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

[go6ooo1212]

..so did the speeds on x11/x13 brake the 7000/4800 kH/s

[1714924132]

1714924132

[1714924132]

1714924132

[1714924132]

1714924132

[1714924132]

1714924132

[1714924132]

1714924132

[djm34]

with WOLF0  .bin and  https://github.com/metalicjames/sgminer-Lyra2RE/releases ......same results

I don't have a bin for Lyra2RE.

yeah but now some people might be mining monero with your cpu miner or some others cpu coin by now  

By the way the latest version of lyra2RE (which should be slightly faster than the one in metalicjames, better groestl256 and one round blake256) can be found here:
https://github.com/djm34/sgminer

[djm34]

with WOLF0  .bin and  https://github.com/metalicjames/sgminer-Lyra2RE/releases ......same results

I don't have a bin for Lyra2RE.

yeah but now some people might be mining monero with your cpu miner or some others cpu coin by now 

By the way the latest version of lyra2RE (which should be slightly faster than the one in metalicjames, better groestl and one round blake256) can be found here:
https://github.com/djm34/sgminer

Groestl still slooooooow 

will do for now  (difficulty is already at the level before knc titan went in)

[litecoinbeast]

with WOLF0  .bin and  https://github.com/metalicjames/sgminer-Lyra2RE/releases ......same results

I don't have a bin for Lyra2RE.

yeah but now some people might be mining monero with your cpu miner or some others cpu coin by now  

By the way the latest version of lyra2RE (which should be slightly faster than the one in metalicjames, better groestl256 and one round blake256) can be found here:
https://github.com/djm34/sgminer

That is what I read abou...binaries somewhere? Also what settings would you recommend for 270x?

[splat44]

All is clear!

I will make few test later on! for meanwhile:

How do I know which threads, intensity and work size to use?

I might a bit understand work size but it might help knowing more on two first things
I couldn't find any information based on my GPU (leading to thread and intensity)

Ok I got that and you're right in your assumption!
You mention using 2 threads and intensity, based on following code, can you implement what you said for me to get started?

Code:

sgminer --algorithm quarkcoin --no-extranonce -I 18 -g 1 -w 256 -o stratum+tcp://some-quark-pool:3333 -u ... -p ...

I know which server to use but it other setting strings I need help with!
your support is greatly appreciated.

I have to be clear here, I don't know which setting to use yet

I tried using just the program as command prompt and I get the following warnings:
GPU_MAX_ALLOC_PERCENT is not specified
GPU_USE_SYNC_OBJECTS is not specified

Once I know those settings that overide those above then it might be easier
Is there way to show listing of option using program?

I do understand it possible mining quarkcoin using this but I am unsure the best setting for Diamond AMD Radeon R7 200 series.
I have latest catalyst and tried using without specifying amount of video ram following stream using sph-sgminer:
GPU_MAX_ALLOC_PERCENT

Too many rejected shares and I am doing better mining on a CPU which I am earning 50 QRK in two week than on gpu.

Since windows binaries is availasble and so I can test it, proper settings would be great.
Any feedback?

dont know what settings your using but i found it works best with 2 threads and intensity no higher than 14
also i never get any HW errors and the only rejects are stale shares

This might help in your .bat file

setx GPU_MAX_ALLOC_PERCENT 100
setx GPU_USE_SYNC_OBJECTS 1

that won't help his rejects/HW errors though
i assume he doesn't know how to set up sgminer at all
try this (from OP, with minor modification)

Code:

setx GPU_MAX_ALLOC_PERCENT 100
setx GPU_USE_SYNC_OBJECTS 1
sgminer --algorithm quarkcoin --no-extranonce -I 14 -g 2 -w 64 -o stratum+tcp://some-quark-pool:3333 -u ... -p ...

              ^
see above |
-I is intensity, -g is threads, -w is work size, -o is your pool address and port, -u is username, -p is password

[pallas]

10 minutes programming, 10% faster :-)

[pallas]

10 minutes programming, 10% faster :-)

I know, right? 

Yes, usual tricks. I was wondering if it can be made to fit the code cache.

[badman74]

All is clear!

I will make few test later on! for meanwhile:

How do I know which threads, intensity and work size to use?

I might a bit understand work size but it might help knowing more on two first things
I couldn't find any information based on my GPU (leading to thread and intensity)

trial and error as far as i know....

[pallas]

How many hardware errors? Me a lot...

[djm34]

10 minutes programming, 10% faster :-)

How many hardware errors? Me a lot...

10% faster with HW errors does really count as 10% faster 

[pallas]

10 minutes programming, 10% faster :-)

How many hardware errors? Me a lot...

10% faster with HW errors does really count as 10% faster  

I had them with stock kernel as well.
Also it's hanging pretty often.

[pallas]

10 minutes programming, 10% faster :-)

How many hardware errors? Me a lot...

10% faster with HW errors does really count as 10% faster 

Nevermind, was just intensity too high 😁

[platinum4]

I like how this thread changed from sharing tribal knowledge on configs, different ways to boost hashrate (keccak unroll, blake compact, etc.) to just bragging about hey look I made shit faster for myself and nobody else, so I'm going to earn a few more cents per day and let everybody else get fucked over on their ROI.

[pallas]

Who said I will not open source it? I already did it with my groestl.

