[ANN] cudaMiner & ccMiner CUDA based mining applications [Windows/Linux/MacOSX] - page 830.

bigjme

sr. member

Activity: 350

Merit: 250

Quote from: archtme on February 09, 2014, 11:00:30 AM

Same! I spent the first 3-4 hours trying to get that stuff going, and when I finally managed to CPU mine I got wind of this thread lol

i had the cudaminer and everything ready. tried to run the wallet and it destroyed my entire installed. couldnt even login again

which version do i need christian?

basic
c#
c++

i believe its c++?

so glad i have a spare 2TB HDD, my SSD is getting very full installing everything lmao

cbuchner1

hero member

Activity: 756

Merit: 502

Quote from: Thirtybird on February 09, 2014, 10:56:33 AM

This explains why it won't compile in MinGW - drats!

nVidia does not support MinGW at all with the CUDA runtime APIs. It may be possible to do something with the driver APIs, but you may have to adapt the CUDA header files for use in MinGW.

the Microsoft specific parallel_for construct could be replaced with OpenMP easily by simply changing the #if WIN32 check to something that explicitly tests for MinGW or gcc and uses OpenMP like under Linux

archtme

newbie

Activity: 19

Merit: 0

Quote from: bigjme on February 09, 2014, 10:52:33 AM

2010 nearly downloaded, im following your guide ManIkWeet

i think for my next release im going to have 2 machines on standby with windows and linux for the wallets
the linux wallet destroyed my install and cost me stupid amounts of coins in the process

Same! I spent the first 3-4 hours trying to get that stuff going, and when I finally managed to CPU mine I got wind of this thread lol

cbuchner1

hero member

Activity: 756

Merit: 502

Quote from: AizenSou on February 09, 2014, 10:55:42 AM

The problem is I don't know how to define "high" N-factor. 10 is already high or from 13 ?? I got 500 khs for GTX 780 with T-config and N=9 (MRC). You think it is ok ?

trial and error I would say. It also depends on how much RAM there is on your card. For a 1 GB card the "high" N-factors start earlier than for 4GB cards.

For MRC I get 500 also using -l T24x24

"high" would start around N-factor 12 I think.

Christian

Thirtybird

hero member

Activity: 693

Merit: 500

Quote from: cbuchner1 on February 09, 2014, 10:31:58 AM

Quote from: bigjme on February 09, 2014, 10:19:51 AM

yes it is, i think it would depend on our speed, as high end amd are getting over 5GH/s

chrisitian is visual studio 2010 still the best for compiling cudaminer in windows? as i have access to every visual basic out in full licensed versions

I use VS2010 because later editions give no benefit at all. We don't need no AVX2 support. We don't need any of the newer .NET stuff, and we don't use funky C++11 constructs. And the parallel_for construct from Microsoft's parallel patterns library is also nicely supported in the 2010 Express edition (unlike OpenMP which requires the licensed pay-for editions).

Also CUDA 5.5 might not even support the very latest editions of Visual Studio yet.

Christian

This explains why it won't compile in MinGW - drats!

AizenSou

hero member

Activity: 938

Merit: 1000

Quote from: cbuchner1 on February 09, 2014, 10:47:03 AM

Quote from: AizenSou on February 09, 2014, 10:39:56 AM

What kind of scrypt-jane coin do u mine right now Christian? The kernel setting for scrypt-jane coin is the same for every N-factor, it is right ?

nope, high N-factor scrypt-jane coins must use lower case kernels (f,k,t) for efficient mining - and also investigate the use of lookup gap.
Yac mining works best with higher lookup gaps in the range of 4-7 on powerful nVidia GPUs.

There should be an optimal kernel config for each N-factor. Sorry about that Wink

Autotune will help a great deal

with low N factor coins use upper case letter kernels and I think the optimal settings will not be so strongly dependent on the exact N factor.

Christian

The problem is I don't know how to define "high" N-factor. 10 is already high or from 13 ?? I got 500 khs for GTX 780 with T-config and N=9 (MRC). You think it is ok ?

bigjme

sr. member

Activity: 350

Merit: 250

2010 nearly downloaded, im following your guide ManIkWeet

i think for my next release im going to have 2 machines on standby with windows and linux for the wallets
the linux wallet destroyed my install and cost me stupid amounts of coins in the process

ManIkWeet

full member

Activity: 182

Merit: 100

Quote from: cbuchner1 on February 09, 2014, 10:31:58 AM

Also CUDA 5.5 might not even support the very latest editions of Visual Studio yet.

