[ANN] sgminer v5 - optimized X11/X13/NeoScrypt/Lyra2RE/etc. kernel-switch miner

[B.T.Coin]

it appears that i can't build this from the current source...
don't know if it is a problem on my end or something in the code

looking forward to rebuilding out my farm with this... will keep watching!

thxs for all the hard work devs and ops.

ok i have it working now
https://mega.co.nz/#!fVIVxSZD!1Fqnlm8vu-N2QbVdwXZvohIPD8arXdYD6wKYbCR4-iU
https://www.dropbox.com/s/20hcs4j5unpjjxy/sgminer_multi.rar

cant get old config working with this
always using cholivas kernel

it should have worked...
here is newest build
https://mega.co.nz/#!jJZwybCA!xZMgjAPymCRHCjGe-gPSrstUjLj4XP9K89Rem0NTaOI
https://www.dropbox.com/s/20hcs4j5unpjjxy/sgminer_multi.rar

Still the same problem. It will not compile the darkcoin or marucoin kernels and just use ckolivas and spit hardware errors.

This one should work on catalyst 14.6 right?

[badman74]

it appears that i can't build this from the current source...
don't know if it is a problem on my end or something in the code

looking forward to rebuilding out my farm with this... will keep watching!

thxs for all the hard work devs and ops.

ok i have it working now
https://mega.co.nz/#!fVIVxSZD!1Fqnlm8vu-N2QbVdwXZvohIPD8arXdYD6wKYbCR4-iU
https://www.dropbox.com/s/20hcs4j5unpjjxy/sgminer_multi.rar

cant get old config working with this
always using cholivas kernel

it should have worked...
here is newest build
https://mega.co.nz/#!jJZwybCA!xZMgjAPymCRHCjGe-gPSrstUjLj4XP9K89Rem0NTaOI
https://www.dropbox.com/s/20hcs4j5unpjjxy/sgminer_multi.rar

Still the same problem. It will not compile the darkcoin or marucoin kernels and just use ckolivas and spit hardware errors.

This one should work on catalyst 14.6 right?

i tried the 14.6 catalyst yesterday and every version of sgminer i had gave nothing but HW errors on every kernel.
so i am now back on 13.2

[badman74]

Seems like it is not working, after 30sec. or so it just quits the miner.
using sgminer_v5_10062014 no problems

i found that trying to use the new profile options caused the 60 sec quit problem

Are y'all trying to use pool-gpu-threads?  This knowingly fucks up on Windows binaries.

my config but got driver crashes and hw errors, not hashing yet

Code:

{
"pools" : [
        {
 		"name" : "am01 x11 multi",
		"url" : "stratum+tcp://am02.eu.trademybit.com:4010",
		"user" : "xx",
		"pass" : "xx",
                "profile":"x11"
	},
        {
		"name" : "am02 x13 multi",
		"url" : "stratum+tcp://am02.eu.trademybit.com:4011",
		"user" : "xx",
		"pass" : "xx",
                "profile":"x13,x13old,x13,x13,x13"
	}
  ],
  "profiles": [
    {
      "name": "x11",
      "algorithm": "darkcoin-mod",
      "intensity": "18",
      "device": "0,1,2,3,4",
      "thread-concurrency": "8192,8000,8192,8192,8192",
      "worksize": "256,64,256,256,256",
      "gpu-threads": "2"
    },
    {
      "name": "x13old",
      "algorithm": "marucoin-modold",
      "device": "1",
      "intensity": "18",
      "shaders" : "1536",
      "thread-concurrency": "8000",
      "worksize": "64",
      "gpu-threads": "2"
    },
    {
      "name": "x13",
      "algorithm": "marucoin-mod",
      "device": "0,2,3,4",
      "intensity": "18",
      "shaders" : "1792",
      "thread-concurrency": "8192",
      "worksize": "256",
      "gpu-threads": "2"
    }
  ],

