[ANN] sgminer v5 - optimized X11/X13/NeoScrypt/Lyra2RE/etc. kernel-switch miner - page 182.

bullus

sr. member

Activity: 294

Merit: 250

Quote from: badman74 on July 28, 2014, 04:14:25 PM

Quote from: bullus on July 28, 2014, 03:53:24 PM

My max with x11 is 6230kh/s but I can't achieve it everytime..usual it is 6160kh/s..
It happen often if I reboot my pc and start hashing with x11, but today that don't even work. Any explanation or do someone get the same?

There are some changes regarding
"blake-compact" : true,
"keccak-unroll" : "8",
"luffa-parallel" : true,

Are there already windows build out?

the mod in my sig
edit: i see it is out on the develop branch, will compile shortly
edit2: develop branch binaries added to my sig
add this to your config

Code:

"blake-compact" : true,
"keccak-unroll" : "8",
"luffa-parallel" : true,

Thanks for compile Wink

Tested 1050/1350 14.7rc1

X11=5930kh/s Sad

x13=4650kh/s
x15=4010kh/s

Tested 1050/1350 14.6rc1 (opencl files)

X11=6160kh/s

Code:

{
"profiles" : [
	{
		"name" : "x11",
		"algorithm" : "darkcoin-mod",
		"xintensity" : "64",
		"gpu-threads" : "2",
		"worksize": "64"
	},
	{
		"name" : "x13",
		"algorithm" : "marucoin-mod",
		"xintensity" : "51",
		"gpu-threads" : "2",
		"worksize": "64"
	},
	{
		"name" : "x15",
		"algorithm" : "bitblock",
		"xintensity" : "51",
		"gpu-threads" : "2",
		"worksize": "64"
	},
	{
		"name" : "nist5",
		"algorithm" : "talkcoin-mod",
		"intensity" : "16",
		"gpu-threads" : "2",
		"worksize": "64"
	}		
],
"default-profile": "x15",
"hamsi-expand-big" : "7",
"hamsi-short" : true,
"blake-compact" : true,
"keccak-unroll" : "8",
"luffa-parallel" : true,
"shaders" : "2816",
"gpu-fan" : "85-100",
"gpu-powertune" : "10",
"gpu-engine" : "1050",
"gpu-memclock" : "1350",
"gpu-vddc" : "0",
"auto-fan" : true,
"failover-only" : true,
"expiry" : "1",
"gpu-dyninterval" : "7",
"hotplug" : "5",
"log" : "5",
"queue" : "0",
"scan-time" : "1",
"temp-hysteresis" : "2",
"shares" : "0",
"no-submit-stale" : false,
"no-restart" : true,
"failover-switch-delay" : "30",
"show-coindiff" : true,
"remove-disabled" : true,
"extranonce-subscription" : true
}

badman74

hero member

Activity: 658

Merit: 500

Quote from: bullus on July 28, 2014, 03:53:24 PM

My max with x11 is 6230kh/s but I can't achieve it everytime..usual it is 6160kh/s..
It happen often if I reboot my pc and start hashing with x11, but today that don't even work. Any explanation or do someone get the same?

There are some changes regarding
"blake-compact" : true,
"keccak-unroll" : "8",
"luffa-parallel" : true,

Are there already windows build out?

the mod in my sig
edit: i see it is out on the develop branch, will compile shortly
edit2: develop branch binaries added to my sig
add this to your config

Code:

"blake-compact" : true,
"keccak-unroll" : "8",
"luffa-parallel" : true,

bullus

sr. member

Activity: 294

Merit: 250

My max with x11 is 6230kh/s but I can't achieve it everytime..usual it is 6160kh/s..
It happen often if I reboot my pc and start hashing with x11, but today that don't even work. Any explanation or do someone get the same?

There are some changes regarding
"blake-compact" : true,
"keccak-unroll" : "8",
"luffa-parallel" : true,

Are there already windows build out?

ebliever

legendary

Activity: 1708

Merit: 1036

Thanks; my rig's on a lease (which I consider hands-off periods) so I'll have to wait to try it, but I'll let you know if it makes a difference pro or con.