Christian

Indeed, Visual Studio 2013 (Yay university) doesn't have CUDA support.

cbuchner1

hero member

Activity: 756

Merit: 502

Quote from: AizenSou on February 09, 2014, 10:39:56 AM

What kind of scrypt-jane coin do u mine right now Christian? The kernel setting for scrypt-jane coin is the same for every N-factor, it is right ?

nope, high N-factor scrypt-jane coins must use lower case kernels (f,k,t) for efficient mining - and also investigate the use of lookup gap.
Yac mining works best with higher lookup gaps in the range of 4-7 on powerful nVidia GPUs.

There should be an optimal kernel config for each N-factor. Sorry about that Wink

Autotune will help a great deal

with low N factor coins use upper case letter kernels and I think the optimal settings will not be so strongly dependent on the exact N factor.

Christian

AizenSou

hero member

Activity: 938

Merit: 1000

Quote from: djm34 on February 09, 2014, 09:18:16 AM

much less profitable if you don't have that hashing power... I think it will be my last day on maxcoin.

Yep I agree. I have stop my cudaminer to Maxcoin and point back to scrypt-jane. We have lost our advantages. AMD is much more efficient and faster for keccak. My R9 280x got 320 Mhs and draws only 230w, for scrypt is 740Khs and draws 300w. Maxcoin is still very profitable for AMD folks.

Like Christian has stated before, nVidia only has advantages with more memory extensive tasks like scrypt-jane.

Quote from: cbuchner1 on February 09, 2014, 09:26:16 AM

you could mine a few more and speculate on a rising coin value later on. That's what I did with MRC. And I still hold a few UTC.
https://crypto-trade.com/trade/utc_btc

Christian

What kind of scrypt-jane coin do u mine right now Christian? The kernel setting for scrypt-jane coin is the same for every N-factor, it is right ?

bigjme

sr. member

Activity: 350

Merit: 250

ok i just downloaded Microsoft Visual Studio 2010 Service Pack 1
uni comes in handy for some things :-)

cbuchner1

hero member

Activity: 756

Merit: 502

Quote from: bigjme on February 09, 2014, 10:33:25 AM

so are you recommending express or sp1?
i want to set my linux back up so i can compile using that but im sick of it dieing now

2010 Express with SP1 using my precompiled dependencies is the quickest way to get going.

Christian

bigjme

sr. member

Activity: 350

Merit: 250

so are you recommending express or sp1?
i want to set my linux back up so i can compile using that but im sick of it dieing now

cbuchner1

hero member

Activity: 756

Merit: 502

Quote from: bigjme on February 09, 2014, 10:19:51 AM

yes it is, i think it would depend on our speed, as high end amd are getting over 5GH/s

chrisitian is visual studio 2010 still the best for compiling cudaminer in windows? as i have access to every visual basic out in full licensed versions

I use VS2010 because later editions give no benefit at all. We don't need no AVX2 support. We don't need any of the newer .NET stuff, and we don't use funky C++11 constructs. And the parallel_for construct from Microsoft's parallel patterns library is also nicely supported in the 2010 Express edition (unlike OpenMP which requires the licensed pay-for editions).

Also CUDA 5.5 might not even support the very latest editions of Visual Studio yet.

Christian

djm34

legendary

Activity: 1400

Merit: 1050

Quote from: bigjme on February 09, 2014, 10:28:37 AM

i meant studio :p

don't think, it will make a difference, it's just a big interface for people who don't like make...

bigjme

sr. member

Activity: 350

Merit: 250

i meant studio :p

djm34

legendary

Activity: 1400

Merit: 1050

Quote from: bigjme on February 09, 2014, 10:19:51 AM

yes it is, i think it would depend on our speed, as high end amd are getting over 5GH/s

chrisitian is visual basic 2010 still the best for compiling cudaminer in windows? as i have access to every visual basic out in full licensed versions

visual basic ?

bigjme

sr. member

Activity: 350

Merit: 250

yes it is, i think it would depend on our speed, as high end amd are getting over 5GH/s

chrisitian is visual studio 2010 still the best for compiling cudaminer in windows? as i have access to every visual basic out in full licensed versions

cbuchner1

hero member

Activity: 756

Merit: 502

Quote from: bigjme on February 09, 2014, 10:05:30 AM

checked prices, their 0.06 micro btc :p
so 0.00006btc each

yeah, but with a steady upwards trend Wink

see the graph I posted.

bigjme

sr. member

Activity: 350

Merit: 250

it would be similar to doing prime95 on a cpu benchmark. but with much faster crunching from a gpu

it will be interesting to see how we perform on blakecoin, its been out much longer and ~~still has a good price~~
#
checked prices, their 0.06 micro btc :p
so 0.00006btc each

Topic: [ANN] cudaMiner & ccMiner CUDA based mining applications [Windows/Linux/MacOSX] - page 830. (Read 3426922 times)