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

[bigblind]

I'm using the Version of badman74 signature, sgminer 5.0 mod with an 14.6 RC3 Catalyst on 3 R9 290s.
Can anybody tell me why the Hashrates are "just" at about 15 and only go up sometimes to about 19-20?

Can't explain that to me. I'm using the cards to provide NIST5 hashs for Nicehash.

Here is my config:

{
"pools" : [
   {
      "name" : "nicehas.nist5",
      "url" : "stratum+tcp://stratum.nicehash.com:3340",
      "user" : "xxxxxx",
      "pass" : "x",
      "pool-priority" : "0",
      "nfactor" : "10",
      "algorithm" : "talkcoin-mod"
   }
],
"api-allow" : "xxxxxx",
"api-listen" : true,
"expiry" : "1",
"failover-only" : true,
"queue" : "0",
"scan-time" : "1",
"gpu-threads" : "2",
"gpu-engine" : "947,947,947",
"gpu-fan" : "70,70,70",
"gpu-memclock" : "1250,1250,1250",
"gpu-powertune" : "20,20,20",
"intensity" : "17",
"temp-target" : "85",
"temp-overheat" : "90",
"temp-cutoff" : "95",
"worksize" : "128",
"lookup-gap" : "2,2,2",
"shaders" : "2560,2560,2560",
"thread-concurrency" : "8192",
"sharethreads" : "32,32,32",
"no-pool-disable" : true,
"tcp-keepalive" : "30"
}

[1714730548]

1714730548

[1714730548]

1714730548

[bigblind]

how much should R9290 usually be hitting?

[OmarGsPools]

Is anyone capable of integrating these kenels into SG-Miner v5? These are heavily optimized compared to the current qubit kernel.

https://github.com/MaxDZ8/M8M/tree/master/M8M/kernels

Also if anyone needs a pool for a coin let me know and I can open one up. Just trying to help contribute to the mining community.

Curious is yours working, the most recent build? I just keep getting socket closed even if I run it as Administrator (Failed Send, Socket Reset)

Working but only when connecting to a P2Pool. A bit unfortunate since I run my own pools.

[SavageWS6]

Is anyone capable of integrating these kenels into SG-Miner v5? These are heavily optimized compared to the current qubit kernel.

https://github.com/MaxDZ8/M8M/tree/master/M8M/kernels

Also if anyone needs a pool for a coin let me know and I can open one up. Just trying to help contribute to the mining community.

Curious is yours working, the most recent build? I just keep getting socket closed even if I run it as Administrator (Failed Send, Socket Reset)

Working but only when connecting to a P2Pool. A bit unfortunate since I run my own pools.

Ah gotcha. I was also connecting to a P2Pool myself, and kept getting that error. I'll check for a later build then since I'm not sure why it's doing it

[hellyeahent]

Can I get more than ~4,5 for r9 290 and ~4,2 for 280x   x11 algo ?

[Zawamiya]

Can I get more than ~4,5 for r9 290 and ~4,2 for 280x   x11 algo ?

Is it possible? Sure, my 280X is currently doing 4.83MH/s on X11.  

EDIT: 4.9MH/s, now.

care to share your config?

[badman74]

Can I get more than ~4,5 for r9 290 and ~4,2 for 280x   x11 algo ?

Is it possible? Sure, my 280X is currently doing 4.83MH/s on X11.  

EDIT: 4.9MH/s, now.

care to share your config?

try the mod version in my sig and be sure to put this in your config (5.2mh/s on 290, 6.2mh/s on 290x)

"xintensity" : "64",
"worksize": "64",
"hamsi-expand-big" : "7",
"hamsi-short" : true,
"blake-compact" : true,
"keccak-unroll" : "8",
"luffa-parallel" : true,
"gpu-threads" : "2",

[badman74]

Can I get more than ~4,5 for r9 290 and ~4,2 for 280x   x11 algo ?

Is it possible? Sure, my 280X is currently doing 4.83MH/s on X11.  