EDIT: The newer builds mentioned here didn't help any, I ended up reverted to my July 25 build for best stability.

I've been playing with engine/memory settings on individual GPUs. After a few hopeful minutes I keep getting SICK GPU's at anything above 1020/1350 on X11.

badman74

hero member

Activity: 658

Merit: 500

Quote from: ebliever on July 27, 2014, 09:35:48 PM

Quote from: badman74 on July 27, 2014, 08:38:09 PM

Quote from: platinum4 on July 27, 2014, 08:28:40 PM

Quote from: badman74 on July 27, 2014, 08:24:39 PM

Quote from: platinum4 on July 27, 2014, 08:18:20 PM

Quote from: badman74 on July 27, 2014, 08:07:27 PM

#define SPH_COMPACT_BLAKE_64 1 seems to give just a tiny bit more

Edited all .cl's; running experiment for uptime & stability; will report back.
Appears to add around 50Kh/s on x11.

seems like
#define SPH_LUFFA_PARALLEL 1 = 2%
#define SPH_COMPACT_BLAKE_64 1 = 1%
#define SPH_KECCAK_UNROLL 6 = 1%
substituted loops in groestl.cl = ~5%

I wonder if this is only for us on these Hawaii architectures.
Would it be worth us submitting a pull request to edit/update these four .cl files, or would it break compatibility for Tahitis & Pitcairns?

same seems to apply for my 7750's

Where can we get .cl's with these changes in the meantime?

check my sig
let me know if you find any that wont build or decrease in hashrate

ebliever

legendary

Activity: 1708

Merit: 1036

Quote from: badman74 on July 27, 2014, 08:38:09 PM

Quote from: platinum4 on July 27, 2014, 08:28:40 PM

Quote from: badman74 on July 27, 2014, 08:24:39 PM

Quote from: platinum4 on July 27, 2014, 08:18:20 PM

Quote from: badman74 on July 27, 2014, 08:07:27 PM

#define SPH_COMPACT_BLAKE_64 1 seems to give just a tiny bit more

Edited all .cl's; running experiment for uptime & stability; will report back.
Appears to add around 50Kh/s on x11.

seems like
#define SPH_LUFFA_PARALLEL 1 = 2%
#define SPH_COMPACT_BLAKE_64 1 = 1%
#define SPH_KECCAK_UNROLL 6 = 1%
substituted loops in groestl.cl = ~5%

I wonder if this is only for us on these Hawaii architectures.
Would it be worth us submitting a pull request to edit/update these four .cl files, or would it break compatibility for Tahitis & Pitcairns?

same seems to apply for my 7750's

Where can we get .cl's with these changes in the meantime?

platinum4

sr. member

Activity: 547

Merit: 250

Quote from: badman74 on July 27, 2014, 08:40:22 PM

Quote from: platinum4 on July 27, 2014, 08:34:25 PM

badman74, are you familiar with #pragma unroll?

Why can't we #pragma unroll 14 for the bottom two loops in groestl.cl? Is it because we are inside of a do-while?

Trying to #pragma unroll 14 the two for loops inside of the do-whiles doesn't work; it won't build the bin.

actually i have no idea
the only reason i found that for groestl.cl was because it was in aznboy84's darkcoin-mod.cl

Yeah if you look in his bitblock.cl, you will see something in the end, under the whirlpool section; take notice of his #pragma unroll 10

I'm wondering why we can't mimic such behavior but with #pragma unroll 14 for the r<14 for loops. I don't think we can unroll do-whiles, though.

platinum4

sr. member

Activity: 547

Merit: 250

Wow, yeah, I hit the mythical 6.2M as well.
The performance increase is a creeper, but it gets there after around 20 minutes.

badman74

hero member

Activity: 658

Merit: 500

Quote from: platinum4 on July 27, 2014, 08:34:25 PM

badman74, are you familiar with #pragma unroll?

