[ANN] cudaMiner & ccMiner CUDA based mining applications [Windows/Linux/MacOSX]

[peri]

on 2x680's gone from 500khs to around 730 !!

http://s5.postimg.org/5n4bpay7b/18_12_2013_12_48_15.jpg


EDIT

on the 480 in another machine its gone from 230 to

"unable to query CUDA driver version ! - Is an nvidia driver installed?"

[1701176156]

1701176156

[cbuchner1]

But bitcoin and litecoin crashed hard today. Meh, this hardly pays for the electricity bill anymore...

[cbuchner1]

on the 480 in another machine its gone from 230 to

"unable to query CUDA driver version ! - Is an nvidia driver installed?"

driver update, maybe?

[peri]

oops......probably, its the wife's machine, I forgot about that.

On the 2x680's I'm still getting the same 'cudaminer has stopped working' if I try any config except auto or k8x24.

Auto tells me 20x12 should be best, but manually entering that just crashes.

cudaminer-2013-12-10\x86>cudaminer.exe -D -H 1 -i 0 -d 0,1 -l k20x12
 -m 1 -C 1 -o stratum+tcp://ltc.give-me-coins.com:3333 -O


[2013-12-18 13:06:18] Starting Stratum on stratum+tcp://ltc.give-me-coins.com:3333
[2013-12-18 13:06:18] 2 miner threads started, using 'scrypt' algorithm.
[2013-12-18 13:06:18] Stratum session id: 50298b34787f00001
[2013-12-18 13:06:18] Stratum difficulty set to 16
[2013-12-18 13:06:19] DEBUG: job_id='1387371954 4569' extranonce2=00000000 ntime=52b19db2
[2013-12-18 13:06:19] GPU #1: GeForce GTX 680 with compute capability 3.0
[2013-12-18 13:06:19] GPU #0: GeForce GTX 680 with compute capability 3.0
[2013-12-18 13:06:21] DEBUG: job_id='1387371988 4570' extranonce2=00000000 ntime=52b19dd4
[2013-12-18 13:06:21] Stratum detected new block

Program crash


Not a great problem, just thought you might like to know


EDIT: How's this for wierd? - although manually entering the 8x24 config worked twice, after trying 'auto', even that now causes the program to 'stop working'. Tried a reboot, and that didn't help - now only 'auto' works?

[mahowi]

Hi!

I'm trying to run cudaminer with a script to use other pools as fallback if primary pool is offline.

Code:

@echo off

:Global
:: Set options - Change Retries, seconds and pool/usernames to those you desire. Make sure to include port numbers

SET retries=1
SET seconds=5

::Change TTL to how long you want to be at the secondary/tertiary pool before trying the primary. Time in seconds.

SET TTL=60

:If you would like to a specific amount of threads, remove the ":" from the next line and specify. Add "--threads=%threads%" to the end of each minerd line.

:SET threads=X

:Primary Pool

SET userpass1=user.pass
SET pool1=stratum+tcp://stratum01.hashco.ws:8888

:Secondary Pool

SET userpass2=user.pass
SET pool2=stratum+tcp://middlecoin.com:3333


:Tertiary Pool

SET userpass3=user.pass
SET pool3=stratum+tcp://pool1.eu.multipool.us:7777

::Actually mining part
:MINE

START /B /wait cudaminer.exe -i 0 -H1 -m1 -l F64x2 --retries=%RETRIES% --retry-pause=%SECONDS% --userpass=%USERPASS1% --url=%pool1%

START /B cudaminer.exe -i 0 -H1 -m1 -l F64x2 --retries=%RETRIES% --retry-pause=%SECONDS% --userpass=%USERPASS2% --url=%pool2%

tasklist /FI "IMAGENAME eq cudaminer.exe" 2>NUL | find /I /N "cudaminer.exe">NUL
if "%ERRORLEVEL%"=="0" PING 1.1.1.1 -n 1 -w %TTL% >NUL

TASKKILL /F /IM "cudaminer.exe"