EDIT: 4.9MH/s, now.

care to share your config?

try the mod version in my sig and be sure to put this in your config

"xintensity" : "64",
"worksize": "64",
"hamsi-expand-big" : "7",
"hamsi-short" : true,
"blake-compact" : true,
"keccak-unroll" : "8",
"luffa-parallel" : true,
"gpu-threads" : "2",

I worked off of yours, I'm pretty sure my X11 is in the ballpark of 35.20% faster. X13 around 38.51%, and X15 about 42.87%.

with mine you should get 5.2mh/s on x11 with 290 (stock clocks) and 6.2mh/s on 290x (oc to 1040/1375)

[badman74]

Can I get more than ~4,5 for r9 290 and ~4,2 for 280x   x11 algo ?

Is it possible? Sure, my 280X is currently doing 4.83MH/s on X11.  

EDIT: 4.9MH/s, now.

care to share your config?

try the mod version in my sig and be sure to put this in your config

"xintensity" : "64",
"worksize": "64",
"hamsi-expand-big" : "7",
"hamsi-short" : true,
"blake-compact" : true,
"keccak-unroll" : "8",
"luffa-parallel" : true,
"gpu-threads" : "2",

I worked off of yours, I'm pretty sure my X11 is in the ballpark of 35.20% faster. X13 around 38.51%, and X15 about 42.87%.

with mine you should get 5.2mh/s on x11 with 290 (stock clocks) and 6.2mh/s on 290x (oc to 1040/1375)

What about 280X?

EDIT: I cloned the develop branch of your repo - is this a joke? 4.55MH/s on X11 on my 290X, tops. It's at 1000/1500.

EDIT2: Though I am liking the xI 64 - getting 6.14MH/s almost on my 290X, 1000/1500.

use the master branch... although it shouldnt be too different
i get 5.8 to 5.9mh/s on my 290x with 1040/1375 and I 16
and 6.18 to 6.2mh/s with xI 64
edit: my mods arent on the develop branch anymore as i started over with the 5.0.0 update

[poiuty]

Wolf0, wow!
how much hashrate on 290 with your kernel fix?

[bullus]

Can I get more than ~4,5 for r9 290 and ~4,2 for 280x   x11 algo ?

Is it possible? Sure, my 280X is currently doing 4.83MH/s on X11.  

EDIT: 4.9MH/s, now.

care to share your config?

Won't help - I modified the kernels for X11, X13, and X15.

Where can we find your kernels?

[qwep]

Can I get more than ~4,5 for r9 290 and ~4,2 for 280x   x11 algo ?

Is it possible? Sure, my 280X is currently doing 4.83MH/s on X11.  

EDIT: 4.9MH/s, now.

care to share your config?

try the mod version in my sig and be sure to put this in your config

"xintensity" : "64",
"worksize": "64",
"hamsi-expand-big" : "7",
"hamsi-short" : true,
"blake-compact" : true,
"keccak-unroll" : "8",
"luffa-parallel" : true,
"gpu-threads" : "2",

I worked off of yours, I'm pretty sure my X11 is in the ballpark of 35.20% faster. X13 around 38.51%, and X15 about 42.87%.

with mine you should get 5.2mh/s on x11 with 290 (stock clocks) and 6.2mh/s on 290x (oc to 1040/1375)

What about 280X?

EDIT: I cloned the develop branch of your repo - is this a joke? 4.55MH/s on X11 on my 290X, tops. It's at 1000/1500.

EDIT2: Though I am liking the xI 64 - getting 6.14MH/s almost on my 290X, 1000/1500.

use the master branch... although it shouldnt be too different
i get 5.8 to 5.9mh/s on my 290x with 1040/1375 and I 16
and 6.18 to 6.2mh/s with xI 64
edit: my mods arent on the develop branch anymore as i started over with the 5.0.0 update

All right, I'll give it a shot later. Getting 6.15 with mine on 290X now, at 1000/1500, so still not bad.

Checked out your darkcoin-mod.cl, and didn't see too much of a difference between it and sgminer's dev branch, but I'll try it later. In the meantime, I completely rewrote one hash - now getting 5.5MH/s out of my 280X

that publish it your miner or kernels

[qwep]

Wolf0, wow!
how much hashrate on 290 with your kernel fix?