Why can't we #pragma unroll 14 for the bottom two loops in groestl.cl? Is it because we are inside of a do-while?

Trying to #pragma unroll 14 the two for loops inside of the do-whiles doesn't work; it won't build the bin.

actually i have no idea
the only reason i found that for groestl.cl was because it was in aznboy84's darkcoin-mod.cl

badman74

hero member

Activity: 658

Merit: 500

Quote from: platinum4 on July 27, 2014, 08:28:40 PM

Quote from: badman74 on July 27, 2014, 08:24:39 PM

Quote from: platinum4 on July 27, 2014, 08:18:20 PM

Quote from: badman74 on July 27, 2014, 08:07:27 PM

#define SPH_COMPACT_BLAKE_64 1 seems to give just a tiny bit more

Edited all .cl's; running experiment for uptime & stability; will report back.
Appears to add around 50Kh/s on x11.

seems like
#define SPH_LUFFA_PARALLEL 1 = 2%
#define SPH_COMPACT_BLAKE_64 1 = 1%
#define SPH_KECCAK_UNROLL 6 = 1%
substituted loops in groestl.cl = ~5%

I wonder if this is only for us on these Hawaii architectures.
Would it be worth us submitting a pull request to edit/update these four .cl files, or would it break compatibility for Tahitis & Pitcairns?

same seems to apply for my 7750's
edit: these are capeverde

bullus

sr. member

Activity: 294

Merit: 250

1050/1350

platinum4

sr. member

Activity: 547

Merit: 250

badman74, are you familiar with #pragma unroll?

Why can't we #pragma unroll 14 for the bottom two loops in groestl.cl? Is it because we are inside of a do-while?

Trying to #pragma unroll 14 the two for loops inside of the do-whiles doesn't work; it won't build the bin.

platinum4

sr. member

Activity: 547

Merit: 250

Quote from: badman74 on July 27, 2014, 08:24:39 PM

Quote from: platinum4 on July 27, 2014, 08:18:20 PM

Quote from: badman74 on July 27, 2014, 08:07:27 PM

#define SPH_COMPACT_BLAKE_64 1 seems to give just a tiny bit more

Edited all .cl's; running experiment for uptime & stability; will report back.
Appears to add around 50Kh/s on x11.

seems like
#define SPH_LUFFA_PARALLEL 1 = 2%
#define SPH_COMPACT_BLAKE_64 1 = 1%
#define SPH_KECCAK_UNROLL 6 = 1%
substituted loops in groestl.cl = ~5%

I wonder if this is only for us on these Hawaii architectures.
Would it be worth us submitting a pull request to edit/update these four .cl files, or would it break compatibility for Tahitis & Pitcairns?

badman74

hero member

Activity: 658

Merit: 500

Quote from: platinum4 on July 27, 2014, 08:18:20 PM

Quote from: badman74 on July 27, 2014, 08:07:27 PM

#define SPH_COMPACT_BLAKE_64 1 seems to give just a tiny bit more

Edited all .cl's; running experiment for uptime & stability; will report back.
Appears to add around 50Kh/s on x11.

seems like
#define SPH_LUFFA_PARALLEL 1 = 2%
#define SPH_COMPACT_BLAKE_64 1 = 1%
#define SPH_KECCAK_UNROLL 6 = 1%
substituted loops in groestl.cl = ~5%

platinum4

sr. member

Activity: 547

Merit: 250

Quote from: badman74 on July 27, 2014, 08:07:27 PM

#define SPH_COMPACT_BLAKE_64 1 seems to give just a tiny bit more

Edited all .cl's; running experiment for uptime & stability; will report back.
Appears to add around 50Kh/s on x11.

badman74

hero member

Activity: 658

Merit: 500

#define SPH_COMPACT_BLAKE_64 1 seems to give just a tiny bit more

badman74

hero member

Activity: 658

Merit: 500

Quote from: Hueristic on July 27, 2014, 05:45:19 PM

Quote from: badman74 on July 27, 2014, 05:10:07 PM

Quote from: Hueristic on July 27, 2014, 03:58:50 PM

Quote from: Hueristic on July 27, 2014, 11:38:41 AM

TLDR

Does this have lasybear improvements?

And where is a link to the best for multi use? Seems to have to many versions around.

also I could use a generic safe setting for r9 270 for each algo.

Thx guys

These are simple answers not covered in the OP.

ok
#1 yes lasybear improvements (darkcoin-mod) is included (in the OP)
#2 not sure if you are looking for a binary download or a website to use it on.... (also in the OP)
#3 the only thing really different with each algo is whether you use 1, 2, 4 threads, and your Intensity settings (most general stuff still in the OP)
unless you want to overclock your cards these are the only things you need to worry about
TC settings and such are only used on scrypt/nscrypt and these need to be found per card type
if all else fails read https://github.com/sgminer-dev/sgminer/blob/v5_0/doc/configuration.md for your configuration needs (found in the OP....)
only other thing that people commonly mess up is the algo's need to be set with algorithm not -k or kernel (as seen in the example conf on the OP)

Thx, completely missed lasybear in op. For some reason all I'm getting is hardware errors even with 13int and only 8196tc single threads on X15? I'll start again tomorrow from scratch.

Ohh can I use the Bin file from a SPH? My x13 is perfect and I don't want to lose it.

you can probably use your old bin, scrypt/nscrypt are the ONLY algo's that look at TC, and are you using --algorithm x15 if you try to use -k or kernel it will build a ckolivas bin instead

here is my configuration

Code:

{
"pools" : [
	{
		"name" : "X11P2P.com",
		"url" : "http://x11p2p.com:7903",
		"user" : "XvRs67Wnw1n1Kwtq4ihhTpv699DdJpxdnt+.02",
		"pass" : "x",
		"priority" : "0",
		"no-extranonce" : true,
		"profile": "x11"
	},
	{
		"name" : "MineP.it DRK",
		"url" : "STRATUM+TCP://DRK.MINEP.IT:3333",
		"user" : "badman74.1",
		"pass" : "x",
		"priority" : "1",
		"profile": "x11"
	},
	{
		"name" : "Xpool X11",
		"url" : "stratum+tcp://mine.xpool.ca:8888",
		"user" : "XvRs67Wnw1n1Kwtq4ihhTpv699DdJpxdnt",
		"pass" : "x",
		"priority" : "2",
		"profile": "x11"
	}
],
"profiles" : [
	{
		"name" : "x11",
		"algorithm" : "x11"
	},
	{
		"name" : "x13",
		"algorithm" : "x13"
	},
	{
		"name" : "x14",
		"algorithm" : "x14"
	},
	{
		"name" : "x15",
		"algorithm" : "x15"
	},
	{
		"name" : "NIST5",
		"algorithm" : "nist5"
	},
	{
		"name" : "scrypt",
		"lookup-gap" : "2",
		"gpu-engine": "1000,1000",
		"gpu-memclock" : "1250,1250",
		"gpu-powertune" : "0",
		"gpu-threads": "1",
		"algorithm" : "zuikkis"
	},
	{
		"name" : "nscrypt",
		"lookup-gap" : "2",
		"gpu-engine": "1000,1000",
		"gpu-memclock" : "1250,1250",
		"gpu-powertune" : "0",
		"algorithm" : "zuikkis",
		"nfactor" : "11"
	},
	{
		"name" : "keccak",
		"intensity" : "12",
		"gpu-threads": "4",
		"algorithm" : "keccak"
	},
	{
		"name" : "qubit",
		"intensity" : "18",
		"algorithm" : "qubitcoin"
	},
	{
		"name" : "quark",
		"intensity" : "18",
		"algorithm" : "quarkcoin"
	}
],
"intensity" : "16",
"worksize": "64",
"hamsi-expand-big" : "7",
"hamsi-short" : true,
"gpu-map" : "0:0,1:1",
"gpu-fan" : "50-100",
"gpu-engine": "1040,1040",
"gpu-memclock" : "1500,1500",
"gpu-threads" : "2",
"gpu-powertune" : "15",
"failover-only" : true,
"temp-cutoff" : "96",
"temp-overheat" : "94",
"temp-target" : "85",
"auto-fan" : true,
"auto-gpu" : false,
"api-allow" : "W:127.0.0.1",
"api-listen" : true,
"api-port" : "4028",
"expiry" : "1",
"gpu-dyninterval" : "7",
"hotplug" : "5",
"log-file" : "logfile.txt",
"log-show-date" : true,
"log" : "5",
"net-delay" : true,
"queue" : "0",
"scan-time" : "1",
"temp-hysteresis" : "5",
"shares" : "0",
"no-submit-stale" : false,
"no-restart" : false,
"failover-switch-delay" : "30",
"show-coindiff" : true,
"remove-disabled" : true
}

