Author

Topic: Nexus - Pure SHA3 + CPU/GPU + nPoS + 15 Active Innovations + More to Come - page 304. (Read 785514 times)

sr. member
Activity: 433
Merit: 250
Reducing the amount of gpu threads in the conf fixed the cpu usage going up to 90% and the hash is not 0 kh/s any more.
legendary
Activity: 1512
Merit: 1000
quarkchain.io
Thanks , Ill check out
EDIT: Stabilized it at 145 MH/s with 5x280X Vapor-X, but did not find any block.
I have a rig of Nvidia which mining with 110MH/s and hits blocks ...
legendary
Activity: 1400
Merit: 1050
Hi djm , which speeds do you reach on 280X and 750TI with your new miner Smiley
EDIT: It crashes on my 280X after a couple of blocks
Code:
{
"connections" :
[
{
"url" : "127.0.0.1:9325",
"user" : "x",
"password" : "x"
}
],

"algorithm" : "sk1024_djm",
"gpu-engine" : "1100",
"gpu-memclock" : "1500",
"gpu-shaders" : "2048",
"gpu-threads" : "2",
"gpu-vddc" : "1200",
"gpu-powertune" : "0",
"intensity" : "19",
"xintensity" : "0",
"rawintensity" : "0",
"thread-concurrency" : "8192",
"lookup-gap" : "2",
"worksize" : "128"
}
please use the config file in directory "Release/Resources/Config" not the one in "Release".
You are currently using the setting of my R9... (by default... I also removed vddc... that's why I know you are not using the correct one...)

Regarding the cards, I don't have a 280x and for the 750ti, I use the other miners (this one won't work actually with an nvidia card) and it gets around 28MH/s with the 750ti
legendary
Activity: 1512
Merit: 1000
quarkchain.io
Hi djm , which speeds do you reach on 280X and 750TI with your new miner Smiley
EDIT: It crashes on my 280X after a couple of blocks
Code:
{
"connections" :
[
{
"url" : "127.0.0.1:9325",
"user" : "x",
"password" : "x"
}
],

"algorithm" : "sk1024_djm",
"gpu-engine" : "1100",
"gpu-memclock" : "1500",
"gpu-shaders" : "2048",
"gpu-threads" : "2",
"gpu-vddc" : "1200",
"gpu-powertune" : "0",
"intensity" : "19",
"xintensity" : "0",
"rawintensity" : "0",
"thread-concurrency" : "8192",
"lookup-gap" : "2",
"worksize" : "128"
}
legendary
Activity: 1400
Merit: 1050
ok here the first version of the miner: (source+binaries) http://ge.tt/7Z2nZ132/v/0?c

I haven't made a github repo yet.
It was compiled with vs2013.
the source is in "Miner", the binaries in "Release" (from the root, there is another Release directory in "Miner" but you won't find anything).
There is a bat example called "test_750ti.bat" (because I am lazy...).
A config file can be found (and must be set to match your card) in Resources/Config
It is a standard sgminer miner config file.
The kernel is in Kernel directory and should remain there.

Currently in the config file, this is my best setting for the my 290x with 14.9 drivers
And leads to 53MH/s before the card throttle down because of high temp.
With power at -30% (where the temp is stable I still get 45~48MH/s)
powertune does not seem to work (actually it does not work either in msi AB, I have been able to set it only from catalyst app).

Right now, it is a one kernel implementation, 2 kernels might improve the hashrate. (got many exception error when trying to implement it, then I settled down to 1 kernel...)

