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

[platinum4]

Yeah but the feasibility of performance won't be there.

I know for a fact that if I do NOT reboot in between switching from X11/X13 to scrypt-N I will see a predictable 20% performance in hashrate.

I doubt the effectiveness of switching between 4 different algorithms, 3 with different TCs and 2 with different worksizes, the intensity changing I like, but I don't the GPUs in practice will be able to handle this effectively.

I'm waiting in for reports before I go about thrashing my cards on some bullshit.

Not only that, but it puts you at the mercy of somebody's profit-switching algorithm or lookup procedure.  If they are edging you out by .001 BTC/MH/day for some shitty X11 over a scrypt-N, that needs to be taken into account.

As it is right now I appreciate Elun for dropping in a mixed build with added features, but have held off since I am waiting for the x13mod implementation.  Copying either x13mod kernel into this new build throws all nonce and HW errors, so I suppose I will need to wait.

[mrbrdo]

Why would you have decreased performance if switching algos? It's simply not the case and there is certainly no technical reason for that to happen.

The implementation itself has nothing to do with "somebody's" profit-switching algorithm. It just allows you to switch to a pool with a different algo without restarting the miner. Simply copying x13-mod of course will not work since it has 13 kernels and not 11.

[dhsc19]

Why would you have decreased performance if switching algos? It's simply not the case and there is certainly no technical reason for that to happen.

The implementation itself has nothing to do with "somebody's" profit-switching algorithm. It just allows you to switch to a pool with a different algo without restarting the miner. Simply copying x13-mod of course will not work since it has 13 kernels and not 11.

I've experienced things like that, too (with sgminer in general not just this particular build)  I sometimes cannot just restart the miner and get full hash rate.  Rebooting the machine restores it.  I can also predictably cut my hashrate in half by remotely accessing my Linux box through Teamviewer.  Even after killing Teamviewer, the hashrate never recovers until I reboot, and I have Teamviewer daemon running all the time.  There may be a memory leak that sgminer is unable to clear.

[kenshirothefist]

Binaries downloads are now available: https://nicehash.com/software/#sgminer

We expect to add x13mod optimized x13 kernel later this week.

[mrbrdo]

Just added X13-mod (named "marucoin-mod" for now).

dhsc19: it could just be related to TeamViewer. it would be easy enough to find out since you say you can always reproduce it - simply reboot, do not run sgminer at all, use TeamViewer and do whatever causes the problem, and then start sgminer. if you still get a lower hashrate then sgminer can't do anything about it.

Old sgminers and cgminer 3.7.2 did indeed have leaks on the GPU, but I fixed them a few months ago when I did the "hot-switching" between Scrypt and Scrypt-N. So any semi-recent build of sgminer should not have leaks on the GPU, it is probably something else.

[phzi]

Old sgminers and cgminer 3.7.2 did indeed have leaks on the GPU, but I fixed them a few months ago when I did the "hot-switching" between Scrypt and Scrypt-N. So any semi-recent build of sgminer should not have leaks on the GPU, it is probably something else.

Indeed.  I've had similar problems many times with sph-sgminer, but never with a recent sgminer.  After running sph-sgminer for X11, hashrate would always drop until reboot.

It seems the algorithm vs kernel changes fixed this for sgminer.  Althought I can't say I have tested the v5 branch extensively yet, I haven't seen this issue crop up (and it happened every time with sph-sgminer).

[infofront]

So I had to manually add darkcoin-mod.cl to the directory where my sgminer script is located.
Now I'm getting, "Error: clSetKernelArg of all params failed."

[platinum4]

Just added X13-mod (named "marucoin-mod" for now).

dhsc19: it could just be related to TeamViewer. it would be easy enough to find out since you say you can always reproduce it - simply reboot, do not run sgminer at all, use TeamViewer and do whatever causes the problem, and then start sgminer. if you still get a lower hashrate then sgminer can't do anything about it.

Old sgminers and cgminer 3.7.2 did indeed have leaks on the GPU, but I fixed them a few months ago when I did the "hot-switching" between Scrypt and Scrypt-N. So any semi-recent build of sgminer should not have leaks on the GPU, it is probably something else.

Tried to drop in marucoin-mod to the last Windows binaries, nothing but HW failures.

Can we get the X13 wrapped up in a new set of binaries please?

[phzi]

Tried to drop in marucoin-mod to the last Windows binaries, nothing but HW failures.

Can we get the X13 wrapped up in a new set of binaries please?

Go compiled it yourself, or try being patient for a whole day/two...

The marucoin-mod edition was just released (~6 hours ago) - if you want bleeding edge, learn to compile from source.  Else, wait like everyone else.

[platinum4]

Cool, so I'll wait for you to compile it for Windows then, I appreciate it!

[Elun]

Cool, so I'll wait for you to compile it for Windows then, I appreciate it!

Windows binaries with marucoin-mod. https://www.dropbox.com/s/z59qwtcc0kb1l2q/sgminer_v5_0.zip
hd5xxx/6xxx still have problems, but lasybear is working on it, will be fixed soon.

[platinum4]

Cool, so I'll wait for you to compile it for Windows then, I appreciate it!

Windows binaries with marucoin-mod. https://www.dropbox.com/s/z59qwtcc0kb1l2q/sgminer_v5_0.zip
hd5xxx/6xxx still have problems, but lasybear is working on it, will be fixed soon.

Yessir, Elun with the fucking save.

brb gonna go be BLEEDING EDGE for a little fucking bit y'all.

Edit: That's a bingo.  x13 is up and going with this miner, will test for 24hours and report back in with any SICK warnings for Hawaii chipsets.

So far so good it appears, not too sure what the difference is between 4.2.1 and 4.1.0 other than the pool settings now.

Looks to be a pretty smooth patching process of migrating the sph-sgminer over to the unified.

Keep up the good work, I'll be waiting in for test results from others!!

Edit 2:  This build seems to be missing the uptime counter [- this is vital for keeping threads alive], and I would assume is missing the ADL GPU idle patch; it does however contain the show-coindiff patch.

DON'T FORGET THAT UPTIME COUNTER Y'ALL.

[phateminer]

Hey all

does anyone have any ideas why i am getting fatal JSON errors here?

{
"pools" : [
        {
      "url" : "stratum+tcp://#server:port#",
      "user" : "#user.worker#",
      "pass" : "#password#",
      "name" : "x13",
      "algorithm" : "marucoin-mod"
                "gpu-engine" : "1145,1114",
      "gpu-memclock" : "1500,1500"
                "thread-concurrency" : "32760",
                "intensity" : "19,20",
                "gpu-threads" : "1"
      "worksize" : "128"
      "no-restart" : "true"
        },
   {
      "url" : "stratum+tcp://#server:port#",
      "user" : "#user.worker#",
      "pass" : "#password#",
      "name" : "x11",
      "algorithm" : "darkcoin-mod"
                "gpu-engine" : "1140,1110",
      "gpu-memclock" : "1500,1500"
                "thread-concurrency" : "32760",
                "intensity" : "20,20",
                "gpu-threads" : "1"
      "worksize" : "128"
      "no-restart" : "true"
        },
   {
      "url" : "stratum+tcp://#server:port#",
      "user" : "#user.worker#",
      "pass" : "#password#",
      "name" : "scrypt",
      "algorithm" : "zuikis"
                "gpu-engine" : "1040,1025",
      "gpu-memclock" : "1500,1500"
                "thread-concurrency" : "32760",
                "intensity" : "21,21",
                "gpu-threads" : "1"
      "worksize" : "256"
      "no-restart" : "true"
        },
]
}

thanks in advance

phate

[Elun]

Hey all

does anyone have any ideas why i am getting fatal JSON errors here?

{
"pools" : [
        {
      "url" : "stratum+tcp://#server:port#",
      "user" : "#user.worker#",
      "pass" : "#password#",
      "name" : "x13",
      "algorithm" : "marucoin-mod"
                "gpu-engine" : "1145,1114",
      "gpu-memclock" : "1500,1500"
                "thread-concurrency" : "32760",
                "intensity" : "19,20",
                "gpu-threads" : "1"
      "worksize" : "128"
      "no-restart" : "true"
        },
   {
      "url" : "stratum+tcp://#server:port#",
      "user" : "#user.worker#",
      "pass" : "#password#",
      "name" : "x11",
      "algorithm" : "darkcoin-mod"
                "gpu-engine" : "1140,1110",
      "gpu-memclock" : "1500,1500"
                "thread-concurrency" : "32760",
                "intensity" : "20,20",
                "gpu-threads" : "1"
      "worksize" : "128"
      "no-restart" : "true"
        },
   {
      "url" : "stratum+tcp://#server:port#",
      "user" : "#user.worker#",
      "pass" : "#password#",
      "name" : "scrypt",
      "algorithm" : "zuikis"
                "gpu-engine" : "1040,1025",
      "gpu-memclock" : "1500,1500"
                "thread-concurrency" : "32760",
                "intensity" : "21,21",
                "gpu-threads" : "1"
      "worksize" : "256"
      "no-restart" : "true"
        },
]
}

thanks in advance

phate

You missed commas.
Service for JSON validate: http://jsonlint.com/

[phateminer]

Hey all

does anyone have any ideas why i am getting fatal JSON errors here?

