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

[cbuchner1]

Would a scrypt-jane version of cudaminer that utilizes SLI make sense?

SLI is strictly for graphics. It literally means "scan line interleave". Anything computation and CUDA related doesn't have scanlines, so the whole SLI concept is moot.

CUDA sees two SLI'd cards as two separate CUDA devices. You can run two cudaminer instances, specifying -d 0 on the first one and -d 1 on the second one. Instant double hash rate.

[1713946420]

1713946420

[1713946420]

1713946420

[1713946420]

1713946420

[1713946420]

1713946420

[cbuchner1]

Heh ok, can't mine with that overclock! Just got myself an GTX 780 which is running nicely at 1250mhz core and currently 1600mhz memory, skynet bios etcetc. Discovered what you did earlier, power draw is off the charts hahah. It wants more even at 125% TDP and that's when the 100% is boosted to 300W+ on this bios >_> Did give me 555k/hash+ though! xD Will have to tweak it a bit  

these mod bioses are made for overvolting - and the stock cooler can't take it. You need water cooling or extra strong aftermarket heat sinks + fans

Christian

[69charger]

Would a scrypt-jane version of cudaminer that utilizes SLI make sense?

SLI is strictly for graphics. It literally means "scan line interleave". Anything computation and CUDA related doesn't have scanlines, so the whole SLI concept is moot.

CUDA sees two SLI'd cards as two separate CUDA devices. You can run two cudaminer instances, specifying -d 0 on the first one and -d 1 on the second one. Instant double hash rate.

I have SLI off for mining. Still having an issue with my first card not wanting to cooperate in Windows 7. The -d 1 card is hashing scrypt-jane just fine (-a scrypt-jane -d 1 -i 0 -l K14x1 -C 0). Cant seem to get the -d 0 card to do much even with separate instances for both.

[Killyx]

hi are there ways to stop cudaminer from using all your gpu power?
like it can only use like 50 percent because I dont want to have an fully exhausted gpu all the time

[bathrobehero]

Am I doing it wrong on a GTX 660 2GB?

-a scrypt-jane -d 1 -i 0 -l K10x1 -C 2

Getting 0.48khash/s anything higher than 10x1 gives me the too much memory error.

Don't use -C 2, use -C 0.

Also, you should try K14x1 or if it gives you out of memory error (due to some background applications using VRAM) go for K13x1 for optimum performance.

[69charger]

Am I doing it wrong on a GTX 660 2GB?

-a scrypt-jane -d 1 -i 0 -l K10x1 -C 2

Getting 0.48khash/s anything higher than 10x1 gives me the too much memory error.

Don't use -C 2, use -C 0.

Also, you should try K14x1 or if it gives you out of memory error (due to some background applications using VRAM) go for K13x1 for optimum performance.

Yeah I followed your advice a few posts back and it worked for my second card. Running on igpu to try to free both cards up for mining. Anybody try that and make it work?

[cbuchner1]

erm... WHAT?

Code:

[2014-01-11 01:03:31] 1 miner threads started, using 'scrypt' algorithm.
[2014-01-11 01:03:48] GPU #0: GeForce GTX 660 with compute capability 2.1
[2014-01-11 01:03:48] GPU #0: interactive: 1, tex-cache: 0 , single-alloc: 0
[2014-01-11 01:03:48] GPU #0: Performing auto-tuning (Patience...)
[2014-01-11 01:03:48] GPU #0: maximum warps: 992

that GTX 660 OEM card with 4GB video RAM is a rebranded Fermi part or what?

[bathrobehero]

erm... WHAT?

Code:

[2014-01-11 01:03:31] 1 miner threads started, using 'scrypt' algorithm.
[2014-01-11 01:03:48] GPU #0: GeForce GTX 660 with compute capability 2.1
[2014-01-11 01:03:48] GPU #0: interactive: 1, tex-cache: 0 , single-alloc: 0
[2014-01-11 01:03:48] GPU #0: Performing auto-tuning (Patience...)
[2014-01-11 01:03:48] GPU #0: maximum warps: 992

that GTX 660 OEM card with 4GB video RAM is a rebranded Fermi part or what?