START /B /wait cudaminer.exe -i 0 -H1 -m1 -l F64x2 --retries=%RETRIES% --retry-pause=%SECONDS% --userpass=%USERPASS1% --url=%pool1%

START /B cudaminer.exe -i 0 --H1 -m1 -l F64x2 --retries=%RETRIES% --retry-pause=%SECONDS% --userpass=%USERPASS3% --url=%pool3%

tasklist /FI "IMAGENAME eq cudaminer.exe" 2>NUL | find /I /N "cudaminer.exe">NUL
IF "%ERRORLEVEL%"=="0" TIMEOUT /T %TTL%

TASKKILL /F /IM "cudaminer.exe"

GOTO MINE

:END

I have found this script in some other forum and it works with cpuminer. But with cudaminer, if primary pool is unreachable, it stops with

Code:

workio thread dead, waiting for workers...

and cudaminer doesn't terminate. So the script can't switch over to the next pool.
It seems cudaminer doesn't shutdown all worker threads.

[SavellM]

I haven't seen any increase with my Titans
Infact the auto goes slower than my old config...
But using my config it goes the same speed...

[trell0z]

Oh and also auto with debug on seems to fail a ton on the 670 now, gives some 1&jhash /similar error config then starts hashing at <20khs. Without debug it's a really low config like 2x1 6x4 or similar, which also sucks.

[dga]

Looks like my GTX580 doesn't like the latest build (both x64 and x86). For some reason the core temperature increases with ~8 degrees, while performing slightly slower. GPU load is 99% all the time.

Just to be clear, the GTX 580 is a compute 2.0 device ("Fermi"), not a Kepler, so what you're seeing may be something else.  Is the change, for example, due to moving the other hashing from the CPU to the GPU?  (is your CPU now almost completely idle when it wasn't before?)

I'm a little more worried about the person who said his Titan isn't running any faster.

Did the module get properly compiled for architecture sm_35?

What config are you using and what speed are you getting on a single card?

[cbuchner1]

oops......probably, its the wife's machine, I forgot about that.

On the 2x680's I'm still getting the same 'cudaminer has stopped working' if I try any config except auto or k8x24.

Auto tells me 20x12 should be best, but manually entering that just crashes.

Do you enter this launch config with an upper-case K ? lower case won't work...

[cbuchner1]

I'm a little more worried about the person who said his Titan isn't running any faster.

Did the module get properly compiled for architecture sm_35?

My OC 780Ti went from 480 kHash/s stable to 560 kHash/s stable. Using x86 Windows binary, and using the Titan kernel of course. I think I had to autotune again and try a few different settings. The one I use now was posted a little earlier.

In a non-verclocked Linux based mining rig my 780Ti went from ~430 kHash/s stable to around 500 kHash/s stable. Sweet. Would someone know how to OC a card on Ubuntu 12.04 without running an x-server on the card? nvidia-smi isn't helpful and other tools refuse to run if the nVidia is not the display card. Meh.

[cbuchner1]

Hi!
I have found this script in some other forum and it works with cpuminer. But with cudaminer, if primary pool is unreachable, it stops with

Code:

workio thread dead, waiting for workers...

and cudaminer doesn't terminate. So the script can't switch over to the next pool.
It seems cudaminer doesn't shutdown all worker threads.

This would be a bug related to stratum. Does it terminate when you try the getwork protocol?

[peri]

Do you enter this launch config with an upper-case K ? lower case won't work...

doh !  I really am a knob at times 

[danws6]

Wow, what an update! Two EVGA 770's with 4GB ram each, went from ~440 KH/s combined to ~580 KH/s

cudaminer.exe -H 1 -d 0,1 -i 0 -l K12x17

I use evga's precision control to drop the power target to 80% to keep the cards at about 73C to avoid system crashes.

Thanks for the hard work!

[SavellM]

I'm a little more worried about the person who said his Titan isn't running any faster.

Did the module get properly compiled for architecture sm_35?