It was tested on testnet, I haven't had the patience to wait for a block on main net (I know that someone with a big farm is running on coinshield... so I didn't insist...  Grin but it should work)

donation address: 2S2pCpRXyb8Lpre52U3Xjq2MguSdaea5YGjVTsJqgZBfL2S24ag
Thank you very much. It's good to see progress.
Unfortunately I have only older AMD GPUs and the miner doesn't seems to work even with my 7950 which is the best I have right now. (the miner reports 0 khash/s)
Another problem that I observed that the server address is hard coded to 127.0.0.1:9325 or at least I wasn't able to modify using the config file.
EDIT: CPU usage is 100% (4 cores)
it's not hardcoded, it can be changed in the config file which is in Release/Ressource/Config (remove the one in Release, it isn't used... ), may-be you can also better parameter for your card.
(I don't have cpu usage to 100%, it goes up to 20% which is already high, which actually is a little less than the nvidia version... may-be playing with the config, could help... )


regarding the IP:PORT design, I don't how it will work with pools... (same question for both miners actually)

edit: actually if it mines at 90%, I could be because it doesn't find the gpu... (can you check if the gpu is really used... I don't think it can use the opencl of the cpu either...)
member
Activity: 92
Merit: 10
ok here the first version of the miner: (source+binaries) http://ge.tt/7Z2nZ132/v/0?c

I haven't made a github repo yet.
It was compiled with vs2013.
the source is in "Miner", the binaries in "Release" (from the root, there is another Release directory in "Miner" but you won't find anything).
There is a bat example called "test_750ti.bat" (because I am lazy...).
A config file can be found (and must be set to match your card) in Resources/Config
It is a standard sgminer miner config file.
The kernel is in Kernel directory and should remain there.

Currently in the config file, this is my best setting for the my 290x with 14.9 drivers
And leads to 53MH/s before the card throttle down because of high temp.
With power at -30% (where the temp is stable I still get 45~48MH/s)
powertune does not seem to work (actually it does not work either in msi AB, I have been able to set it only from catalyst app).

Right now, it is a one kernel implementation, 2 kernels might improve the hashrate. (got many exception error when trying to implement it, then I settled down to 1 kernel...)

It was tested on testnet, I haven't had the patience to wait for a block on main net (I know that someone with a big farm is running on coinshield... so I didn't insist...  Grin but it should work)

donation address: 2S2pCpRXyb8Lpre52U3Xjq2MguSdaea5YGjVTsJqgZBfL2S24ag

Great! Thx Smiley

I got error when run the bat file: First problem was the "msvcp120.dll" is missing. After installing dll "the application was unable to start correctly 0xc00007b"

My system is Win 7 64bit , AMD MSI 280x .... Any help ?

Install vcredist_x86.exe from here: http://www.microsoft.com/en-us/download/details.aspx?id=40784


I did. Work ! Thank Smiley
sr. member
Activity: 291
Merit: 250
ok here the first version of the miner: (source+binaries) http://ge.tt/7Z2nZ132/v/0?c

I haven't made a github repo yet.
It was compiled with vs2013.
the source is in "Miner", the binaries in "Release" (from the root, there is another Release directory in "Miner" but you won't find anything).
There is a bat example called "test_750ti.bat" (because I am lazy...).
A config file can be found (and must be set to match your card) in Resources/Config
It is a standard sgminer miner config file.
The kernel is in Kernel directory and should remain there.

Currently in the config file, this is my best setting for the my 290x with 14.9 drivers
And leads to 53MH/s before the card throttle down because of high temp.
With power at -30% (where the temp is stable I still get 45~48MH/s)
powertune does not seem to work (actually it does not work either in msi AB, I have been able to set it only from catalyst app).

Right now, it is a one kernel implementation, 2 kernels might improve the hashrate. (got many exception error when trying to implement it, then I settled down to 1 kernel...)

It was tested on testnet, I haven't had the patience to wait for a block on main net (I know that someone with a big farm is running on coinshield... so I didn't insist...  Grin but it should work)

donation address: 2S2pCpRXyb8Lpre52U3Xjq2MguSdaea5YGjVTsJqgZBfL2S24ag
Thank you very much. It's good to see progress.
Unfortunately I have only older AMD GPUs and the miner doesn't seems to work even with my 7950 which is the best I have right now. (the miner reports 0 khash/s)
Another problem that I observed that the server address is hard coded to 127.0.0.1:9325 or at least I wasn't able to modify using the config file.
EDIT: CPU usage is 100% (4 cores)
sr. member
Activity: 433
Merit: 250
Code:
platform name Advanced Micro Devices, Inc. index 0
threadconcurrency 1030
ADL Overdrive 5 Was Found!
ADL Overdrive 6 Was Found!
ADL index 0, id 65353856 - BIOS partno.: 113-AD55200-101, version: 015.032.000.001, date: 2013/03/31
 22:04
ADL_Overdrive5_FanSpeed_Get failed for default value!
GPU 0 call to fanspeed get failed!
vecgpudata size 1

