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

[pallas]

Look like only the 750ti on cuda 7.5 doesn't profit so much

yes. doesn't work. slow.. back to the drawing board... Donate a 750ti to pallas so that he can advance from being a black belt developer.

Hehe! sp_ you'd better be kind with me, you know I've got a weapon against you ;-)

[1713554171]

1713554171

[1713554171]

1713554171

[joblo]

Is this legit?

I don't see this version on your releases page. Someone recently posted having downloading an infected
miner from CMB so I'm suspicious.

http://cryptomining-blog.com/7679-updated-ccminer-1-7-6-fork-by-tpruvot-with-improved-decred-support/

[Epsylon3]

i dont know who write on CMB, and its NOT the final 1.7.6

[USScrypto]

It says compiled form latest git source, not that it is final release. Don't see too much of hashrate improvement with it though.

[pallas]

It says compiled form latest git source, not that it is final release. Don't see too much of hashrate improvement with it though.

Looks like on Windows it doesn't perform as well as on Linux. Will try to fix it asap.

[CodyF86]

Pallas the 10% commit is giving me this error when I run it.

reject reason: submit() takes exactly 6 arguments (7 given)

for every share.

[pallas]

Pallas the 10% commit is giving me this error when I run it.

reject reason: submit() takes exactly 6 arguments (7 given)

for every share.

this is not a problem of my kernel but of the stratum code/pool, are you on suprnova?

[sp_]

This is because the decred protocol has changed with the vote parameter in 1.7.5/1.7.6 and the pools haven't updated their software.

Yiimp.com is working fine, and the fee is reduced to 4%.

[Vaccinus]

so when the new version will be officially released?

[flipclip]

so when the new version will be officially released?

I believe Epsylon3 response will be "when it is ready". 

[Epsylon3]

i already answered i'm waiting for decred vote tests... to release it. I dont think Suprnova is ready for the moment.

Also stay tuned, possible nvidia news annoncements today

[Schleicher]

Also stay tuned, possible nvidia news annoncements today

New Tesla card: P100
Too bad we don't need floating point math :-)

[Epsylon3]

Tesla P100 AND Pascal GP100

also Cuda 8 as expected : https://devblogs.nvidia.com/parallelforall/cuda-8-features-revealed/

[myagui]

Tesla P100 AND Pascal GP100

also Cuda 8 as expected : https://devblogs.nvidia.com/parallelforall/cuda-8-features-revealed/

No consumer models announced yet 
And so the wait continues 

[bathrobehero]

Tesla P100 AND Pascal GP100

also Cuda 8 as expected : https://devblogs.nvidia.com/parallelforall/cuda-8-features-revealed/

No consumer models announced yet 
And so the wait continues 

They will probably announce them in June at Computex.

[CodyF86]

Sorry I didn't answer sooner was on the way out the door.

Just running long poll for now, it is about a 10% improvement for me with cuda 7.5.

Feel bad not mining on your yiimp, but I need a constant block rate, it's been financing a hobby of mine hehe.

[pallas]

Sorry I didn't answer sooner was on the way out the door.

Just running long poll for now, it is about a 10% improvement for me with cuda 7.5.

Feel bad not mining on your yiimp, but I need a constant block rate, it's been financing a hobby of mine hehe.

the donation address is in my signature ;-)

[alexis78]

@pallas

Hi and nice to meet you here too .
Again, nice work on the decred algo.
Though i have a small objection.
You didn't credit the logic which led you get the boost and since the code is on an open source world, credits are all that matters.

If you want some inspiration here is a 1.5% fee binary which reports locally a ~+6% boost from the latest decred commit on 970

http://s000.tinyupload.com/index.php?file_id=07209038321464518069
A little hint: The devil hides in kernel and midstate computation

I'll opensource it in the near future, it's just the fact that too much work was put into it

[pallas]

@pallas

Hi and nice to meet you here too .
Again, nice work on the decred algo.
Though i have a small objection.
You didn't credit the logic which led you get the boost and since the code is on an open source world, credits are all that matters.

If you want some inspiration here is a 1.5% fee binary which reports locally a ~+6% boost from the latest decred commit on 970

http://s000.tinyupload.com/index.php?file_id=07209038321464518069
A little hint: The devil hides in kernel and midstate computation

I'll opensource it in the near future, it's just the fact that too much work was put into it

No Sir, the kernel I posted is all my own work. I saw your kernel AFTER finishing mine. Believe it or not, the multiple nonces per round was a novel idea to me, which I had cause of insomnia :-)
And I also have other speedups which I may opensource in the future.

[alexis78]

I didnt know of that. Anyway, no bad feelings.
I'm very curious to know what led you do this thing 

Since you saw my kernel, what do you think about the way i divide the scanning ranges of threads?

Isn't it a more accurate way to experiment with parameters
since you can use non powers (or even multiples) of 2 in intensity, blocks, nonces_per_thread and actually dont hash nonces > maxNonce while maintaing 100% threads of the block for the kernel's life span? (except on the last iterations)