[ANN] cudaMiner & ccMiner CUDA based mining applications [Windows/Linux/MacOSX] - page 1104.

Lacan82

sr. member

Activity: 247

Merit: 250

Quote from: Stoneysilence on April 30, 2013, 03:22:11 PM

Well downloading a new Stratum didn't help. Got the same error message. Added the -pa part too. Stratum is connecting fine to the pool. But Cudaminer isn't connecting properly to Stratum. Stratum keeps throwing up errors about unable to decode http://1 or something. I had the error copied but somehow it didn't copy right.

Will try again later tonight when I have more time.

What are you typing in the stratum proxy exactly?

Stoneysilence

member

Activity: 104

Merit: 10

Well downloading a new Stratum didn't help. Got the same error message. Added the -pa part too. Stratum is connecting fine to the pool. But Cudaminer isn't connecting properly to Stratum. Stratum keeps throwing up errors about unable to decode http://1 or something. I had the error copied but somehow it didn't copy right.

Will try again later tonight when I have more time.

cbuchner1

hero member

Activity: 756

Merit: 502

The boost (if any) in this version is limited to GT6xx cards. For your Tesla with Compute Capability 2.0 there is no difference.

The multiprocessors (SMX) on Kepler cards have really small shared memory in relation to the number of active warps that you need to throw at them to saturate them. So I had to find ways to reduce my shared memory use further so that I could run more warps on each SMX. I made it so that 2 warps now have to compete for one shared memory buffer - and I use spinlocks to safeguard the buffer. The buffers are held only for a relatively short time while doing transfers from and to the card's VRAM.

The technique has some overhead, and hence slows down all cards of non-Kepler type. So I only used it in the "S" kernels.

mhps

hero member

Activity: 516

Merit: 500

CAT.EX Exchange

I just installed 4-22, tried it out, read all 30+ pages, and was about to report my result. Well I find the 4-30 release! That was fast.
OK on my c2075 card I get 168 mH/s. On cgminer or ufasoft I get 75mH/s. That is 120% boost!

I don't find any performance difference between the 4-22 and the 4-30 version (actually the CPU usage reported by win8 increased from 17% to 18%).

Code:

C:\Users\mh\cm>cudaminer -o http://ltc.kattare.com:9332 -O xx:xx -d 1 -l 112x8 -i 0 -C 2
*** CudaMiner for nVidia GPUs by Christian Buchner ***
This is version 2013-04-30 (alpha)
based on pooler-cpuminer 2.2.3 (c) 2010 Jeff Garzik, 2012 pooler
Cuda additions Copyright 2013 Christian Buchner
My donation address: LKS1WDKGED647msBQfLBHV3Ls8sveGncnm

[2013-05-01 02:42:36] 1 miner threads started, using 'scrypt' algorithm.
[2013-05-01 02:42:39] Long-polling activated for http://ltc.kattare.com:9332/LP
[2013-05-01 02:42:39] GPU #1: Tesla C2075 with compute capability 2.0
[2013-05-01 02:42:39] GPU #1: interactive: 0, tex-cache: 2D, single-alloc: 1
[2013-05-01 02:42:39] GPU #1: using launch configuration 112x8
[2013-05-01 02:42:40] GPU #1: Tesla C2075, 28672 hashes, 71.26 khash/s
[2013-05-01 02:43:05] GPU #1: Tesla C2075, 4214784 hashes, 168.68 khash/s
[2013-05-01 02:43:06] LONGPOLL detected new block
[2013-05-01 02:43:06] GPU #1: Tesla C2075, 172032 hashes, 145.91 khash/s
[2013-05-01 02:43:45] GPU #1: Tesla C2075, 6651904 hashes, 169.07 khash/s
[2013-05-01 02:43:46] accepted: 1/1 (100.00%), 169.07 khash/s (yay!!!)
[2013-05-01 02:44:04] LONGPOLL detected new block
[2013-05-01 02:44:04] GPU #1: Tesla C2075, 3153920 hashes, 168.32 khash/s

Just for comparison this is from 4-22 version

Code:

C:\Users\mh\cm>cudaminer -o http://ltc.kattare.com:9332 -O xx:xx -d 1 -l 112x8 -i 0 -C 2
*** CudaMiner for nVidia GPUs by Christian Buchner ***
This is version 2013-04-22 (alpha)
based on pooler-cpuminer 2.2.3 (c) 2010 Jeff Garzik, 2012 pooler
Cuda additions Copyright 2013 Christian Buchner
My donation address: LKS1WDKGED647msBQfLBHV3Ls8sveGncnm

