[ANN] ccminer 2.3 - opensource - GPL (tpruvot)

[pallas]

From my first tests with cuda 8 rc, sm61 doesnt improve anything... over 5.2, maybe this rc is not yet optimized for the GTX, but only for the Tesla 6.0

I tested cuda 8 RC as well:

- performance apparently the same as 7.5
- on lyra2re, when building for compute 6.1, 1070 was slower; same speed as 7.5 when the card runs compute 5.2 code made with cuda 8
- faster compiling cuda code
- less register spill so possible improvement in some algos

[1715617170]

1715617170

[1715617170]

1715617170

[1715617170]

1715617170

[1715617170]

1715617170

[Epsylon3]

tx, yep same results

[rednoW]

Sorry for noob question. How to enable gpu temp showing in miner window? Temp readings are ok - i tested with --max-temp=N option and it works as expected. But I want to see temp like in sgminer or claymore miner. Thanks!

[Epsylon3]

i dont display it in the miner output... the terminal doesnt handle the "°" symbol on windows

[tbearhere]

from --help :
      --mem-clock=3505  Set the gpu memory boost clock
      --gpu-clock=1178  Set the gpu engine boost clock       << 970gtx normal clock speed my model.
      --plimit=100      Set the gpu power limit in percentage
      --tlimit=80       Set the gpu thermal limit in degrees

Nice Epsylon3  now if only you have fan speed.. checking the help files now.

What is the command to send the help file to my printer plz.  thx
And or make a notepad file that can be saved and read or printed.

ccminer1.8.exe --help
pause:

[rednoW]

What is the command to send the help file to my printer plz.  thx
And or make a notepad file that can be saved and read or printed.

ccminer1.8.exe --help
pause:

ccminer1.8.exe --help > ready_to_print.txt

))

[bensam1231]

Wait wait

Keep Calm   .

Now...

the zotac Wotan then Wipeout are more powerful than all gtx1070 ?

They do 400mhs against 270mhs of gigabyte g1 gaming.

And all this thanks to the double feed slot? and a bios?

It would be nice if everyone more calm, and try to understand .

Something's very fishy about that speed. Is there any confirmed proofs of such differences

Yeah, going to call BS on this too. There is no way to gain +50% performance from a OC without literally OCing an additional 50%. The cards would be running at around 3300Mhz. The only other options are it's a personal kernel (someone coded for him) or it's erroneous.

Lbry doesn't depend super heavily on memory either, so it can't be some weird timing tweaks like with Eth.


Just as a FYI, while Amph is a troll, I have noticed Chinese love gloating about non-existent things. It's like part of their culture to lie and brag about things that don't exist from politics, to science, to technology, to just random shit in daily life all to try and raise their prominence. Take anything they say with a huge grain of salt.

That being said there is some ingenuity that comes out of there and it's entirely possible this guy is really talented, just highly unlikely.

then how come he is the only one doing that hash? you have a zotac yourself but it is not doing 400MH like you said, only 270MH, which is ok it's possible with oc alone, but not 400

haha Amph i thing it was a loong night for u ?
cant sleep?

He s not a Dev, not a coder, a hacker, he even dont know anything about mining 4 month ago.
Just a guy who knows how to oc a nvidia card. Thats it.

But u will still not getit right?
That someone got better Gpus then u!
It is blowing up ur mind that some one is better then u ,
that someone can do someting better then u!
that someone knows more that u!
and again... it is justa oc nothing more not any magic, hoodoo vodoo!

Screen shots or this is BS. Even then it's really easy to fake screenshots.

Link to the original post Amph.

[jonesskill]

Guys, i have one question:

Now im mining lyra2rev2, in my miner promt shows 46 Mh/s, its OK.
But in the Pool the Mh/s float from 30 to 60-80, never stabilizes.

Its normal? or i miss something?

In my config file i only put the addres, my user and pass and --cpu-intensivity 4

[Epsylon3]

jonesskill: yes its normal


bensam: ive had another idea, as i said its not possible to use a kernel on a card and not the other... But I think it could be someone testing the Zotac extreme bios on a bigger card...