"extranonce-subscribe" : true,
"remove-disabled" : true,
"api-allow" : "W:127.0.0.1",
"api-listen" : true,
"api-mcast-port" : "4028",
"api-port" : "4028",
"expiry" : "10",
"failover-only" : true,
"log" : "5",
"queue" : "1",
"scan-time" : "7",
"shares" : "0",
"failover-switch-delay" : "10",
"gpu-threads" : "2",
"gpu-dyninterval" : "7",
"gpu-engine" : "0",
"gpu-fan" : "65",
"gpu-platform" : "0",
"gpu-memclock" : "0",
"gpu-memdiff" : "0",
"gpu-powertune" : "8",
"gpu-vddc" : "0",
"temp-target" : "75",
"temp-overheat" : "85",
"temp-cutoff" : "95",
"temp-hysteresis" : "3",
"vectors" : "1",
"lookup-gap" : "0",
"hamsi-expand-big" : "4",
"no-pool-disable" : true
}

7950-6970-7950-7950-7950

I have nothing good to offer; conceptually this feels as if it should parse ok; the only issue I see is the 8000 TC and worksize 64, I'm not sure if it will accept such settings (but unfamiliar with it, so don't take my word as gold)  

Is there a reason you are specifying gpu-threads in each profile, and is there a reason you are running 2 threads as opposed to 1?  If you specify it globally at the bottom, then I would omit it elsewhere if you aren't changing it.

Running it for ages with TCC 8000 and -w 64 for my 6970 
I can change that, but I notice it is not creating the darkcoin-mod or marucoin-mod bin files and it's using the ckolivas bin as default, that's why I have a lots of hw-errors and driver crash

try removing TC alltogether and see if it will work
it doesnt seem to use it for anything but scrypt

[phzi]

Ok, need to check that..
It is possible to use multi profiles like?

"profile" : "x13,x13old,x13,x13"

I doubt it... that configuration doesn't really make any sense.  If you need to specify per-GPU settings, you do it IN the profile. 

e.g.

Code:

"profiles": [
 {
      "name": "x13",
      "algorithm": "marucoin-mod,marucoin-modold,marucoin-mod,marucoin-mod",
      "intensity": "18",
      "shaders" : "1792,1536,1792,1792",
      "thread-concurrency": "8192,8000,8192,8192",
      "worksize": "256,64,256,256",
      "gpu-threads": "2"
    }

[bullus]

Still it's using the ckovalis kernel and not compiling the darcoin or marucoin bins

Code:

{
"pools" : [
        {
 		"name" : "am01 x11 multi",
		"url" : "stratum+tcp://am02.eu.trademybit.com:4010",
		"user" : "xxx",
		"pass" : "xxx",
                "profile" : "x11"
	},
        {
		"name" : "am02 x13 multi",
		"url" : "stratum+tcp://am02.eu.trademybit.com:4011",
		"user" : "xx",
		"pass" : "xx",
                "profile" : "x13"
	}
],
  "profiles": [
    {
      "name": "x11",
      "algorithm" : "darkcoin-mod"
    },
    {
      "name": "x13",
      "algorithm" : "marucoin-mod,marucoin-modold,marucoin-mod,marucoin-mod,marucoin-mod"
    }
  ],
"extranonce-subscribe" : true,
"default-profile" : "darkcoin-mod",
"remove-disabled" : true,
"api-allow" : "W:127.0.0.1",
"api-listen" : true,
"api-mcast-port" : "4028",
"api-port" : "4028",
"expiry" : "10",
"failover-only" : true,
"log" : "5",
"queue" : "1",
"scan-time" : "7",
"shares" : "0",
"kernel-path" : "/usr/local/bin",
"failover-switch-delay" : "10",
"gpu-threads" : "2",
"gpu-dyninterval" : "7",
"gpu-engine" : "0",
"gpu-fan" : "65",
"gpu-platform" : "0",
"gpu-memclock" : "0",
"gpu-memdiff" : "0",
"gpu-powertune" : "8",
"gpu-vddc" : "0",
"temp-target" : "75",
"temp-overheat" : "85",
"temp-cutoff" : "95",
"temp-hysteresis" : "3",
"vectors" : "1",
"lookup-gap" : "0",
"gpu-threads": "2",
"device": "0,1,2,3,4",
"hamsi-expand-big" : "4",
"no-pool-disable" : true
}