[2013-05-01 02:20:11] 1 miner threads started, using 'scrypt' algorithm.
[2013-05-01 02:20:12] Long-polling activated for http://ltc.kattare.com:9332/LP
[2013-05-01 02:20:12] GPU #1: Tesla C2075 with compute capability 2.0
[2013-05-01 02:20:12] GPU #1: interactive: 0, tex-cache: 2D, single-alloc: 1
[2013-05-01 02:20:12] GPU #1: using launch configuration 112x8
[2013-05-01 02:20:12] GPU #1: Tesla C2075, 28672 hashes, 74.79 khash/s
[2013-05-01 02:20:37] GPU #1: Tesla C2075, 4186112 hashes, 168.67 khash/s
[2013-05-01 02:20:38] accepted: 1/1 (100.00%), 168.67 khash/s (yay!!!)
[2013-05-01 02:21:05] GPU #1: Tesla C2075, 4730880 hashes, 168.81 khash/s
[2013-05-01 02:21:06] accepted: 2/2 (100.00%), 168.81 khash/s (yay!!!)
[2013-05-01 02:21:26] GPU #1: Tesla C2075, 3239936 hashes, 168.99 khash/s
[2013-05-01 02:21:26] accepted: 3/3 (100.00%), 168.99 khash/s (yay!!!)

I have tried to optimize the -l S option. At S224x4 there was a maximum that is slightly slower than simple -l 112x8.

The -i 0 -C 2 option seems to add a few mH/s.

Thank you Christian. Donation is coming your way (once I have any in hand)

cbuchner1

hero member

Activity: 756

Merit: 502

well then

http://lmgtfy.com/?q=download+scrypt+stratum+proxy

Stoneysilence

member

Activity: 104

Merit: 10

Quote from: Tobi042 on April 30, 2013, 04:19:13 AM

Quote from: Stoneysilence on April 29, 2013, 05:11:50 PM

Quote from: sokol815 on April 29, 2013, 04:48:31 PM

sounds like your pool is having problems... I have cudaminer running with stratum_proxy and it is working just fine. Verify that your stratum_proxy is working by testing with cpuminer running on 127.0.0.1:8332

Also important is making sure that stratum proxy starts with -pa scrypt e.g. mine looks like:

Code:

stratum_proxy.exe -pa scrypt -o fc.altpool.com -p 3333
pause

I was using localhost instead of 127.0.0.1 and I didn't use the -pa scrypt

Will try again later tonight.

Also, the regular stratum proxy doesn't support scrypt, so if you get a message saying that "a" is an illegal port, you are using the bitcoin-only version of the program.

Tobi

That would probably be another of my problems. I downloaded this Stratum when I was on Slush's pool. Where would I find the scrypt version of Stratum?

cbuchner1

hero member

Activity: 756

Merit: 502

in the arguments given to configure, expand CFLAGS and CXXFLAGS with -m64, possibly also the LDFLAGS.
Try forcing it to build 64 bit. Maybe also edit the Makefile to include -m64 in the arguments given to NVCC.

Christian

ffernandex

full member

Activity: 143

Merit: 100

Quote from: Lacan82 on April 30, 2013, 11:51:26 AM

Quote from: ffernandex on April 30, 2013, 11:47:58 AM

Hi, I'm trying to compile cudaMiner for OS X 10.8.3

I'm obtaining the following errors when try to "make",

make all-recursive
Making all in compat
make[3]: Nothing to be done for `all-am'.
g++ -DHAVE_CONFIG_H -I. -pthread -fno-strict-aliasing -msse2 -fopenmp -O3 -MT cudaminer-scrypt.o -MD -MP -MF .deps/cudaminer-scrypt.Tpo -c -o cudaminer-scrypt.o `test -f 'scrypt.cpp' || echo './'`scrypt.cpp
scrypt.cpp:46:20: error: malloc.h: No such file or directory
make[2]: *** [cudaminer-scrypt.o] Error 1
make[1]: *** [all-recursive] Error 1
make: *** [all] Error 2

Any ideas or suggestions? thank you in advance.

it is complaining this is missing: malloc.h: No such file or directory

Thanks, I fixed the missing malloc.h.

Then it complained because it didn't find the folder -L/usr/local/cuda/lib64
I checked and my cuda installation had the drivers in /usr/local/cuda/lib, so I changed the link in the makefile.