                ********* Coinshield sk-1024 miner for amd GPUs by djm34 *********
                                     original design by bitslapper 2014
                                           Copyright 2014 djm34
                donation address CSD: 2S2pCpRXyb8Lpre52U3Xjq2MguSdaea5YGjVTsJqgZBfL2S24ag

Connected to 127.0.0.1:9325...
[MASTER] Coinshield Network | New Block 32636
[METERS] 0 Hashes | 0.000000 KHash/s | Height = 32636

CPU goes to 90%.
sr. member
Activity: 291
Merit: 250
ok here the first version of the miner: (source+binaries) http://ge.tt/7Z2nZ132/v/0?c

I haven't made a github repo yet.
It was compiled with vs2013.
the source is in "Miner", the binaries in "Release" (from the root, there is another Release directory in "Miner" but you won't find anything).
There is a bat example called "test_750ti.bat" (because I am lazy...).
A config file can be found (and must be set to match your card) in Resources/Config
It is a standard sgminer miner config file.
The kernel is in Kernel directory and should remain there.

Currently in the config file, this is my best setting for the my 290x with 14.9 drivers
And leads to 53MH/s before the card throttle down because of high temp.
With power at -30% (where the temp is stable I still get 45~48MH/s)
powertune does not seem to work (actually it does not work either in msi AB, I have been able to set it only from catalyst app).

Right now, it is a one kernel implementation, 2 kernels might improve the hashrate. (got many exception error when trying to implement it, then I settled down to 1 kernel...)

It was tested on testnet, I haven't had the patience to wait for a block on main net (I know that someone with a big farm is running on coinshield... so I didn't insist...  Grin but it should work)

donation address: 2S2pCpRXyb8Lpre52U3Xjq2MguSdaea5YGjVTsJqgZBfL2S24ag

Great! Thx Smiley

I got error when run the bat file: First problem was the "msvcp120.dll" is missing. After installing dll "the application was unable to start correctly 0xc00007b"

My system is Win 7 64bit , AMD MSI 280x .... Any help ?

Install vcredist_x86.exe from here: http://www.microsoft.com/en-us/download/details.aspx?id=40784
member
Activity: 92
Merit: 10
ok here the first version of the miner: (source+binaries) http://ge.tt/7Z2nZ132/v/0?c

I haven't made a github repo yet.
It was compiled with vs2013.
the source is in "Miner", the binaries in "Release" (from the root, there is another Release directory in "Miner" but you won't find anything).
There is a bat example called "test_750ti.bat" (because I am lazy...).
A config file can be found (and must be set to match your card) in Resources/Config
It is a standard sgminer miner config file.
The kernel is in Kernel directory and should remain there.

Currently in the config file, this is my best setting for the my 290x with 14.9 drivers
And leads to 53MH/s before the card throttle down because of high temp.
With power at -30% (where the temp is stable I still get 45~48MH/s)
powertune does not seem to work (actually it does not work either in msi AB, I have been able to set it only from catalyst app).

Right now, it is a one kernel implementation, 2 kernels might improve the hashrate. (got many exception error when trying to implement it, then I settled down to 1 kernel...)

