[ANN][X11] X11/Darkcoin Gpu Miner - updated to use leaked binaries - even faster

[phzi]

I think it's time for girino to release his source code as promised.

done. (added zip with code to the shared dropbox folder).

Thank-you for respecting the open source community and licenses by releasing the source code for your modifications.

[Sglasio]

sgminer.exe -o stratum+tcp://stratum-eu.suchpool.pw:3325 -u Sglasio.1 -p x --thread-concurrency 8193 -g 4 -I 18 -w 128 --gpu-memclock 1500 --gpu-powertune 20 --gpu-engine 1150 --gpu-fan 50-100 --shaders 2048 --lookup-gap 2
Intensity can 19 or 18 for youre choice
Get 3.3 Mhs
Temp is ok

[Amph]

I think it's time for girino to release his source code as promised.

done. (added zip with code to the shared dropbox folder).

I'm more interested in an explanation  why it's so much faster!
Your crack and Lasybear's work are "sufficient" enough to copy what he has done, but CL-code itself is 100% identical and doesn't explain the speed boost.....  Only thing different is the way CL-functions are called, and the use of an additional memory-block (plus pointers instead of arrays).
Still like to learn what Girino's breakthrough was and his line of thought archiving this (this knowledge, our community can use for further improvements).

There are many things that make it faster.

GPU's have very little "private" memory, and all declared local variables are allocated in "private" memory space, that quickly fills up and is then cached into global (yes, global) memory. So big functions with lots of variables are REALLY BAD for GPU. Breaking up the functions the way i did can give you huge speed improvements. Sequencial accesses to global memory are quite fast as they are cached in local memory by the compiler, so the use of global memory to store the intermediate hashes is still fast enough for the improvements to be worth it.

in your experience it is possible to optimize even more? maybe a 20% more hashrate?

[Elenelen]

I think it's time for girino to release his source code as promised.

done. (added zip with code to the shared dropbox folder).

I'm more interested in an explanation  why it's so much faster!
Your crack and Lasybear's work are "sufficient" enough to copy what he has done, but CL-code itself is 100% identical and doesn't explain the speed boost.....  Only thing different is the way CL-functions are called, and the use of an additional memory-block (plus pointers instead of arrays).
Still like to learn what Girino's breakthrough was and his line of thought archiving this (this knowledge, our community can use for further improvements).

There are many things that make it faster.

GPU's have very little "private" memory, and all declared local variables are allocated in "private" memory space, that quickly fills up and is then cached into global (yes, global) memory. So big functions with lots of variables are REALLY BAD for GPU. Breaking up the functions the way i did can give you huge speed improvements. Sequencial accesses to global memory are quite fast as they are cached in local memory by the compiler, so the use of global memory to store the intermediate hashes is still fast enough for the improvements to be worth it.

Thank you for the zip and the information, Girino !
I will try to optimize it some more with your hints about Global-mem.

[Amph]

it's possible to optimize the jackpoint algo for amd? there is too much difference between nvidia and amd there(2MH/ vs 4MH/s)

In the past this was because AMD was much more efficient with integer operations whereas Nvidia excelled at Floating point stuff.

so nothing can be done? sad...

[ozzy1926]

will there be an update for other coins like boostcoin?

[Undland]

i have only 400-900Kh/s with 7790 and 260X on BAMT 1.6.2 (same on SMOS 1.2). On windows 8.1 i had 1200Kh/s. Why so different?  

"intensity" : "17", - up or dawn is reduces hashrate
"kernel" : "x11mod", - same darkcoin no effect
"lookup-gap" : "2", - no effect
"gpu-threads" : "1", - if 2, then rise hashrate to 1100Kh/s for 7790 (260X no effect), but some GPU is turning off
"gpu-engine" : "1100", - if gpu-thread=1, then no effect
"gpu-memclock" : "1500"
http://prntscr.com/3nbyot
 

P.S. What difference between x11-sgminer-bamt-build2.tgz and x11-sgminer-linux-32bit-build1.tgz?

[Amph]

will there be an update for other coins like boostcoin?

remember the 5 ecc(sry your nick is too long to be remembered lol), is tweaking the 11 algo of the x13

[ozzy1926]

will there be an update for other coins like boostcoin?

remember the 5 ecc(sry your nick is too long to be remembered lol), is tweaking the 11 algo of the x13

great news!

[kahir]

i get 750 k/hash on normal sgminer ....for 260x


but when i use this SGminer


1 of them go to  1 M/hash ... and thats great

but the another one stay at  150 k/hash


anyhelp ?

[kahir]

i have only 400-900Kh/s with 7790 and 260X on BAMT 1.6.2 (same on SMOS 1.2). On windows 8.1 i had 1200Kh/s. Why so different?  