Now I'm getting this...

make all-recursive
Making all in compat
make[3]: Nothing to be done for `all-am'.
g++ -O3 -pthread -L/usr/local/cuda/lib -o cudaminer cudaminer-cuda-miner.o cudaminer-util.o cudaminer-sha2.o cudaminer-scrypt.o salsa_kernel.o titan_kernel.o -lcurl -ljansson -lpthread -lcudart -fopenmp
ld: warning: ld: warning: ignoring file salsa_kernel.o, file was built for i386 which is not the architecture being linked (x86_64): salsa_kernel.oignoring file titan_kernel.o, file was built for i386 which is not the architecture being linked (x86_64): titan_kernel.o

I guess I'm reaching my limit here. Is cudaminer only for 32 bit? if yes, how can I compile on my mac?

Thanks for any tips.

cbuchner1

hero member

Activity: 756

Merit: 502

Try out my April-30 release. CPU use might be a bit more stable now. Kepler devices now actually use 99-100% GPU on Windows. Previously the GPU utilization was hovering around 75-90% - for me at least.

Everyone with a Kepler device. Try prefixing your so far best launch config with an S to see if there is any improvement.

Everyone who has been using an S config previously: You're out of luck. The new S kernels now kater to Kepler. Wink

Not sure what kind of performance users of compute 1.x devices can now expect on Windows. Just try.

There is no change for users of Titan cards.

I have been able to get solid 186 kHash out of a 660Ti recently with slight overclock (10% power, +150 MHz core). Driver used was 320.00 Beta. Beats my 560 Ti (but not the 560 Ti 448 Core edition).

I am still sober, so if I borked anything today - there's time left to fix it Wink

Christian

cbuchner1

hero member

Activity: 756

Merit: 502

Quote from: Lacan82 on April 30, 2013, 11:51:26 AM

Quote from: ffernandex on April 30, 2013, 11:47:58 AM

Hi, I'm trying to compile cudaMiner for OS X 10.8.3
scrypt.cpp:46:20: error: malloc.h: No such file or directory
make[2]: *** [cudaminer-scrypt.o] Error 1
make[1]: *** [all-recursive] Error 1
make: *** [all] Error 2
Any ideas or suggestions? thank you in advance.

it is complaining this is missing: malloc.h: No such file or directory

it was never tested on MAC OS X. You're on your own Wink

Lacan82

sr. member

Activity: 247

Merit: 250

Quote from: ffernandex on April 30, 2013, 11:47:58 AM

Hi, I'm trying to compile cudaMiner for OS X 10.8.3

I'm obtaining the following errors when try to "make",

make all-recursive
Making all in compat
make[3]: Nothing to be done for `all-am'.
g++ -DHAVE_CONFIG_H -I. -pthread -fno-strict-aliasing -msse2 -fopenmp -O3 -MT cudaminer-scrypt.o -MD -MP -MF .deps/cudaminer-scrypt.Tpo -c -o cudaminer-scrypt.o `test -f 'scrypt.cpp' || echo './'`scrypt.cpp
scrypt.cpp:46:20: error: malloc.h: No such file or directory
make[2]: *** [cudaminer-scrypt.o] Error 1
make[1]: *** [all-recursive] Error 1
make: *** [all] Error 2

Any ideas or suggestions? thank you in advance.

it is complaining this is missing: malloc.h: No such file or directory

ffernandex

full member

Activity: 143

Merit: 100

Hi, I'm trying to compile cudaMiner for OS X 10.8.3

I'm obtaining the following errors when try to "make",

make all-recursive
Making all in compat
make[3]: Nothing to be done for `all-am'.
g++ -DHAVE_CONFIG_H -I. -pthread -fno-strict-aliasing -msse2 -fopenmp -O3 -MT cudaminer-scrypt.o -MD -MP -MF .deps/cudaminer-scrypt.Tpo -c -o cudaminer-scrypt.o `test -f 'scrypt.cpp' || echo './'`scrypt.cpp
scrypt.cpp:46:20: error: malloc.h: No such file or directory
make[2]: *** [cudaminer-scrypt.o] Error 1
make[1]: *** [all-recursive] Error 1
make: *** [all] Error 2

Any ideas or suggestions? thank you in advance.

Tobi042

newbie

Activity: 16

Merit: 0

Quote from: Stoneysilence on April 29, 2013, 05:11:50 PM

Quote from: sokol815 on April 29, 2013, 04:48:31 PM

sounds like your pool is having problems... I have cudaminer running with stratum_proxy and it is working just fine. Verify that your stratum_proxy is working by testing with cpuminer running on 127.0.0.1:8332

Also important is making sure that stratum proxy starts with -pa scrypt e.g. mine looks like:

Code:

stratum_proxy.exe -pa scrypt -o fc.altpool.com -p 3333
pause

I was using localhost instead of 127.0.0.1 and I didn't use the -pa scrypt

Will try again later tonight.