[platinum4]

Sorry man, maybe I'm getting a bit frustrated; I've been here since Page 1 of this thread, meticulously bugged ystarnaud to implement a ton of fucking features and bugfixes and improvements, and have provided him with great feedback and also this entire community on the thread with multiple configs, test cases, driver version feedbacks and a ton of other shit.  Yes, I don't have a build environment so I can just tweak what I want, and badman74 and Elun and Elbandi have provided me with so many versions of sgminer5 by now it's hard to count.  I feel like any of the advice, configs, trial and error I have done and posted results here for, has only really shot me in the foot.  It appears in the mining game there's no reason ever to show anybody else how to get their shit set up right, because if there's isn't set up right and mine is, then that would mean I'm making more than they are.  I thought it was essential to giving back to the community.  I don't make the miners, the kernels, but sometimes I do find out what makes them faster, and I share it.  And now it's making me regret it, since we end up having to wait for table scraps ("leaked" kernels as opposed to shared) and having to deal with listening to fuckers brag about programming skills when they should just crowdfund it for a price to open source it or something.  I don't know.  As I said, frustrated.  You can see how this happened with neoscrypt.  It was profitable for a few days at best, and now I haven't seen it on westhash as most profitable in the last week.  It got ran into the ground, and we weren't even at its most optimized yet.  The field is way different than when scrypt was here - everybody shared there - it was a common struggle.

[pallas]

BTW I got another 10℅, programming on the phone.

[adaseb]

Well most of them all eventually get run into the ground because every starts mining with the kernel..

Same will happen with Lyra and it wont be profitable anymore.

[jch9678]

A couple weird things going on with my sgminer.  First, as the miner changes to a new algo, strange bins are being generated without changing anything in the confg file.  To test I deleted all the bins, started mining x13 with the correct bin, marucoin-modHawaiigw64l4ku0big7hs.  Nicehash switches algos to x11, but sgminer decides to make another x13 bin, marucoin-modHawaiigw64l4ku0 without big7hs.  I get  the clsetkernelarg of all params failed warning, but then it builds an x11 bin.  Here are all the bins getting produced:

darkcoin-modHawaiigw64l4ku0.bin
darkcoin-modHawaiigw64l4ku0big7hs.bin
darkcoin-modHawaiigw64l4tc8192.bin
marucoin-modHawaiigw64l4ku0.bin
marucoin-modHawaiigw64l4ku0big7hs.bin
marucoin-modHawaiigw64l4tc8192.bin
neoscryptHawaiigw64l4tc8192.bin

Some of those bins don't make sense.  How is it trying to make an x11 bin with thread concurrency?  Second problem but may be related is that sgminer just quits, it looks like it happens when the algo switches, but I'm not sure.  I get this from my output.log:

Code:

[14:07:01] Building binary neoscryptHawaiigw64l4ku0big7hs.bin
[14:07:01] Error -11: Building Program (clBuildProgram)
[14:07:01] "C:\Users\ANIMAL~1\AppData\Local\Temp\OCL4772T27.cl", line 368: warning: 
          variable "t" was declared but never referenced
  	uint4 t, st[4];
  	      ^

"C:\Users\ANIMAL~1\AppData\Local\Temp\OCL4772T27.cl", line 495: error: 
          identifier "MAX_GLOBAL_THREADS" is undefined
  	__global ulong16 *V = (__global ulong16 *)(padcache + (0x8000 * (get_global_id(0) % MAX_GLOBAL_THREADS)));
  	                                                                                    ^

"C:\Users\ANIMAL~1\AppData\Local\Temp\OCL4772T27.cl", line 513: warning: 
          argument of type "__global ulong16 *" is incompatible with parameter
          of type "__global uint16 *"
  		SMix(X, V, flag);
  		        ^

1 error detected in the compilation of "C:\Users\ANIMAL~1\AppData\Local\Temp\OCL4772T27.cl".

Frontend phase failed compilation.

Any ideas?  Could CGWatcher be interfering somehow when bins are made?

[MTJ151]

Sorry man, maybe I'm getting a bit frustrated; I've been here since Page 1 of this thread, meticulously bugged ystarnaud to implement a ton of fucking features and bugfixes and improvements, and have provided him with great feedback and also this entire community on the thread with multiple configs, test cases, driver version feedbacks and a ton of other shit.  Yes, I don't have a build environment so I can just tweak what I want, and badman74 and Elun and Elbandi have provided me with so many versions of sgminer5 by now it's hard to count.  I feel like any of the advice, configs, trial and error I have done and posted results here for, has only really shot me in the foot.  It appears in the mining game there's no reason ever to show anybody else how to get their shit set up right, because if there's isn't set up right and mine is, then that would mean I'm making more than they are.  I thought it was essential to giving back to the community.  I don't make the miners, the kernels, but sometimes I do find out what makes them faster, and I share it.  And now it's making me regret it, since we end up having to wait for table scraps ("leaked" kernels as opposed to shared) and having to deal with listening to fuckers brag about programming skills when they should just crowdfund it for a price to open source it or something.  I don't know.  As I said, frustrated.  You can see how this happened with neoscrypt.  It was profitable for a few days at best, and now I haven't seen it on westhash as most profitable in the last week.  It got ran into the ground, and we weren't even at its most optimized yet.  The field is way different than when scrypt was here - everybody shared there - it was a common struggle.

I shared Neoscrypt - that's WHY it got ran into the ground.

Shared it as in it's public now?

[badman74]

My sgminer versions are updated with Lyra2RE now for those that want to use it