It was tested on testnet, I haven't had the patience to wait for a block on main net (I know that someone with a big farm is running on coinshield... so I didn't insist...  Grin but it should work)

donation address: 2S2pCpRXyb8Lpre52U3Xjq2MguSdaea5YGjVTsJqgZBfL2S24ag

Great! Thx Smiley

I got error when run the bat file: First problem was the "msvcp120.dll" is missing. After installing dll "the application was unable to start correctly 0xc00007b"

My system is Win 7 64bit , AMD MSI 280x .... Any help ?
member
Activity: 92
Merit: 10
ok here the first version of the miner: (source+binaries) http://ge.tt/7Z2nZ132/v/0?c

I haven't made a github repo yet.
It was compiled with vs2013.
the source is in "Miner", the binaries in "Release" (from the root, there is another Release directory in "Miner" but you won't find anything).
There is a bat example called "test_750ti.bat" (because I am lazy...).
A config file can be found (and must be set to match your card) in Resources/Config
It is a standard sgminer miner config file.
The kernel is in Kernel directory and should remain there.

Currently in the config file, this is my best setting for the my 290x with 14.9 drivers
And leads to 53MH/s before the card throttle down because of high temp.
With power at -30% (where the temp is stable I still get 45~48MH/s)
powertune does not seem to work (actually it does not work either in msi AB, I have been able to set it only from catalyst app).

Right now, it is a one kernel implementation, 2 kernels might improve the hashrate. (got many exception error when trying to implement it, then I settled down to 1 kernel...)

It was tested on testnet, I haven't had the patience to wait for a block on main net (I know that someone with a big farm is running on coinshield... so I didn't insist...  Grin but it should work)

donation address: 2S2pCpRXyb8Lpre52U3Xjq2MguSdaea5YGjVTsJqgZBfL2S24ag

Great! Thx Smiley
sr. member
Activity: 318
Merit: 250
Nexus pool : http://nexusniropool.cestballot.fr/
ok here the first version of the miner: (source+binaries) http://ge.tt/7Z2nZ132/v/0?c

I haven't made a github repo yet.
It was compiled with vs2013.
the source is in "Miner", the binaries in "Release" (from the root, there is another Release directory in "Miner" but you won't find anything).
There is a bat example called "test_750ti.bat" (because I am lazy...).
A config file can be found (and must be set to match your card) in Resources/Config
It is a standard sgminer miner config file.
The kernel is in Kernel directory and should remain there.

Currently in the config file, this is my best setting for the my 290x with 14.9 drivers
And leads to 53MH/s before the card throttle down because of high temp.
With power at -30% (where the temp is stable I still get 45~48MH/s)
powertune does not seem to work (actually it does not work either in msi AB, I have been able to set it only from catalyst app).

Right now, it is a one kernel implementation, 2 kernels might improve the hashrate. (got many exception error when trying to implement it, then I settled down to 1 kernel...)

It was tested on testnet, I haven't had the patience to wait for a block on main net (I know that someone with a big farm is running on coinshield... so I didn't insist...  Grin but it should work)

donation address: 2S2pCpRXyb8Lpre52U3Xjq2MguSdaea5YGjVTsJqgZBfL2S24ag

Thx a lot

Best regards
legendary
Activity: 1400
Merit: 1050
ok here the first version of the miner: (source+binaries) http://ge.tt/7Z2nZ132/v/0?c

I haven't made a github repo yet.
It was compiled with vs2013.
the source is in "Miner", the binaries in "Release" (from the root, there is another Release directory in "Miner" but you won't find anything).
There is a bat example called "test_750ti.bat" (because I am lazy...).
A config file can be found (and must be set to match your card) in Resources/Config
It is a standard sgminer miner config file.
The kernel is in Kernel directory and should remain there.

Currently in the config file, this is my best setting for the my 290x with 14.9 drivers
And leads to 53MH/s before the card throttle down because of high temp.
With power at -30% (where the temp is stable I still get 45~48MH/s)
powertune does not seem to work (actually it does not work either in msi AB, I have been able to set it only from catalyst app).

Right now, it is a one kernel implementation, 2 kernels might improve the hashrate. (got many exception error when trying to implement it, then I settled down to 1 kernel...)