Also, the regular stratum proxy doesn't support scrypt, so if you get a message saying that "a" is an illegal port, you are using the bitcoin-only version of the program.

Tobi

Stoneysilence

member

Activity: 104

Merit: 10

Installed it on my Laptop Running a GTX 460M

It's doing 52-53Khash/s

Also tried it on my older laptop running a 8600M GT, it got 4-5Khash/s.

EDIT: The GTX 460M was autotuned and running on Win8x64 using the 4-22 miner. The 8600M GT was Win Vistax64 autotuned using the 4-22 miner.

bugilt

member

Activity: 85

Merit: 10

Haven't tweeked any settings yet.

[2013-04-29 21:27:37] GPU #0: GeForce GTX 560, 265216 hashes, 124.22 khash/s
[2013-04-29 21:27:37] GPU #1: GeForce 8800 GT, 34944 hashes, 16.75 khash/s

The BGA on the 8800gt has separated once. Cooked it back together in the over.

Waramp22

sr. member

Activity: 321

Merit: 250

Quote from: Lacan82 on April 29, 2013, 08:39:48 PM

Quote from: Waramp22 on April 29, 2013, 08:29:13 PM

Quote from: Schleicher on April 27, 2013, 06:12:38 PM

Quote from: Waramp22 on April 27, 2013, 01:24:04 PM

I dont see the "No Thanks" option anywhere. Im either blind or retarded.

Screenshot:
https://dl.dropboxusercontent.com/u/34410241/mega-co-nz-screenshot.jpg

Hmm weird. This is what i get. http://i.imgur.com/h73IGa5.jpg

Nothing wrong, You're using IE. Please Upgrade your browser Tongue

But its on my mining machine, the less crap i get on here the less i will want to use it for everyday things. Thats what the laptop is for. Welp, Firefox it is.

termhn

full member

Activity: 126

Merit: 100

Quote from: Lacan82 on April 29, 2013, 08:39:48 PM

Quote from: Waramp22 on April 29, 2013, 08:29:13 PM

Quote from: Schleicher on April 27, 2013, 06:12:38 PM

Quote from: Waramp22 on April 27, 2013, 01:24:04 PM

I dont see the "No Thanks" option anywhere. Im either blind or retarded.

Screenshot:
https://dl.dropboxusercontent.com/u/34410241/mega-co-nz-screenshot.jpg

Hmm weird. This is what i get. http://i.imgur.com/h73IGa5.jpg

Nothing wrong, You're using IE. Please Upgrade your browser Tongue

+1

Lacan82

sr. member

Activity: 247

Merit: 250

Quote from: Waramp22 on April 29, 2013, 08:29:13 PM

Quote from: Schleicher on April 27, 2013, 06:12:38 PM

Quote from: Waramp22 on April 27, 2013, 01:24:04 PM

I dont see the "No Thanks" option anywhere. Im either blind or retarded.

Screenshot:
https://dl.dropboxusercontent.com/u/34410241/mega-co-nz-screenshot.jpg

Hmm weird. This is what i get. http://i.imgur.com/h73IGa5.jpg

Nothing wrong, You're using IE. Please Upgrade your browser Tongue

Waramp22

sr. member

Activity: 321

Merit: 250

Quote from: Schleicher on April 27, 2013, 06:12:38 PM

Quote from: Waramp22 on April 27, 2013, 01:24:04 PM

I dont see the "No Thanks" option anywhere. Im either blind or retarded.

Screenshot:
https://dl.dropboxusercontent.com/u/34410241/mega-co-nz-screenshot.jpg

Hmm weird. This is what i get. http://i.imgur.com/h73IGa5.jpg

sokol815

member

Activity: 76

Merit: 10

Quote from: Stoneysilence on April 29, 2013, 05:11:50 PM

Quote from: sokol815 on April 29, 2013, 04:48:31 PM

sounds like your pool is having problems... I have cudaminer running with stratum_proxy and it is working just fine. Verify that your stratum_proxy is working by testing with cpuminer running on 127.0.0.1:8332

Also important is making sure that stratum proxy starts with -pa scrypt e.g. mine looks like:

Code:

stratum_proxy.exe -pa scrypt -o fc.altpool.com -p 3333
pause

I was using localhost instead of 127.0.0.1 and I didn't use the -pa scrypt

Will try again later tonight.

That sounds like it was the problem... cool. Good luck!

Topic: [ANN] cudaMiner & ccMiner CUDA based mining applications [Windows/Linux/MacOSX] - page 1104. (Read 3426947 times)