Don't have a 290, but here, the GPUs are, in order from 0 - 3, a 270X, 290X, 7950, and a 280X. It's nsfw, as always: https://ottrbutt.com/tmp/x11localrigwolf.png

Can your miner

[poiuty]

Wolf0, wow!
how much hashrate on 290 with your kernel fix?

Don't have a 290, but here, the GPUs are, in order from 0 - 3, a 270X, 290X, 7950, and a 280X. It's nsfw, as always: https://ottrbutt.com/tmp/x11localrigwolf.png

What settings do you use for 270?

get 3Mh/s

"kernel": "darkcoin-mod",
"xintensity" : "102",
"gpu-threads" : "2",
"lookup-gap" : "2",
"worksize" : "128",
"gpu-engine" : "1250",
"gpu-memclock" : "1250",
"gpu-fan" : "35-85",

With xintensity -> 64 slightly less than 102

Please, share your kernel fix

[qwep]

Wolf0, wow!
how much hashrate on 290 with your kernel fix?

Don't have a 290, but here, the GPUs are, in order from 0 - 3, a 270X, 290X, 7950, and a 280X. It's nsfw, as always: https://ottrbutt.com/tmp/x11localrigwolf.png

What settings do you use for 270?

get 3Mh/s

"kernel": "darkcoin-mod",
"xintensity" : "102",
"gpu-threads" : "2",
"lookup-gap" : "2",
"worksize" : "128",
"gpu-engine" : "1250",
"gpu-memclock" : "1250",
"gpu-fan" : "35-85",

With xintensity -> 64 slightly less than 102

Please, share your kernel fix

+1

[qwep]

Wolf0, wow!
how much hashrate on 290 with your kernel fix?

Don't have a 290, but here, the GPUs are, in order from 0 - 3, a 270X, 290X, 7950, and a 280X. It's nsfw, as always: https://ottrbutt.com/tmp/x11localrigwolf.png

What settings do you use for 270?

get 3Mh/s

"kernel": "darkcoin-mod",
"xintensity" : "102",
"gpu-threads" : "2",
"lookup-gap" : "2",
"worksize" : "128",
"gpu-engine" : "1250",
"gpu-memclock" : "1250",
"gpu-fan" : "35-85",

With xintensity -> 64 slightly less than 102

Please, share your kernel fix

That engine clock is pretty high... I'm running 1165/1500.

Please, share your miner

[badman74]

280X above 5.75MH/s with new settings!

EDIT: Actually, 5.8MH+ sustained with a clock bump.

I really would like to see what you did the the darkcoin kernel to get those hash rates, maybe I could incorporate the speedups in the other kernels as well.
Unfortunately I don't know enough to build the kernels myself without at least something to dissect, even then it doesn't always work for me.

[Elun]

280X above 5.75MH/s with new settings!

EDIT: Actually, 5.8MH+ sustained with a clock bump.

I really would like to see what you did the the darkcoin kernel to get those hash rates, maybe I could incorporate the speedups in the other kernels as well.
Unfortunately I don't know enough to build the kernels myself without at least something to dissect, even then it doesn't always work for me.

I think He will not share his code to public soon. And He will be right. Because, what he will get from it? I think that will be only higher hashrates and diff in the networks.

[badman74]

280X above 5.75MH/s with new settings!

EDIT: Actually, 5.8MH+ sustained with a clock bump.

I really would like to see what you did the the darkcoin kernel to get those hash rates, maybe I could incorporate the speedups in the other kernels as well.
Unfortunately I don't know enough to build the kernels myself without at least something to dissect, even then it doesn't always work for me.

I think He will not share his code to public soon. And He will be right. Because, what he will get from it? I think that will be only higher hashrates and diff in the networks.

true but that doesn't stop me from wanting it....

[sp_]

It's easy to improve the speed of the current X11, x13 and x15 miners since other coders has published their optimized miners:

MaxDZ8
https://github.com/MaxDZ8/M8M/tree/master/M8M/kernels

cubehash
Echo
luffa
shavite
simd

pallas (groestl)

https://bitcointalk.org/index.php?topic=779598

5 out ot 11 x11 hashfunctions optimized.

on R9 280x I get the best stable results with

-w 64 -I 17 -g 4

gpu clock 1100 and memclock 1500

5800 is good wolf0, but it's not optimal.