[B.T.Coin]

i tried the 14.6 catalyst yesterday and every version of sgminer i had gave nothing but HW errors on every kernel.
so i am now back on 13.2

That's true, but still your latest version does not compile the kernels and just runs with ckolivas.
I thought the latest SGminer had a fix for catalyst 14.6, that's why I'm trying this version.
What's all this talk then about people claiming huge gains when using 14.6? What miner version are they using?

[platinum4]

Still it's using the ckovalis kernel and not compiling the darcoin or marucoin bins

Code:

{
"pools" : [
        {
 		"name" : "am01 x11 multi",
		"url" : "stratum+tcp://am02.eu.trademybit.com:4010",
		"user" : "xxx",
		"pass" : "xxx",
                "profile" : "x11"
	},
        {
		"name" : "am02 x13 multi",
		"url" : "stratum+tcp://am02.eu.trademybit.com:4011",
		"user" : "xx",
		"pass" : "xx",
                "profile" : "x13"
	}
],
  "profiles": [
    {
      "name": "x11",
      "algorithm" : "darkcoin-mod"
    },
    {
      "name": "x13",
      "algorithm" : "marucoin-mod,marucoin-modold,marucoin-mod,marucoin-mod,marucoin-mod"
    }
  ],
"extranonce-subscribe" : true,
"default-profile" : "darkcoin-mod",
"remove-disabled" : true,
"api-allow" : "W:127.0.0.1",
"api-listen" : true,
"api-mcast-port" : "4028",
"api-port" : "4028",
"expiry" : "10",
"failover-only" : true,
"log" : "5",
"queue" : "1",
"scan-time" : "7",
"shares" : "0",
"kernel-path" : "/usr/local/bin",
"failover-switch-delay" : "10",
"gpu-threads" : "2",
"gpu-dyninterval" : "7",
"gpu-engine" : "0",
"gpu-fan" : "65",
"gpu-platform" : "0",
"gpu-memclock" : "0",
"gpu-memdiff" : "0",
"gpu-powertune" : "8",
"gpu-vddc" : "0",
"temp-target" : "75",
"temp-overheat" : "85",
"temp-cutoff" : "95",
"temp-hysteresis" : "3",
"vectors" : "1",
"lookup-gap" : "0",
"gpu-threads": "2",
"device": "0,1,2,3,4",
"hamsi-expand-big" : "4",
"no-pool-disable" : true
}

Shit guys, maybe open a github issue for it?  This is major fucking bad hiccup in the development of this unified sgminer if we can't get it back to spitting out the kernels correctly.  Is it fucking up for everybody, or just since the addition of ystarnaud's config parser and profiles?

As I stated before I'm running some sort of drag and drop sgminer_v5_multi that badman had wrapped up.  It might have been the multi most recently released, dragged and dropped into the link from 6/16.  Yes, it doesn't use profiles, but then again, yes it does perform hashing on the specified kernels/algos and flips algos softly and nicely.

Something fucked up in the github itself?  No reason why 2 or 3 different people compiling are running in to the same shit over and over again.

[bullus]

Still it's using the ckovalis kernel and not compiling the darcoin or marucoin bins

Code:

{
"pools" : [
        {
 		"name" : "am01 x11 multi",
		"url" : "stratum+tcp://am02.eu.trademybit.com:4010",
		"user" : "xxx",
		"pass" : "xxx",
                "profile" : "x11"
	},
        {
		"name" : "am02 x13 multi",
		"url" : "stratum+tcp://am02.eu.trademybit.com:4011",
		"user" : "xx",
		"pass" : "xx",
                "profile" : "x13"
	}
],
  "profiles": [
    {
      "name": "x11",
      "algorithm" : "darkcoin-mod"
    },
    {
      "name": "x13",
      "algorithm" : "marucoin-mod,marucoin-modold,marucoin-mod,marucoin-mod,marucoin-mod"
    }
  ],
"extranonce-subscribe" : true,
"default-profile" : "darkcoin-mod",
"remove-disabled" : true,
"api-allow" : "W:127.0.0.1",
"api-listen" : true,
"api-mcast-port" : "4028",
"api-port" : "4028",
"expiry" : "10",
"failover-only" : true,
"log" : "5",
"queue" : "1",
"scan-time" : "7",
"shares" : "0",
"kernel-path" : "/usr/local/bin",
"failover-switch-delay" : "10",
"gpu-threads" : "2",
"gpu-dyninterval" : "7",
"gpu-engine" : "0",
"gpu-fan" : "65",
"gpu-platform" : "0",
"gpu-memclock" : "0",
"gpu-memdiff" : "0",
"gpu-powertune" : "8",
"gpu-vddc" : "0",
"temp-target" : "75",
"temp-overheat" : "85",
"temp-cutoff" : "95",
"temp-hysteresis" : "3",
"vectors" : "1",
"lookup-gap" : "0",
"gpu-threads": "2",
"device": "0,1,2,3,4",
"hamsi-expand-big" : "4",
"no-pool-disable" : true
}

Shit guys, maybe open a github issue for it?  This is major fucking bad hiccup in the development of this unified sgminer if we can't get it back to spitting out the kernels correctly.  Is it fucking up for everybody, or just since the addition of ystarnaud's config parser and profiles?

As I stated before I'm running some sort of drag and drop sgminer_v5_multi that badman had wrapped up.  It might have been the multi most recently released, dragged and dropped into the link from 6/16.  Yes, it doesn't use profiles, but then again, yes it does perform hashing on the specified kernels/algos and flips algos softly and nicely.

Something fucked up in the github itself?  No reason why 2 or 3 different people compiling are running in to the same shit over and over again.

Can u post ur config?
With the new profiles update I can't get it working yet. (windows 8.1, driver 14.4)

edit

Instead of using profiles, I used "algorithm" : "darcoin-mod"  and that worked for x11, when switching to x13 no hashrates...back to miner14.6

[kenshirothefist]

We have published new binaries (linux, windows), build today ... you already know the most notable updates:

- completely new configuration options with profiles (remember to replace any pool-xxx setting with only xxx), take your time and read thoroughly all the details here: https://github.com/sgminer-dev/sgminer/blob/v5_0/doc/configuration.md
- support for 14.6 drivers
- new x11 (darkcoin-mod) and x13 (marucoin-mod) kernels (make sure you delete your old .bin files before running new version!)
- optimized nist5 (talkcoin-mod) kernel
- numerous bugfixes, also related to algorithm switching

https://www.nicehash.com/software/#sgminer

[B.T.Coin]

Still it's using the ckovalis kernel and not compiling the darcoin or marucoin bins

Code:

{
"pools" : [
        {
 		"name" : "am01 x11 multi",
		"url" : "stratum+tcp://am02.eu.trademybit.com:4010",
		"user" : "xxx",
		"pass" : "xxx",
                "profile" : "x11"
	},
        {
		"name" : "am02 x13 multi",
		"url" : "stratum+tcp://am02.eu.trademybit.com:4011",
		"user" : "xx",
		"pass" : "xx",
                "profile" : "x13"
	}
],
  "profiles": [
    {
      "name": "x11",
      "algorithm" : "darkcoin-mod"
    },
    {
      "name": "x13",
      "algorithm" : "marucoin-mod,marucoin-modold,marucoin-mod,marucoin-mod,marucoin-mod"
    }
  ],
"extranonce-subscribe" : true,
"default-profile" : "darkcoin-mod",
"remove-disabled" : true,
"api-allow" : "W:127.0.0.1",
"api-listen" : true,
"api-mcast-port" : "4028",
"api-port" : "4028",
"expiry" : "10",
"failover-only" : true,
"log" : "5",
"queue" : "1",
"scan-time" : "7",
"shares" : "0",
"kernel-path" : "/usr/local/bin",
"failover-switch-delay" : "10",
"gpu-threads" : "2",
"gpu-dyninterval" : "7",
"gpu-engine" : "0",
"gpu-fan" : "65",
"gpu-platform" : "0",
"gpu-memclock" : "0",
"gpu-memdiff" : "0",
"gpu-powertune" : "8",
"gpu-vddc" : "0",
"temp-target" : "75",
"temp-overheat" : "85",
"temp-cutoff" : "95",
"temp-hysteresis" : "3",
"vectors" : "1",
"lookup-gap" : "0",
"gpu-threads": "2",
"device": "0,1,2,3,4",
"hamsi-expand-big" : "4",
"no-pool-disable" : true
}

Shit guys, maybe open a github issue for it?  This is major fucking bad hiccup in the development of this unified sgminer if we can't get it back to spitting out the kernels correctly.  Is it fucking up for everybody, or just since the addition of ystarnaud's config parser and profiles?

As I stated before I'm running some sort of drag and drop sgminer_v5_multi that badman had wrapped up.  It might have been the multi most recently released, dragged and dropped into the link from 6/16.  Yes, it doesn't use profiles, but then again, yes it does perform hashing on the specified kernels/algos and flips algos softly and nicely.

Something fucked up in the github itself?  No reason why 2 or 3 different people compiling are running in to the same shit over and over again.

I think I figured out what's wrong.
CGwatcher will change the "algorithm" argument to "pool-algorithm" when used with a pool (and possibly also in a profile, I don't use these so can not test)
So we either have to wait for a new version of CGwatcher or run without CGwatcher

[bullus]

Still it's using the ckovalis kernel and not compiling the darcoin or marucoin bins

Code:

{
"pools" : [
        {
 		"name" : "am01 x11 multi",
		"url" : "stratum+tcp://am02.eu.trademybit.com:4010",
		"user" : "xxx",
		"pass" : "xxx",
                "profile" : "x11"
	},
        {
		"name" : "am02 x13 multi",
		"url" : "stratum+tcp://am02.eu.trademybit.com:4011",
		"user" : "xx",
		"pass" : "xx",
                "profile" : "x13"
	}
],
  "profiles": [
    {
      "name": "x11",
      "algorithm" : "darkcoin-mod"
    },
    {
      "name": "x13",
      "algorithm" : "marucoin-mod,marucoin-modold,marucoin-mod,marucoin-mod,marucoin-mod"
    }
  ],
"extranonce-subscribe" : true,
"default-profile" : "darkcoin-mod",
"remove-disabled" : true,
"api-allow" : "W:127.0.0.1",
"api-listen" : true,
"api-mcast-port" : "4028",
"api-port" : "4028",
"expiry" : "10",
"failover-only" : true,
"log" : "5",
"queue" : "1",
"scan-time" : "7",
"shares" : "0",
"kernel-path" : "/usr/local/bin",
"failover-switch-delay" : "10",
"gpu-threads" : "2",
"gpu-dyninterval" : "7",
"gpu-engine" : "0",
"gpu-fan" : "65",
"gpu-platform" : "0",
"gpu-memclock" : "0",
"gpu-memdiff" : "0",
"gpu-powertune" : "8",
"gpu-vddc" : "0",
"temp-target" : "75",
"temp-overheat" : "85",
"temp-cutoff" : "95",
"temp-hysteresis" : "3",
"vectors" : "1",
"lookup-gap" : "0",
"gpu-threads": "2",
"device": "0,1,2,3,4",
"hamsi-expand-big" : "4",
"no-pool-disable" : true
}

Shit guys, maybe open a github issue for it?  This is major fucking bad hiccup in the development of this unified sgminer if we can't get it back to spitting out the kernels correctly.  Is it fucking up for everybody, or just since the addition of ystarnaud's config parser and profiles?

As I stated before I'm running some sort of drag and drop sgminer_v5_multi that badman had wrapped up.  It might have been the multi most recently released, dragged and dropped into the link from 6/16.  Yes, it doesn't use profiles, but then again, yes it does perform hashing on the specified kernels/algos and flips algos softly and nicely.

Something fucked up in the github itself?  No reason why 2 or 3 different people compiling are running in to the same shit over and over again.

I think I figured out what's wrong.
CGwatcher will change the "algorithm" argument to "pool-algorithm" when used with a pool (and possibly also in a profile, I don't use these so can not test)
So we either have to wait for a new version of CGwatcher or run without CGwatcher

I'm not using cgwatcher either, but it seems the latest build is running ok without using profiles, but got lower hashrates than the 14.6miner

[platinum4]

I'm not using cgwatcher either, but it seems the latest build is running ok without using profiles, but got lower hashrates than the 14.6miner

Deleted your .bin files right, and checked your hamsi-expand-big?

[bullus]

I'm not using cgwatcher either, but it seems the latest build is running ok without using profiles, but got lower hashrates than the 14.6miner

Deleted your .bin files right, and checked your hamsi-expand-big?

I didn't install 14.6yet but just copied the opencl and amdocl files into the miner but no increase, with 14.6miner it was working like that and offcourse deleted the bins

[platinum4]

I'm not using cgwatcher either, but it seems the latest build is running ok without using profiles, but got lower hashrates than the 14.6miner

Deleted your .bin files right, and checked your hamsi-expand-big?

I didn't install 14.6yet but just copied the opencl and amdocl files into the miner but no increase, with 14.6miner it was working like that and offcourse deleted the bins

So wait, you saw performance increase on sgminer-14.6, even though you were using 14.4 drivers and not had 14.6 installed?

That's kind of confusing.

[DarkAGeS]

you don't need to install 14.6
just use right bin-files and thats all.

for those who have hw in 14.6
change in darkcoin-mod.cl and marucoin-mod.cl

#define SPH_T32(x) ((x) & SPH_C32(0xFFFFFFFF))
#define SPH_ROTL32(x, n) SPH_T32(((x) << (n)) | ((x) >> (32 - (n))))

#define SPH_T64(x) ((x) & SPH_C64(0xFFFFFFFFFFFFFFFF))
#define SPH_ROTL64(x, n) SPH_T64(((x) << (n)) | ((x) >> (64 - (n))))

to:

#define SPH_T32(x) (as_uint(x))
#define SPH_ROTL32(x, n) rotate(as_uint(x), as_uint(n))

#define SPH_T64(x) (as_ulong(x))
#define SPH_ROTL64(x, n) rotate(as_ulong(x), (n) & 0xFFFFFFFFFFFFFFFFUL)

delete your bin-files and try again

and please put this bin-files somewhere and give a link

[oktay50000]

anyone can post working config with this new miner??/
 i cant get it to work
always building cholivash kernel

[platinum4]

anyone can post working config with this new miner??/
 i cant get it to work
always building cholivash kernel

That's what badman74 & bullus are reporting in too.

Open a github issue?

[Ninetoe]

after 60 seconds the following happens

thread was not cancelled in 60 seconds after restart_mining_threads

driver 14.6
5x 280x

i fixed the default of ckolivas to not specify a default profile and put the algo in the global config.

[oktay50000]

after 60 seconds the following happens

thread was not cancelled in 60 seconds after restart_mining_threads

driver 14.6
5x 280x

i fixed the default of ckolivas to not specify a default profile and put the algo in the global config.

how fixed cholivash???

[B.T.Coin]

anyone can post working config with this new miner??/
 i cant get it to work
always building cholivash kernel

Do not use "pool-algorithm" anymore but use "algorithm" (also in the pool section) or it will not work.

So on my machine it does build the kernels now but my drivers (14.6) keep crashing for one of my cards no matter what settings I try.
It's not just a simple copy-paste your old config when you want to step up to 14.6 I'm afraid