "adres", "password" : "x", "nicehash" : true, "max_difficulty" : 1700000},

heavy:
{"pool_use_tls" : false, "pool" : "cryptonightv7.eu.nicehash.com:3364", "wallet" : "adres", "password" : "x", "nicehash" : true, "max_difficulty" : 1700000},

and use  config file identical like nano or other algo

U use wrong port ...

Nice one - thanks for the config too - nice and clear.

"intensity" : 112 or 120 , "worksize" : 8 or 16, for v7
"intensity" : 56, "worksize" : 8 for heavy

[Iamtutut]

Hello @doktor83,

I'm trying to use my CPU(Threadripper 1950x) also with SRBMiner v1.6.0 but i got an OpenCL error. I used the default config from the readme file and the --usecpuopencl arg on the command line. I've tried different parameters inside "cpu_conf" but it doesn't want to compile the kernel.

Code:

"cpu_conf" :
[
{ "intensity" : 1, "worksize" : 1, "threads" : 1}
]

The error is like this:


Thank you.

Use JCE CN miner

[diansari]

what algo for ETN??

[Iamtutut]

normalv7

[dingdongtobias]

Link must be broken, new one https://mega.nz/#!OaYRVJTI!Z5dsgQTfMxlVrn9fNdqdNvG36eIAx4smfpagUxMiM2U. Watchdog has done a failed reset. It didn't mine after that last part so...

this is v 1.5.8, please use latest as maybe the bug you are experiencing is fixed.

Doktor why with versión 1.4.2 nofee i get 67-72 kh/s stellite and with 1.6.0 only 58-70 (miner show 70.5) same pools? A lot of less, not only 0.85%.

Tried 1.4.2 with Kraft and works fine.
Tried 1.6.0 with Kraft and down a lot the hashrate with same settings .

The miner show a bit more hashrate on 1.6.0, but in the pools show a lot less.

1.4.2 imagreedybastard -24 hours average hashrate 67-69kh/s
1.6.0                                  -24 hours average hashrate 60-62kh/s

You must be a magician if you can mine stellite with 1.4.2 version.

Also there is no Kraft coin. There is craft but that's not cryptonight algo.

sorry is Graft. and i have saved the hasrate in old algo stellite with 1.4.2. with the new algo mi hashrate drop 6-8 kh/s in pool, in srbminer shows more

All good here, 1.6.0

Miner avg : 4530 hs
Pool 24h : 4491 hs

https://i.imgur.com/rDcTeZU.jpg

[zot57]

How to get log file?

[dragonmike]

How to get log file?

add the following to your bat file: --logfile minerlog.txt

[doktor83]

V1.6.1
- Changed IPBC/Bittube old algo to new BitTubeV2 algo (from block 110000)
- Max difficulty reconnecting mechanism is now triggering instantly
- Added 'worksize' option in config.txt, that is applied to every GPU (only if not set in gpu_conf)
- Hopefuly fixed a bug with job timeout reconnecting
- Hashrate watchdog now triggers even when user not connected to pool


+ BitTube (former IPBC) forks on block 110.000, and their new algo will be BitTube V2.
You can use the new algo only after they forked, use "cryptonight_type" : "bittubev2". The old IPBC algo has been removed, so until the fork please use the previous version of SRBMiner.

+ Added 'worksize' option in config.txt (use it on top of config.txt , where intensity and double_threads is). If set, it will be applied to every GPU that is found. Of course if you have a gpu_conf set, those settings will be used, insted of this global setting.

+ Some had app crash on job_timeout triggering, i hope i managed to fix it

+ Hashrate watchdog was set to trigger only if user is connected to the pool. Now this is changed, watchdog will monitor GPU'S all the time.

[Kgonla]

Same problem under Server 2008 R2:

[2018-06-22 15:05:58] CryptonightV7 mode enabled
[2018-06-22 15:06:40] DevFee pools loaded
[2018-06-22 15:06:40] DevFee pool SET
[2018-06-22 15:06:40] DevFee address SET

42 sec. to load devfee pools, not firewalled.

[UnclWish]

Thanks for update, doc!

It's strange, but kernels from older versions didn't work with 1.6.1... Hope 1.6.1 can build fast kernels...

[doktor83]

Thanks for update, doc!

It's strange, but kernels from older versions didn't work with 1.6.1... Hope 1.6.1 can build fast kernels...

I did some optimisations in ocl kernels, that why old binaries wont work. well if its not fast enough for your taste you can use older versions or whatever is faster. 

[UnclWish]

Thanks for update, doc!

It's strange, but kernels from older versions didn't work with 1.6.1... Hope 1.6.1 can build fast kernels...

I did some optimisations in ocl kernels, that why old binaries wont work. well if its not fast enough for your taste you can use older versions or whatever is faster.  

Yes, I can't achieve max speed with new kernel builded by 1.6.1, rolling back to 1.6.0 with kernels created by 1.5.8... Your optimizations have something for 580 cards?

[doktor83]

Thanks for update, doc!

It's strange, but kernels from older versions didn't work with 1.6.1... Hope 1.6.1 can build fast kernels...

I did some optimisations in ocl kernels, that why old binaries wont work. well if its not fast enough for your taste you can use older versions or whatever is faster.  

Yes, I can't achieve max speed with new kernel builded by 1.6.1, rolling back to 1.6.0 with kernels created by 1.5.8... Your optimizations have something for 580 cards?

yes i managed to squeeze out +2hs on my 580 8g on v7 . its not much but hey, step by step. Also you are probably talking about heavy algo, let it run for 10 minutes on 1.6.1, i got the +1000hs in 90% of time on it. Still researching why is this happening, but had to push new version  because of the soon forking BitTube coin.