...yesterday I though I was unlucky that I couldn't find them anywhere to buy (around UK).

[trell0z]

Heh ok, can't mine with that overclock! Just got myself an GTX 780 which is running nicely at 1250mhz core and currently 1600mhz memory, skynet bios etcetc. Discovered what you did earlier, power draw is off the charts hahah. It wants more even at 125% TDP and that's when the 100% is boosted to 300W+ on this bios >_> Did give me 555k/hash+ though! xD Will have to tweak it a bit  

these mod bioses are made for overvolting - and the stock cooler can't take it. You need water cooling or extra strong aftermarket heat sinks + fans

Christian

Ah ye forgot to mention, it's an gigabyte WF, so the cooling isn't a problem. It's the power draw ^^ Way crazy. Gonna see if I can manage around 500khash/s at a less dangerous power draw

[cbuchner1]

erm... WHAT?

...yesterday I though I was unlucky that I couldn't find them anywhere to buy (around UK).

buchner@miner2:~/CudaMiner$ ./cudaminer --algo=scrypt-jane:14 -l 15x2 -i 0 -C 2 --benchmark

Code:

[2014-01-11 01:40:32] Nfactor is 14 (N=32768)!
[2014-01-11 01:40:32] GPU #0: GeForce GTX 660 with compute capability 2.1
[2014-01-11 01:40:32] GPU #0: interactive: 0, tex-cache: 2D, single-alloc: 1
[2014-01-11 01:40:32] GPU #0: using launch configuration F15x2
[2014-01-11 01:40:35] GPU #0: GeForce GTX 660, 1.45 khash/s
[2014-01-11 01:40:35] Total: 1.45 khash/s

well... not completely worthless... and a good excuse for optimizing the Fermi kernels more...

[bathrobehero]

Ah ye forgot to mention, it's an gigabyte WF, so the cooling isn't a problem. It's the power draw ^^ Way crazy. Gonna see if I can manage around 500khash/s at a less dangerous power draw

You can dowclock the memory pretty heavily for scrypt and could try undervolt the GPU.

[manofcolombia]

well with K13x1 my 660 ti gets 1.33 khash on windows...I'm really tempted to throw another drive in my rig and put mint on it and see the difference but I've only got until Sunday before I gotta drive back to school so I'm just gonna be relaxing haha.

Right now I'm in a pool. How would I solo mine Yacoin?...I wanna test my luck
And as far as overclocking goes, memory OC if anything for jane? But doubt that will help much since its more mem capacity and the way windows is utilizing it correct?

[cbuchner1]

well with K13x1 my 660 ti gets 1.33 khash on windows...I'm really tempted to throw another drive in my rig and put mint on it and see the difference but I've only got until Sunday before I gotta drive back to school so I'm just gonna be relaxing haha.

I use -l K4x4 -b 8192 -C 1 -i 0 and get 2.65 Khash but I also overclock...

[manofcolombia]

well with K13x1 my 660 ti gets 1.33 khash on windows...I'm really tempted to throw another drive in my rig and put mint on it and see the difference but I've only got until Sunday before I gotta drive back to school so I'm just gonna be relaxing haha.

I use -l K4x4 -b 8192 -C 1 -i 0 and get 2.65 Khash but I also overclock...

K5x2 with -b 8192 gives me 1.80 khash anything higher crashes the drivers
What specifically do you focus your OC on? Memory?

EDIT: Actually getting a few that dont validate with the cpu

[cdogster]

Using the recently linked to 32bit Windows build I am seeing 0.64 kh/s per each of my GTX 580 cards.

Code:

cudaminer.exe -a scrypt-jane -i 1 -C 2 -l F10x1 -o http://yac.coinmine.pl:8882 -O cdogster.gtx580x2:123
           *** CudaMiner for nVidia GPUs by Christian Buchner ***
                     This is version 2013-12-18 (beta)
        based on pooler-cpuminer 2.3.2 (c) 2010 Jeff Garzik, 2012 pooler
               Cuda additions Copyright 2013 Christian Buchner
           My donation address: LKS1WDKGED647msBQfLBHV3Ls8sveGncnm

[2014-01-10 21:18:59] 2 miner threads started, using 'scrypt-jane' algorithm.
[2014-01-10 21:18:59] Long-polling activated for http://yac.coinmine.pl:8882/LP
[2014-01-10 21:18:59] Current Nfactor is 14 (N=32768)!
[2014-01-10 21:19:00] GPU #0: GeForce GTX 580 with compute capability 2.0
[2014-01-10 21:19:00] GPU #0: interactive: 1, tex-cache: 2D, single-alloc: 1
[2014-01-10 21:19:00] GPU #0: using launch configuration F10x1
[2014-01-10 21:19:00] GPU #1: GeForce GTX 580 with compute capability 2.0
[2014-01-10 21:19:00] GPU #1: interactive: 1, tex-cache: 2D, single-alloc: 1
[2014-01-10 21:19:00] GPU #1: using launch configuration F10x1
[2014-01-10 21:19:06] GPU #1: GeForce GTX 580, 0.64 khash/s
[2014-01-10 21:19:06] GPU #0: GeForce GTX 580, 0.64 khash/s
[2014-01-10 21:19:13] LONGPOLL detected new block
[2014-01-10 21:19:13] GPU #1: GeForce GTX 580, 0.64 khash/s
[2014-01-10 21:19:13] GPU #0: GeForce GTX 580, 0.64 khash/s
[2014-01-10 21:19:54] GPU #0: GeForce GTX 580 result does not validate on CPU (i=15, s=1)!
[2014-01-10 21:19:56] LONGPOLL detected new block
[2014-01-10 21:19:57] GPU #1: GeForce GTX 580, 0.68 khash/s
[2014-01-10 21:19:57] GPU #0: GeForce GTX 580, 0.68 khash/s
[2014-01-10 21:19:58] LONGPOLL detected new block

Using my GTX 650 I get around 1.08 kh/s and a lot more yays!

Code:

cudaminer.exe -a scrypt-jane -i 1 -C 2 -o http://yac.coinmine.pl:8882 -O cdogster.gtx650x1:123
           *** CudaMiner for nVidia GPUs by Christian Buchner ***
                     This is version 2013-12-18 (beta)
        based on pooler-cpuminer 2.3.2 (c) 2010 Jeff Garzik, 2012 pooler
               Cuda additions Copyright 2013 Christian Buchner
           My donation address: LKS1WDKGED647msBQfLBHV3Ls8sveGncnm

[2014-01-10 21:44:36] 1 miner threads started, using 'scrypt-jane' algorithm.
[2014-01-10 21:44:37] Long-polling activated for http://yac.coinmine.pl:8882/LP
[2014-01-10 21:44:37] Current Nfactor is 14 (N=32768)!
[2014-01-10 21:44:37] GPU #0: GeForce GTX 650 with compute capability 3.0
[2014-01-10 21:44:37] GPU #0: interactive: 1, tex-cache: 2D, single-alloc: 1
[2014-01-10 21:44:37] GPU #0: Performing auto-tuning (Patience...)
[2014-01-10 21:44:37] GPU #0: maximum warps: 9
[2014-01-10 21:44:50] GPU #0:    1.27 khash/s with configuration K9x1
[2014-01-10 21:44:50] GPU #0: using launch configuration K9x1
[2014-01-10 21:44:54] GPU #0: GeForce GTX 650, 0.24 khash/s
[2014-01-10 21:45:02] LONGPOLL detected new block
[2014-01-10 21:45:02] GPU #0: GeForce GTX 650, 1.06 khash/s
[2014-01-10 21:45:28] GPU #0: GeForce GTX 650, 1.08 khash/s
[2014-01-10 21:45:28] accepted: 1/1 (100.00%), 1.08 khash/s (yay!!!)
[2014-01-10 21:45:53] GPU #0: GeForce GTX 650, 1.07 khash/s
[2014-01-10 21:45:53] accepted: 2/2 (100.00%), 1.07 khash/s (yay!!!)
[2014-01-10 21:45:57] LONGPOLL detected new block
[2014-01-10 21:45:57] GPU #0: GeForce GTX 650, 1.02 khash/s
[2014-01-10 21:46:09] LONGPOLL detected new block
[2014-01-10 21:46:09] GPU #0: GeForce GTX 650, 1.07 khash/s
[2014-01-10 21:46:16] GPU #0: GeForce GTX 650, 1.05 khash/s
[2014-01-10 21:46:16] accepted: 3/3 (100.00%), 1.05 khash/s (yay!!!)
[2014-01-10 21:46:40] LONGPOLL detected new block
[2014-01-10 21:46:40] GPU #0: GeForce GTX 650, 1.08 khash/s
[2014-01-10 21:47:04] LONGPOLL detected new block
[2014-01-10 21:47:04] GPU #0: GeForce GTX 650, 1.08 khash/s
[2014-01-10 21:48:00] LONGPOLL detected new block
[2014-01-10 21:48:00] GPU #0: GeForce GTX 650, 1.09 khash/s
[2014-01-10 21:48:43] LONGPOLL detected new block
[2014-01-10 21:48:43] GPU #0: GeForce GTX 650, 1.09 khash/s
[2014-01-10 21:49:43] GPU #0: GeForce GTX 650, 1.09 khash/s
[2014-01-10 21:49:56] LONGPOLL detected new block
[2014-01-10 21:49:56] GPU #0: GeForce GTX 650, 1.07 khash/s
[2014-01-10 21:50:39] LONGPOLL detected new block
[2014-01-10 21:50:39] GPU #0: GeForce GTX 650, 1.09 khash/s
[2014-01-10 21:51:39] GPU #0: GeForce GTX 650, 1.09 khash/s
[2014-01-10 21:52:35] LONGPOLL detected new block
[2014-01-10 21:52:35] GPU #0: GeForce GTX 650, 1.09 khash/s
[2014-01-10 21:53:28] LONGPOLL detected new block
[2014-01-10 21:53:28] GPU #0: GeForce GTX 650, 1.09 khash/s
[2014-01-10 21:53:56] GPU #0: GeForce GTX 650, 1.08 khash/s
[2014-01-10 21:53:56] accepted: 4/4 (100.00%), 1.08 khash/s (yay!!!)
[2014-01-10 21:54:04] GPU #0: GeForce GTX 650, 1.06 khash/s
[2014-01-10 21:54:04] accepted: 5/5 (100.00%), 1.06 khash/s (yay!!!)

Is there anything wrong with my configurations that I could improve?

[bigjme]

Did you get the full rig running Christian?
Does -C do much or anything on the titan kernal with scrypt-jane?

Manofcolombia
I would try linux. I had mint 16 and that was the one that died because it was using ubuntu 13.10. My ubuntu 12.04 LTS is rock solid. Compiles easy and runs well. Plus the huge spike in performance and it uses the memory better :-)

I would use a decebt hard drive though. My yacoin wallet can take 15 minutes to load sometimes. And that is loading off a 550MB/s read and write ssd

Only uses 1 thread to load though. Be nice if I could tell it to use all 8

[bathrobehero]

Anybody has a reference of how well any 7xx series 4GB card performs in scrypt-jane?

[bigjme]

My 780 3gb gets 3.77khash/s using ubuntu 12.04
As gpu performance doesnt play a huge part I would say maybe a little more then my 780?

What 700 series card are you looking at?

[bathrobehero]

My 780 3gb gets 3.77khash/s using ubuntu 12.04
As gpu performance doesnt play a huge part I would say maybe a little more then my 780?

What 700 series card are you looking at?

Thinking about getting some 4GB 760's.

[bigjme]

I would say perhaps 4khash/s under linux if you can get it to use all the memory. I believe memory bandwidth is starting to become an issue with over 3gb cards.

Im sure someone posted the result on their 760 the other day but I think it was a 2gb version.

What overclocks are we all using to get better scrytp-jane performance? As clock speed isnt a big factor. And I dont see memory frewuency playing a big part as it is mainly down to how much it is given?

If memory clock speeds do make a difference then a 760 could get a very good score. I would reccomend trying the cpu out aswell :-) as mine together get over 4.4khash/s