Topic: [ANN] sgminer v5 - optimized X11/X13/NeoScrypt/Lyra2RE/etc. kernel-switch miner - page 211. (Read 877859 times)

That build doesn't allow for --kernel i don't think, builds after 6/25 do, but some are broken in the v5_0 tree for the next day or two

I'm having a problem and I'm out of ideas.

Using sgminer-5.0-pre-release-2014-06-25-win32 only 1 GPU will mine and the others will be listed as "OFF". Also, it forces marucoin-mod even when x11mod is set as the algo in the config. (Using x11mod because I can't get marucoin-mod to work).
The same config works with sph-sgminer_x11mod (not including kernel vs algo).

I'm running 4x270x's

I'm going to switch to Nicehash after this is working.


{
"pools" : [
   {
      "url" : "stratum+tcp://east01.us.trademybit.com:4440",
      "no-extranonce" : true,
      "user" : "x",
      "pass" : "x",
      "algorithm" : "x11mod"
   },
   {
      "url" : "stratum+tcp://east02.us.trademybit.com:4440",
      "no-extranonce" : true,
      "user" : "x",
      "pass" : "x",
      "algorithm" : "x11mod"
   }
]
,
"api-listen" : true,
"expiry" : "28",
"failover-only" : true,
"no-submit-stale" : true,
"queue" : "1",
"scan-time" : "5",
"device" : "0,1,2,3",
"gpu-threads" : "2",
"gpu-engine" : "1070",
"gpu-fan" : "0-100",
"gpu-memclock" : "1470",
"gpu-powertune" : "20",
"gpu-dyninterval" : "7",
"temp-target" : "70",
"temp-overheat" : "85",
"temp-cutoff" : "95",
"temp-hysteresis" : "3",
"worksize" : "128",
"lookup-gap" : "2",
"vectors" : "1",
"thread-concurrency" : "8192",
"xintensity" : "300",
"tcp-keepalive" : "30",
"api-allow" : "W:127.0.0.1"
}

http://imgur.com/rmpcuwf

I got cgwatcher working with the newest version.

Get CGWatcher 1.3.9

In the settings there is something you need to disable...

Settings>Miner>
CHECK - Disable temporary config file creation during miner start

Now, you are not done. Launch SGMiner through CGWatcher and it should instaclose.

Go to your config file and you should see a 2 new lines in there:

"api-listen" : true,

"api-allow" : "W:127.0.0.1"

For whatever reason it will add the lines needed to monitor, but it wont add the comma that will cause a JSON error and make the miner crash. For me, it was the "api-allow" : "W:127.0.0.1" it would continuously delete the comma, so the fix is to add that line to the VERY bottom:

EXAMPLE

"api-listen" : true,
"intensity" : "19",
"worksize" : "256",
"lookup-gap" : "2",
"thread-concurrency" : "8193",
"gpu-threads" : "2",
"gpu-fan" : "40,60,60,40,40",
"gpu-memclock" : "1500",
"gpu-powertune" : "20",
"expiry" : "10",
"gpu-dyninterval" : "7",
"log" : "5",
"no-pool-disable" : true,
"tcp-keepalive" : "30",
"temp-hysteresis" : "3",
"failover-only" : true,
"failover-switch-delay" : "1",
"no-submit-stale" : true,
"kernel-path" : "/usr/local/bin",
"api-allow" : "W:127.0.0.1"
}

Problem solved. You can now use CGWatcher when using multi algo switching.

I sincerely hope this helps someone.

Thanks.

New sgminer v5 is great, thanks devs very much for great effort!
But if you're looking for any reason for old sgminer builds you can get them here https://sites.google.com/site/coinstodam00n/builds

You are correct, I had no idea nvidia cards were mining X11 so efficiently.   In that case, keep up the progress!

Cgwatcher is no longer working with this version.Any backup plan?It's hot in summer and sgminer like to hang.

Is it possible to use more than one algo in the same time ?

like this:

],
"failover-only" : true,
"api-allow" : "W:127.0.0.1",
"api-listen" : true,
"intensity" : "19,19,19,19,19",
"worksize" : "128,128,128,128,128",
"thread-concurrency" : "8192,8192,8192,8192,8192",
"gpu-threads" : "4,4,4,4,2",
"kernel": "x13mod,x13mod,x13mod,x13mod,x13modold"


(280x,280x,280x,280x,HD5870)

Not sure if that's true, but I've counted now five different versions of x15 miners floating around.