"intensity" : "17", - up or dawn is reduces hashrate
"kernel" : "x11mod", - same darkcoin no effect
"lookup-gap" : "2", - no effect
"gpu-threads" : "1", - if 2, then rise hashrate to 1100Kh/s for 7790 (260X no effect), but some GPU is turning off
"gpu-engine" : "1100", - if gpu-thread=1, then no effect
"gpu-memclock" : "1500"
http://prntscr.com/3nbyot
 

P.S. What difference between x11-sgminer-bamt-build2.tgz and x11-sgminer-linux-32bit-build1.tgz?

please i need help i got same config as you ....  260x and 7790 ....


when i open sgminer .... 1 of them is really slow and another is fast

[wacko]

i get 750 k/hash on normal sgminer ....for 260x


but when i use this SGminer


1 of them go to  1 M/hash ... and thats great

but the another one stay at  150 k/hash


anyhelp ?

Whenever you see unrealistically low hashrate from some cards while others work at full speed, there's a good chance that your cards (the "slow" ones) aren't stable. Lower your overclock, increase the voltage etc, delete all the .bin files from the miner's directory and try again.

[girino]

in your experience it is possible to optimize even more? maybe a 20% more hashrate?

Possibly yes, but now it involves other tricks, like using vector types (int2, int4 and int8) and to use amd specific functions (bitalign, etc). It will be a really slow work and possibly take some months to get to 20% improvement.

and still, the slower algorithms are the ones using lookup tables (groestl, shavite, echo), which are hard to improve due to the size of the tables (groestl uses 8 tables of 1Kb each). So i wouldn't expect huge gains in the next versions, but slow and steady improvement.

[kahir]

girino  


i get more hash ...thx you for dev


but



i keep getting


stratum from pool 0 requested work restart



and my hash rate at pool is lower than normal  SGminer lol

[Undland]

please i need help i got same config as you ....  260x and 7790 ....

when i open sgminer .... 1 of them is really slow and another is fast

I don't know, man.   try gpu_threads=2, maybe all of your cards will work fine

1*7950=3*7790  that's not really funny

[kahir]

i keep getting this


stratum from pool 0 requested work restart
stratum from pool 0 requested work restart
stratum from pool 0 requested work restart
stratum from pool 0 requested work restart
stratum from pool 0 requested work restart
stratum from pool 0 requested work restart
stratum from pool 0 requested work restart



i rarely get any accepted .... tried diff pools ... my hashrate show really low on any pool

should i do something

[atp1916]

i keep getting this


stratum from pool 0 requested work restart
stratum from pool 0 requested work restart
stratum from pool 0 requested work restart
stratum from pool 0 requested work restart
stratum from pool 0 requested work restart
stratum from pool 0 requested work restart
stratum from pool 0 requested work restart



i rarely get any accepted .... tried diff pools ... my hashrate show really low on any pool

should i do something

That appears to be the dev pool messing things up.

Put just your pool in the batch file, and set your miner to fail-over in the pool commands and/or use CGWatcher poolguard to ensure your miner stays on the pools of your choice.

[uzitgc]

any help runing this with defult bamt commands?(intergrate into bamt)
mine start etc....

i installed it but have to start manully and i want to run it but defult mine commands any help?

yes this would be HUGELY appreciated.
how do you even run it manually??? I have installed it using the instructions, just don't know how to run it! 

[kahir]

i keep getting this


stratum from pool 0 requested work restart
stratum from pool 0 requested work restart
stratum from pool 0 requested work restart
stratum from pool 0 requested work restart
stratum from pool 0 requested work restart
stratum from pool 0 requested work restart
stratum from pool 0 requested work restart



i rarely get any accepted .... tried diff pools ... my hashrate show really low on any pool

should i do something

That appears to be the dev pool messing things up.

Put just your pool in the batch file, and set your miner to fail-over in the pool commands and/or use CGWatcher poolguard to ensure your miner stays on the pools of your choice.

im kinda new to mining .... how i do that

[chadwickx16]

i keep getting this


stratum from pool 0 requested work restart
stratum from pool 0 requested work restart
stratum from pool 0 requested work restart
stratum from pool 0 requested work restart
stratum from pool 0 requested work restart
stratum from pool 0 requested work restart
stratum from pool 0 requested work restart



i rarely get any accepted .... tried diff pools ... my hashrate show really low on any pool

should i do something

That appears to be the dev pool messing things up.

Put just your pool in the batch file, and set your miner to fail-over in the pool commands and/or use CGWatcher poolguard to ensure your miner stays on the pools of your choice.

im kinda new to mining .... how i do that

Inside the mining window type P>C>0

P=Pool Management
C=Change Strategy
0=Failover (It's not displayed but it will still work)