[jonesskill]

jonesskill: yes its normal


bensam: ive had another idea, as i said its not possible to use a kernel on a card and not the other... But I think it could be someone testing the Zotac extreme bios on a bigger card...

Thank you, Epsylon3!

[Amph]

Wait wait

Keep Calm   .

Now...

the zotac Wotan then Wipeout are more powerful than all gtx1070 ?

They do 400mhs against 270mhs of gigabyte g1 gaming.

And all this thanks to the double feed slot? and a bios?

It would be nice if everyone more calm, and try to understand .

Something's very fishy about that speed. Is there any confirmed proofs of such differences

Yeah, going to call BS on this too. There is no way to gain +50% performance from a OC without literally OCing an additional 50%. The cards would be running at around 3300Mhz. The only other options are it's a personal kernel (someone coded for him) or it's erroneous.

Lbry doesn't depend super heavily on memory either, so it can't be some weird timing tweaks like with Eth.


Just as a FYI, while Amph is a troll, I have noticed Chinese love gloating about non-existent things. It's like part of their culture to lie and brag about things that don't exist from politics, to science, to technology, to just random shit in daily life all to try and raise their prominence. Take anything they say with a huge grain of salt.

That being said there is some ingenuity that comes out of there and it's entirely possible this guy is really talented, just highly unlikely.

then how come he is the only one doing that hash? you have a zotac yourself but it is not doing 400MH like you said, only 270MH, which is ok it's possible with oc alone, but not 400

haha Amph i thing it was a loong night for u ?
cant sleep?

He s not a Dev, not a coder, a hacker, he even dont know anything about mining 4 month ago.
Just a guy who knows how to oc a nvidia card. Thats it.

But u will still not getit right?
That someone got better Gpus then u!
It is blowing up ur mind that some one is better then u ,
that someone can do someting better then u!
that someone knows more that u!
and again... it is justa oc nothing more not any magic, hoodoo vodoo!

Screen shots or this is BS. Even then it's really easy to fake screenshots.

Link to the original post Amph.

https://bitcointalk.org/index.php?topic=1572003.0

https://bitcointalk.org/index.php?topic=1572003.msg15791191#msg15791191

he is using a bios mod to make the 1080 look like a 1070, probably name change only, the mystery is solved, it was fake like i said in that thread, but everyone was believing the opposite just because of the hate...

[tbearhere]

What is the command to send the help file to my printer plz.  thx
And or make a notepad file that can be saved and read or printed.

ccminer1.8.exe --help
pause:

ccminer1.8.exe --help > ready_to_print.txt

))

Thx rednoW much appreciated. 

[_javi_]

GPU #0: scanhash_lbry:133 misaligned address
Cuda error in func 'scanhash_lbry' at line 115 : misaligned address.
GPU #0: scanhash_lbry:136 misaligned address
...


MSi 1070, no OC, dr368.81, w7 64 .. was working fine until these lines appeared.

what´s wrong now??

[Epsylon3]

hard to know, but look like other miners (including me) doesn't have this problem

[Epsylon3]

try the newer release (1.8.1)

[pallas]

Thanks tpruvot for the new release! :-)
The shared mem of sha256 was, in fact, useless ;-)

[Epsylon3]

was not exactly useless, shared mem is like a "Café Gourmant", its often better but cost a lot ^^ and you can only eat it if you have enough space in your stomac

[Delphix]

try the newer release (1.8.1)

5% increase on GTX1070 for LBRY algo. Thnx mate.

[Amph]

same around 4-5% iincrease, i wonder if x64 can not be build with cuda 8?

[pallas]

was not exactly useless, shared mem is like a "Café Gourmant", its often better but cost a lot ^^ and you can only eat it if you have enough space in your stomac

Interesting parallel, but IRRC, in the sha256 case shared mem was used to store constants which could easily fit directly into the code.