It was tested on testnet, I haven't had the patience to wait for a block on main net (I know that someone with a big farm is running on coinshield... so I didn't insist...  Grin but it should work)

donation address: 2S2pCpRXyb8Lpre52U3Xjq2MguSdaea5YGjVTsJqgZBfL2S24ag
sr. member
Activity: 308
Merit: 250
Djm you are working on miner?

He's trying to get the OpenCL miner to be competitive.

at last it works....
Current speed over my R9-290x 45MH/s (then the temp increase to 94°C and it down clock... getting fed up of this card... but I think it is rather a problem with my card...). The funny thing is that changing the driver from 14.7 to 14.9 increased the speed by around 10MH/s
So I would encourage amd user moving to that version (you want to keep using your old crappy driver: you're on your own... Grin for some reason amd users, always ask for things like that: I still use windows xp, and the new drivers don't work with windows XP...  Grin time to upgrade  and enjoy boiling eggs with your cards !!!)

I think some of the perf issue could have been due to printf inside the kernel during testing (however I can't prove it and won't test the old kernel either... as it isn't my problem  Grin) even within conditional statement, it decreases the speed down to 35kH/s)

I will release soon the source and windows exe after I am done with cleaning and removing some hard-coded crap in the code... (which is kind of funny when a full config file is inputed to a program...)

they won't be any support for linux as I can't test it (and none was included in the files I got...  Grin) but it should be straightforward.

Main issue: cpu usage which is rather high (same problem with the nvidia version actually...)

I asked are you working on it because if you do i knew that you will make it. If someone can do it, it is you! Good job!
legendary
Activity: 1400
Merit: 1050
Djm you are working on miner?

He's trying to get the OpenCL miner to be competitive.

at last it works....
Current speed over my R9-290x 45MH/s (then the temp increase to 94°C and it down clock... getting fed up of this card... but I think it is rather a problem with my card...). The funny thing is that changing the driver from 14.7 to 14.9 increased the speed by around 10MH/s
So I would encourage amd user moving to that version (you want to keep using your old crappy driver: you're on your own... Grin for some reason amd users, always ask for things like that: I still use windows xp, and the new drivers don't work with windows XP...  Grin time to upgrade  and enjoy boiling eggs with your cards !!!)

I think some of the perf issue could have been due to printf inside the kernel during testing (however I can't prove it and won't test the old kernel either... as it isn't my problem  Grin) even within conditional statement, it decreases the speed down to 35kH/s)

I will release soon the source and windows exe after I am done with cleaning and removing some hard-coded crap in the code... (which is kind of funny when a full config file is inputed to a program...)

they won't be any support for linux as I can't test it (and none was included in the files I got...  Grin) but it should be straightforward.

Main issue: cpu usage which is rather high (same problem with the nvidia version actually...)
legendary
Activity: 868
Merit: 1058
Creator of Nexus http://nexus.io
Code:
	/** Breaks the remainder of last composite in Prime Cluster into an integer. 
Larger numbers are more rare to find, so a proportion can be determined
to give decimal difficulty between whole number increases. **/
unsigned int GetFractionalDifficulty(CBigNum composite)
{
/** Break the remainder of Fermat test to calculate fractional difficulty [Thanks Sunny] **/
return ((composite - FermatTest(composite, 2) << 24) / composite).getuint();
}

The fractional Remainder is more than just a Fermat Test. The result is 0.000 in rare cases the proportion is larger than 1.0; this may be the cases you are saying where +2 is prime.

Code:
		double fractionalRemainder = 1000000.0 / GetFractionalDifficulty(next);
if(fractionalRemainder > 1.0 || fractionalRemainder < 0.0)
fractionalRemainder = 0.0;

Viz.
sr. member
Activity: 291
Merit: 250
Hi fellas

I find that the sieve in the Supercomputing miner may be a bit more efficient when you increase the number of offsets going into the sieve. In 3 hours my three GPUs found 9 blocks, which is about 50% more efficient than before.

[...content deleted...]