Hueristic

legendary

Activity: 3836

Merit: 4969

Doomed to see the future and unable to prevent it

Quote from: badman74 on July 27, 2014, 05:10:07 PM

Quote from: Hueristic on July 27, 2014, 03:58:50 PM

Quote from: Hueristic on July 27, 2014, 11:38:41 AM

TLDR

Does this have lasybear improvements?

And where is a link to the best for multi use? Seems to have to many versions around.

also I could use a generic safe setting for r9 270 for each algo.

Thx guys

These are simple answers not covered in the OP.

ok
#1 yes lasybear improvements (darkcoin-mod) is included (in the OP)
#2 not sure if you are looking for a binary download or a website to use it on.... (also in the OP)
#3 the only thing really different with each algo is whether you use 1, 2, 4 threads, and your Intensity settings (most general stuff still in the OP)
unless you want to overclock your cards these are the only things you need to worry about
TC settings and such are only used on scrypt/nscrypt and these need to be found per card type
if all else fails read https://github.com/sgminer-dev/sgminer/blob/v5_0/doc/configuration.md for your configuration needs (found in the OP....)
only other thing that people commonly mess up is the algo's need to be set with algorithm not -k or kernel (as seen in the example conf on the OP)

Thx, completely missed lasybear in op. For some reason all I'm getting is hardware errors even with 13int and only 8196tc single threads on X15? I'll start again tomorrow from scratch.

Ohh can I use the Bin file from a SPH? My x13 is perfect and I don't want to lose it.

badman74

hero member

Activity: 658

Merit: 500

Quote from: Hueristic on July 27, 2014, 03:58:50 PM

Quote from: Hueristic on July 27, 2014, 11:38:41 AM

TLDR

Does this have lasybear improvements?

And where is a link to the best for multi use? Seems to have to many versions around.

also I could use a generic safe setting for r9 270 for each algo.

Thx guys

These are simple answers not covered in the OP.

ok
#1 yes lasybear improvements (darkcoin-mod) is included (in the OP)
#2 not sure if you are looking for a binary download or a website to use it on.... (also in the OP)
#3 the only thing really different with each algo is whether you use 1, 2, 4 threads, and your Intensity settings (most general stuff still in the OP)
unless you want to overclock your cards these are the only things you need to worry about
TC settings and such are only used on scrypt/nscrypt and these need to be found per card type
if all else fails read https://github.com/sgminer-dev/sgminer/blob/v5_0/doc/configuration.md for your configuration needs (found in the OP....)
only other thing that people commonly mess up is the algo's need to be set with algorithm not -k or kernel (as seen in the example conf on the OP)

Hueristic

legendary

Activity: 3836

Merit: 4969

Doomed to see the future and unable to prevent it

Quote from: Hueristic on July 27, 2014, 11:38:41 AM

TLDR

Does this have lasybear improvements?

And where is a link to the best for multi use? Seems to have to many versions around.

also I could use a generic safe setting for r9 270 for each algo.

Thx guys

These are simple answers not covered in the OP.

Topic: [ANN] sgminer v5 - optimized X11/X13/NeoScrypt/Lyra2RE/etc. kernel-switch miner - page 182. (Read 877859 times)