I'd be very carefull with the X15 AMD miner floating around,
it could be trojan/virus. 

does this ave nvida support or just amd?

Optimized or not, hashrates are better right now for most AMD miners that have been following along with the x11mod/x13mod updates.  Anybody who has complaints should go back to sgminer 4.1.0 or sph-sgminer without modded kernels, and appreciate exactly how far the fuck we have come in eight weeks.

Nope, don't agree.  It is not rational to think that a 290 @ 2.5mhs is "optimized" when a 750 Ti is pulling in the same hash rate vicinity but spending only 30w doing so.

With clocks exactly the same / drivers exactly the same, the optimized lasybear / sgminer v5 builds fall within 5-10% power usage (more) of the most current phm build on all my 290, 270, 280x, and 7950 rigs. Temps are virtually the same (including VRMs).  I do not use 5/6xxx cards, so it may be different with those architectures.

My take is that X11 has simply been under-optimized for AMD hardware since phm stopped developing a while back.  These current enhancements are bringing the performance up to where it should be.  

  I agree.Hooray!Hooray!Hooray!

Following up from my last post, my other rig:

3 x 290 MSI Gaming 4G
These GPU's have their quirks.  I've found that removing the setx commands from my bat file and removing the environment variables for X11/X13 have improved its performance.  This rig doesnt handle switching well at all with any algorithm.  It's very hit or miss, even when switching X11>X13 and back.  Lowering the thread-concurrency for Scrypt/Scrypt-N resolved the error enqueing -4 error

X11 - darkcoin-mod  3.75MH/s sustained solid
X13 - marucoin-mod 2.85MH/s sustained solid

    intensity - 19
    worksize - 512
    gpu-engine - 1060
    gpu-memclock - 1250
    gpu-powertune - 10
    gpu-threads - 2
    hamsi-expand-big - 1

I'm not defining a thread concurrency for X11/X13.  Is it even necessary? I do define shaders - 2560 though.

Scrypt - zukkis - 850KH/s sustained solid
Scrypt-N - zukkis - 430KH/s sustained solid

    intensity - 19
    worksize - 512
    gpu-engine - 1000
    gpu-memclock - 1250
    gpu-powertune - 10
    gpu-threads - 1
    thread-concurrency - 20480

Keccak - maxcoin - 430MH/s sustained -- 10% rejects

    intensity - 14
    worksize - 256
    gpu-engine - 1080
    gpu-memclock - 1250
    gpu-powertune - 10
    gpu-threads - 1

no thread-concurrency defined.

Very nice miner, I may have to setup my GPU again just so I can test this out.
Thanks for all the work OP!

Man, where do i begin.  I've tweaked and changed so many things just to see how stable I can get things, here is where I am at:

4 x 280x - MSI Gaming 4G
This thing runs solid.  X11 & X13 switching works great. Scrypt to Scrypt-N works great. Keccak works great but rejects are a little high.. ~10%-ish.
I dont have ALL of the algos on one conf file and I am stopping and starting the mining manually as I see profitability changing... but I do miss changes too.

X11 - darkcoin-mod  3.25MH/s sustained solid
X13 - marucoin-mod 2.55MH/s sustained solid

intensity - 17
worksize - 256
gpu-engine - 1125
gpu-memclock - 1500
gpu-powertune - -20
gpu-threads - 2
hamsi-expand-big - 4

I'm not defining a thread concurrency for X11/X13.  Is it even necessary? I do define shaders - 2048 though.

Scrypt - zukkis - 730KH/s sustained solid
Scrypt-N - zukkis - 355KH/s sustained solid

xintensity - 4
worksize - 256
gpu-engine - 1030
gpu-memclock - 1500
gpu-powertune - -20
gpu-threads - 2
thread-concurrency - 8193

Keccak - maxcoin - 350MH/s sustained -- 10% rejects

intensity - 14
worksize - 256
gpu-engine - 1125
gpu-memclock - 1020
gpu-powertune - -20
gpu-threads - 2

no thread-concurrency defined.

For sure.  But with the increased energy costs due to "progress" we're that much closer to no profits.  Oh well, can't fight it I guess

Delete bins?  What CCC version you on 13 or 14?

How do we get this guy's pull request to go through?

Still better than no profits, which is current for mining BTC/LTC/FTC/Quark, etc.