{
"pools" : [
        {
      "url" : "stratum+tcp://#server:port#",
      "user" : "#user.worker#",
      "pass" : "#password#",
      "name" : "x13",
      "algorithm" : "marucoin-mod"
                "gpu-engine" : "1145,1114",
      "gpu-memclock" : "1500,1500"
                "thread-concurrency" : "32760",
                "intensity" : "19,20",
                "gpu-threads" : "1"
      "worksize" : "128"
      "no-restart" : "true"
        },
   {
      "url" : "stratum+tcp://#server:port#",
      "user" : "#user.worker#",
      "pass" : "#password#",
      "name" : "x11",
      "algorithm" : "darkcoin-mod"
                "gpu-engine" : "1140,1110",
      "gpu-memclock" : "1500,1500"
                "thread-concurrency" : "32760",
                "intensity" : "20,20",
                "gpu-threads" : "1"
      "worksize" : "128"
      "no-restart" : "true"
        },
   {
      "url" : "stratum+tcp://#server:port#",
      "user" : "#user.worker#",
      "pass" : "#password#",
      "name" : "scrypt",
      "algorithm" : "zuikis"
                "gpu-engine" : "1040,1025",
      "gpu-memclock" : "1500,1500"
                "thread-concurrency" : "32760",
                "intensity" : "21,21",
                "gpu-threads" : "1"
      "worksize" : "256"
      "no-restart" : "true"
        },
]
}

thanks in advance

phate

You missed commas.
Service for JSON validate: http://jsonlint.com/

absolute legend. thanks Elun.

[phateminer]

this now switches and uses the correct algorithm, but does not change any of the gpu settings such as intensity or engine. any ideas?


EDIT: it will change the intensity and gpu engine if you use the "pool-gpu-engine" and "pool-intensity" paramters instead of "gpu-engine" and "intesnity". does not set the intensity correctly when opening (sets it to unless you switch to another pool and then switch back. setmemclock error listed, but i have my memory set to the same speed for all anyway so i shouldn't need that. after a couple of minutes the miner shuts down and displays a summary of works.

[12:38:26]
Summary of runtime statistics:

[12:38:26] Started at [2014-06-03 12:36:36]
[12:38:26] Runtime: 0 hrs : 0 mins : 48 secs
[12:38:26] Average hashrate: 6.4 Megahash/s
[12:38:26] Solved blocks: 0
[12:38:26] Best share difficulty: 0.057
[12:38:26] Share submissions: 28
[12:38:26] Accepted shares: 28
[12:38:26] Rejected shares: 0
[12:38:26] Accepted difficulty shares: 0
[12:38:26] Rejected difficulty shares: 0
[12:38:26] Reject ratio: 0.0%
[12:38:26] Hardware errors: 0
[12:38:26] Utility (accepted shares / min): 36.69/min
[12:38:26] Work Utility (diff1 shares solved / min): 0.11/min

[12:38:26] Stale submissions discarded due to new blocks: 0
[12:38:26] Unable to get work from server occasions: 0
[12:38:26] Work items generated locally: 50
[12:38:26] Submitting work remotely delay occasions: 0
[12:38:26] New blocks detected on network: 5

Phate

[platinum4]

Again, questioning the effectiveness or usefulness in switching between [scrypt, scrypt-N] and [x11,x13].

Still skeptical as to whether you will see peak performance without having to experience a reboot.

Also, are there any pools that have a muliport stratum, for multiports?  i.e. the pool analyzes which algo+coin is the top breadwinner and serves the work through different ports?

We're missing something here.

[kenshirothefist]

NiceHash brought you another innovation: NiceHash.com Best Profit Auto-Switching Multi-Algorithm GPU Mining. You can get more information here: https://www.nicehash.com/multialgo/

[bullus]

It look nice but I'm gonna wait until the powerconsumption is also in the calculation..x11 is about 0.5 less power than scrypt. So the profit should be minimal twice as much between scrypt and x11.

Is it possible to only switch between x11 and x13 ? And if the algo choose a scrypt coin u go to your own failover?

I also have to wait untill my hd6970 is working with the miner (x13) I have a rig with 7950 and 6970 in it.

[mrbrdo]

Am I the only one that cannot mine on the X11/X13 ports anymore since today?