EDIT: Meh, after a run time of 12 hours I think I can say it must have been a fluke. The first few blocks arrived faster purely by chance.

I've also played with these parameters and had similar experience. Sometime I get an influx of blocks and then nothing, and I can't decide which configuration is good or not. In any case I'm almost sure that  keeping the sieve on 6 even if the difficulty goes down to 5.8+ is still better.
hero member
Activity: 756
Merit: 502
The algo of CPU Channel is the same as XPM?

No it is finding Dense Prime Clusters separated by a maximum gap of +12. XPM is finding Cunningham Chains and Bi-Twin Chains, they are completely different Prime Number Patterns.

Code:
	/** Determines the difficulty of the Given Prime Number.
Difficulty is represented as so V.X
V is the whole number, or Cluster Size, X is a proportion
of Fermat Remainder from last Composite Number [0 - 1] **/
double GetPrimeDifficulty(CBigNum prime, int checks)
{
if(!PrimeCheck(prime, checks))
return 0.0; ///difficulty of a composite number

CBigNum lastPrime = prime;
CBigNum next = prime + 2;
unsigned int clusterSize = 1;

///largest prime gap in cluster can be + 12
///this was determined by previously found clusters up to 17 primes
for( next ; next <= lastPrime + 12; next += 2)
{
if(PrimeCheck(next, checks))
{
lastPrime = next;
++clusterSize;
}
}

///calulate the rarety of cluster from proportion of fermat remainder of last prime + 2
///keep fractional remainder in bounds of [0, 1]
double fractionalRemainder = 1000000.0 / GetFractionalDifficulty(next);
if(fractionalRemainder > 1.0 || fractionalRemainder < 0.0)
fractionalRemainder = 0.0;

return (clusterSize + fractionalRemainder);
}

Thank You,
Viz.



Bug report

the comment says ///calulate the rarety of cluster from proportion of fermat remainder of last prime + 2

yet you're taking the proportion of the fermat remainder of the last number tested +2 before leaving
this loop when the distance exceeds 12.

This is significant because the last number tested is clearly not a prime (because it would be
part of the chain otherwise), and after adding 2 there is a chance that this one *is* a prime,
having a fermat remainder of exactly 1.

I think this is one of the reasons some found constellations get assigned a difficulty of 4.000000, 5.00000,
6.000000 etc. which somewhat reduces the chance of submitting a share for found constellations.

Unfortunately fixing this would require a hardfork.

Christian
legendary
Activity: 868
Merit: 1058
Creator of Nexus http://nexus.io
The algo of CPU Channel is the same as XPM?

No it is finding Dense Prime Clusters separated by a maximum gap of +12. XPM is finding Cunningham Chains and Bi-Twin Chains, they are completely different Prime Number Patterns.

Code:
	/** Determines the difficulty of the Given Prime Number.
Difficulty is represented as so V.X
V is the whole number, or Cluster Size, X is a proportion
of Fermat Remainder from last Composite Number [0 - 1] **/
double GetPrimeDifficulty(CBigNum prime, int checks)
{
if(!PrimeCheck(prime, checks))
return 0.0; ///difficulty of a composite number

CBigNum lastPrime = prime;
CBigNum next = prime + 2;
unsigned int clusterSize = 1;

///largest prime gap in cluster can be + 12
///this was determined by previously found clusters up to 17 primes
for( next ; next <= lastPrime + 12; next += 2)
{
if(PrimeCheck(next, checks))
{
lastPrime = next;
++clusterSize;
}
}

///calulate the rarety of cluster from proportion of fermat remainder of last prime + 2
///keep fractional remainder in bounds of [0, 1]
double fractionalRemainder = 1000000.0 / GetFractionalDifficulty(next);
if(fractionalRemainder > 1.0 || fractionalRemainder < 0.0)
fractionalRemainder = 0.0;

return (clusterSize + fractionalRemainder);
}

Thank You,
Viz.

sr. member
Activity: 392
Merit: 250
The algo of CPU Channel is the same as XPM?
Jump to: