CCminer(SP-MOD) Modded GPU kernels.

[scryptr]

sp_

I confirm that built from your github source pre-release 45 gives 4540khs instead of the best previous 4418khs of r43.
R44 was no good for me except whirpool and x15.
Quark seems to be very close to R43 record 5717khs but R43 still better for quark

oced 750non-ti 1gb,

comparing v43 vs. v44 for my gtx750(non-ti) (windows binaries from github)
x15: 1966 vs 1976
x13: 2272 vs 2270
x11: 2886 vs 2883
quark: 5700 vs 5660
qubit: 4418 vs 4419 (with drops)
lyra: 882 vs 881

So, only x15 gain for me

v45-pre (?)

After allowing my 6x750ti rig to run Quark 24hr+, I notice that the rig performs better than previously reported.  When I came home today, the rig was running at 36600+ kh/s, and it is currently running at 36580kh/s as I type this.  This is better than yesterday's report of 36500kh/s.  Individual cards run from 6050kh/s to 6160kh/s .       --scryptr

[DragonSlayer]

I just picked up a gtx980 over the weekend.  Using version 44 I am getting 10200kh/s.  One thing I notice the pools are only registering around 8000kh/s.  Anyone else notice this as well?  I tried out 2 different pools with same results.  Oh, this is on x11.

Thanks

[sp_]

Wich pools? (PLNS or PROP payout?)
Windows/linux?
Did you build it yourself?
if you build yourself you must add compute 5.2 in the project file.
Do you get any (does not validate on the cpu)?
Did you try on lower clockrates? on a standard card(Non OC) you should get around 9.2MHASH

What is the reject rate in percent.
What is the fee in the pool.

Could you show some screenshots?

[sp_]

Submitted a 10% increase in skeincoin.
The change was done in the sha-256 implementation
-Reduced register usage by moving constants from constmem to the instruction   cache
-bether launch configuration
-bether default intensity

The last pass of skeincoin is a pure SHA-256 pass(bitcoin) Probobly it would be bether to use the bitcoin kernal implemented by klaus_t.
Or any ASIC miner in hybrid mining.

NVIDIA does the SKEIN and the asic's does the sha256 and the search.

[sp_]

Submitted a 10% increase in skeincoin.
The change was done in the sha-256 implementation
-Reduced register usage by moving constants from constmem to the instruction   cache
-bether launch configuration
-bether default intensity

The last pass of skeincoin is a pure SHA-256 pass(bitcoin) Probobly it would be bether to use the bitcoin kernal implemented by klaus_t.
Or any ASIC miner in hybrid mining.
NVIDIA does the SKEIN and the asic's does the sha256 and the search.

I'm guessing you don't know how ASIC miners are implemented...
It returns winning nonces, not hashes - it functions as a BTC-mining black-box.

I dont need the hash, i need the wining nounce. (in skeincoin the SHA-256 is the last pass)
But I need to feed the Hardware with input data, and this is not supported or is it?

[sp_]

I just put the data of a 750ti light overclock in a  Myriadcoin Profitability Calculator
I I can boost the hashrate from 60 to 82 MHASH(36.6%) it will be more profitable than mining qubit.


http://birdonwheels5.no-ip.org/profit-calc/
 

[chrysophylax]

I just put the data of a 750ti light overclock in a

Myriadcoin Profitability Calculator

http://birdonwheels5.no-ip.org/profit-calc/
 

these are NEVER accurate sp ...

its based on the current diff and the current availability of miners ...

the moment you put another miner on there - especially if its a farm - that blows the so called 'profitability' out the window ...

i do like quibit tho ...

#crysx

[sp_]

these are NEVER accurate sp ...
its based on the current diff and the current availability of miners ...
the moment you put another miner on there - especially if its a farm - that blows the so called 'profitability' out the window ...
i do like quibit tho ...
#crysx

I know, but calculators say something about the speed of the hashing algo. People usually don't mine when the miner use more power than it gains. Some of the miners have
faster mining software or hardware.

Bitcoin is not profitable on GPU's (ASIC's)
Scrypt is not profitable on GPU's (ASIC's)
groestl is not profitable on NVIDIA GPU's. (AMD has a fast opensource implementation).
SKEIN could be profitable with some work on NVIDIA. What is the speed on AMD cards?
QUBIT is faster on NVIDIA

Also note that qubit ccminer spmod version 44 is

1750% more profitable to mine than SCRYPT in cudaminer. (on the same hardware. 750ti OC)

[chrysophylax]

these are NEVER accurate sp ...
its based on the current diff and the current availability of miners ...
the moment you put another miner on there - especially if its a farm - that blows the so called 'profitability' out the window ...
i do like quibit tho ...
#crysx

I know, but calculators say something about the speed of the hashing algo. People usually don't mine when the miner use more power than it gains. Some of the miners have
faster mining software or hardware.

Bitcoin is not profitable on GPU's (ASIC's)
Scrypt is not profitable on GPU's (ASIC's)
groestl is not profitable on NVIDIA GPU's. (AMD has a fast opensource implementation).
SKEIN could be profitable with some work on NVIDIA. What is the speed on AMD cards?
QUBIT is faster on NVIDIA

Also note that qubit ccminer spmod version 44 is

1750% more profitable to mine than SCRYPT in cudaminer. (on the same hardware. 750ti OC)

agreed ... if its about the profitability ...

though wouldnt it make sense to focus on the algos that are implemented in ccminer ( nvidia ) and optimize them to the fullest extent they can go? ...

possibly implement more algos to your fork and make it an all round better miner - even if amd have better opensource implementations ...

unless of course it is possible for ccminer to have both nvidia and amd implementations ...

#crysx

[sp_]

agreed ... if its about the profitability ...
though wouldnt it make sense to focus on the algos that are implemented in ccminer ( nvidia ) and optimize them to the fullest extent they can go? ...
possibly implement more algos to your fork and make it an all round better miner - even if amd have better opensource implementations ...
unless of course it is possible for ccminer to have both nvidia and amd implementations ...
#crysx

I will focus on the algos that give a profit. But the profit keeps changing, so in the end most of the algos are faster.

But in small steps. Optimizing is time consuming.

After Wolf0's x11,x13 and x15 binaries leaked he stopped optimizing them and moved over to other algos.. It doesn't meen that the wolf0's x11 implentation is the fastest possible.

[chrysophylax]

agreed ... if its about the profitability ...
though wouldnt it make sense to focus on the algos that are implemented in ccminer ( nvidia ) and optimize them to the fullest extent they can go? ...
possibly implement more algos to your fork and make it an all round better miner - even if amd have better opensource implementations ...
unless of course it is possible for ccminer to have both nvidia and amd implementations ...
#crysx

I will focus on the algos that give a profit. But the profit keeps changing, so in the end most of the algos are faster.

But in small steps. Optimizing is time consuming.

After Wolf0's x11,x13 and x15 binaries leaked he stopped optimizing them and moved over to other algos.. It doesn't meen that the wolf0's x11 implentation is the fastest possible.

of course ... i guess the whole concept of optimization is spending time and effort to make the algo perform better - and faster ...

profitability plays a big part in that area - otherwise its not worth it - i understand ...

so if i understand correctly - there is more to be had in the way of better and faster performance with most of these algos - including x11 x13 x15? ...

#crysx

[ol92]

agreed ... if its about the profitability ...
though wouldnt it make sense to focus on the algos that are implemented in ccminer ( nvidia ) and optimize them to the fullest extent they can go? ...
possibly implement more algos to your fork and make it an all round better miner - even if amd have better opensource implementations ...
unless of course it is possible for ccminer to have both nvidia and amd implementations ...
#crysx

I will focus on the algos that give a profit. But the profit keeps changing, so in the end most of the algos are faster.

But in small steps. Optimizing is time consuming.

After Wolf0's x11,x13 and x15 binaries leaked he stopped optimizing them and moved over to other algos.. It doesn't meen that the wolf0's x11 implentation is the fastest possible.

ccminer was profitable with whirlpoolx, but smelter released a miner with >500mh/s for 1 280X (https://bitcointalk.org/index.php?topic=977245.900). Now this an amd only coin. A GTX 980 can do almost 300mh/s which used to be good, few weeks ago.

If you can do something like 600mh/s for a GTX 980, it will be profitable again.

[sp_]

QUARK algo @yaamp multipool: (13-april-2015)

AMD GPU miners: 0.72%
CPU miners: 0.09%
ccminer: 99,19% (sp-mod: 95,79%)

Quark does 20MHASH on the gtx 980 with overclocking in the latest sp-mod

Smolen's  "Smelter - GPU miner for Quark-based coins" is doing 2 MHASH on the 280x

The same power consumption

[sp_]

of course ... i guess the whole concept of optimization is spending time and effort to make the algo perform better - and faster ...
profitability plays a big part in that area - otherwise its not worth it - i understand ...
so if i understand correctly - there is more to be had in the way of better and faster performance with most of these algos - including x11 x13 x15? ...
#crysx

x11,x13 and x15 can be alot faster on NVIDIA if all of the AES implementations are rewritten.

In my quark implementation the chained algorithms are running in parallel, and the speed of the total hash is just a little bit slower than the slowest of the algos wich is groestl.

This meens that quark is not much more secure than groestl with a proper implementation. When kernals are running in parallell, they use 0 time.

Go and write a paper about it.

On bitcointalk.org all we care about it profit and hashrate.

[MaxDZ8]

You probably mean they run pipelined (pipelined != parallel).

[chrysophylax]

of course ... i guess the whole concept of optimization is spending time and effort to make the algo perform better - and faster ...
profitability plays a big part in that area - otherwise its not worth it - i understand ...
so if i understand correctly - there is more to be had in the way of better and faster performance with most of these algos - including x11 x13 x15? ...
#crysx

x11,x13 and x15 can be alot faster on NVIDIA if all of the AES implementations are rewritten.

In my quark implementation the chained algorithms are running in parallel, and the speed of the total hash is just a little bit slower than the slowest of the algos wich is groestl.

This meens that quark is not much more secure than groestl with a proper implementation. When kernals are running in parallell, they use 0 time.

Go and write a paper about it.

On bitcointalk.org all we care about it profit and hashrate.

mate - im not in the business of writing papers - nor are we in the business of making it all about profit for the short term ...

its about performance and hashrate - agreed ... but also consistency and reliability ... once that is optimal - the profit and the rest follow ...

you and the other devs have a great deal of respect from us lowly miners ... ...

but thats one thing we know how to do REALLY well ... mine! ...

#crysx

[tbearhere]

Yesterday I found out that the hefty 1 algo (mjolnir,heavycoin,securecoin?) is producing wrong results.
Are there any other coins that are using this algo?
If not, I will remove the support in my fork

unfortunately i do not know of any others ...
but i do have a question ... are you going to implement the spreadcoinx11 algo ( amongst some others ) in this fork as you previously mentioned? ...
tanx sp ...
#crysx

I might add scrypt-n..

Just checked, not many profitable scrypt-n coins out there..

Spreadcoin is still only for the donators. New version is coming soon with more hash.

tanx sp ...

we are one of the donors ...

but i thought i would ask due to it being a pain in the backside to continue changing miners on EVERY worker in the farm when we want to mine sprx11 ...

would be so much easier if it was within the one miner ...

tanx again ...

#crysx

Yes for me too. It would be easier if sp put them into one miner. How about it sp.

[antonio8]

Yesterday I found out that the hefty 1 algo (mjolnir,heavycoin,securecoin?) is producing wrong results.
Are there any other coins that are using this algo?
If not, I will remove the support in my fork

unfortunately i do not know of any others ...
but i do have a question ... are you going to implement the spreadcoinx11 algo ( amongst some others ) in this fork as you previously mentioned? ...
tanx sp ...
#crysx

I might add scrypt-n..

Just checked, not many profitable scrypt-n coins out there..

Spreadcoin is still only for the donators. New version is coming soon with more hash.

tanx sp ...

we are one of the donors ...

but i thought i would ask due to it being a pain in the backside to continue changing miners on EVERY worker in the farm when we want to mine sprx11 ...

would be so much easier if it was within the one miner ...

tanx again ...

#crysx

Yes for me too. It would be easier if sp put them into one miner. How about it sp.

Then non donors would have SPR.

I agree to keep them separate.

[sp_]

I will add some more algos.

I'm Planning to add  Bitcredits and scryptn.

[djm34]

I will add some more algos.

I'm Planning to add  Bitcredits and scryptn.

bitcredits ? what is that ?