My OC 780Ti went from 480 kHash/s stable to 560 kHash/s stable. Using x86 Windows binary, and using the Titan kernel of course. I think I had to autotune again and try a few different settings. The one I use now was posted a little earlier.

In a non-verclocked Linux based mining rig my 780Ti went from ~430 kHash/s stable to around 500 kHash/s stable. Sweet. Would someone know how to OC a card on Ubuntu 12.04 without running an x-server on the card? nvidia-smi isn't helpful and other tools refuse to run if the nVidia is not the display card. Meh.

Hey Christian,

Is there somewhere we can talk about my Titans?
Before using 12-10 I was getting 465-475KH/s per Titan, using config T112x4
With the 12-18 on Auto it was showing 1040KH/s during the config but then started mining and is only 440-450KH/s.
If I use my T112x4 it goes up to 470-480KH/s.

My whole config is:
-i 0 -m 1 -H 2 -l auto

Also when it starts its says: GTX TITAN with compute capability 3.5 - Is this correct?

Any ideas?

[cbuchner1]

Hey Christian,

Is there somewhere we can talk about my Titans?

We're not talking about an STD, drugs, or an extra kinky fetish so I guess we can talk _right here_.

Before using 12-10 I was getting 465-475KH/s per Titan, using config T112x4
With the 12-18 on Auto it was showing 1040KH/s during the config but then started mining and is only 440-450KH/s.
If I use my T112x4 it goes up to 470-480KH/s.

My whole config is:
-i 0 -m 1 -H 2 -l auto

Also when it starts its says: GTX TITAN with compute capability 3.5 - Is this correct?

Any ideas?

if I use  -l T15x24  with good results on 780Ti with 15 SMX you should be able to use -l T14x24 on the Titan. I believe it has 14 SMX enabled, whereas the GTX 780 has 12 SMX.

Also be sure the card is not configured for double precision computation (it makes it slower for single precision work loads)

I also had that one bizarre autotune result with 1000 kHash/s, but that was too good to be true... and in fact it wasn't working for me either.

[SavellM]

Hehe but I just love to talk about STD's

T14x24 = 465-470KH/s.

Everytime I do an auto tune I get like 900-1050KH/s reported lol... like everytime...

I have tried on multiple titans and dont get much more than 480KH/s

"Also be sure the card is not configured for double precision computation (it makes it slower for single precision work loads)"
In the startup it say: single-alloc: 1 is that what you are referring to?

[mahowi]

Hi!
I have found this script in some other forum and it works with cpuminer. But with cudaminer, if primary pool is unreachable, it stops with

Code:

workio thread dead, waiting for workers...

and cudaminer doesn't terminate. So the script can't switch over to the next pool.
It seems cudaminer doesn't shutdown all worker threads.

This would be a bug related to stratum. Does it terminate when you try the getwork protocol?

I don't know how to test this. The only server having problems with downtimes is hashcows and it uses only stratum protocol.

[cbuchner1]

"Also be sure the card is not configured for double precision computation (it makes it slower for single precision work loads)"
In the startup it say: single-alloc: 1 is that what you are referring to?

I found a blurb on the net about the single/double switching feature (I don't have it on GTX 780Ti, so this is hearsay):

Double precision wise, to unlock full performance you must open the NVIDIA Control Panel, navigate to “Manage 3D Settings”. In the Global Settings box you will find an option titled “CUDA – Double Precision”, but... GeForce GTX Titan runs at reduced clock speeds when full double-precision is enabled. Still a great option if you are working on CUDA applications.

Also, if you're really into talking about STDs: http://www.healthboards.com/boards/sexually-transmitted-diseases/

[Ness]

Did a bit more testing, had to back off my overclock TDP since temps were high, and I found that K14x16 seems to be the optimum config so far for my 670 FTWs. 630-640 kh/s range combined on my 2 cards.

[SavellM]

HAHA no not into STD's...

Ok so Double was not in use.

I have just enabled it as a test to check it out.
Will report back asap.

*EDIT* OK with Double enabled it drops to 420KH/s.

Wonder what I'm missing to get better performance.