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

[lagur]

Here is my hashpower on HVC with two 660TI's

Code:

[2014-04-01 15:17:49]
[2014-04-01 15:20:33] accepted: 3798/3799 (99.97%), 20303 khash/s (yay!!!)
[2014-04-01 15:20:34] thread 0: 80886912 hashes, 10511 khash/s
[2014-04-01 15:20:34] accepted: 3799/3800 (99.97%), 21317 khash/s (yay!!!)
[2014-04-01 15:20:34] thread 1: 14671833 hashes, 10202 khash/s
[2014-04-01 15:20:34] accepted: 3800/3801 (99.97%), 20713 khash/s (yay!!!)
[2014-04-01 15:20:35] thread 1: 9452806 hashes, 10219 khash/s
[2014-04-01 15:20:35] accepted: 3801/3802 (99.97%), 20730 khash/s (yay!!!)

If you have MSI 660Ti than you can add 1Mh/s and get ~11500 khash/s from 1 card. If you have open case then can add (+150 core, +114 power) in AfterBurner.
Total profit for 2x660ti will be 2Mh/s -> ~23000 khash/s.
My result for 1xMSI 660Ti and 2xMSI 750Ti:

What's the clock setting of your 750ti?
Nice hashrate btw.

Afterburner
Core Clock: +250, Memory: -502, Fan speed: fixed 40%
Open case -> gpu temperature: 40-43C

Unlocked bios?

[bigjme]

Unlocked bios?

the MSI cards naturally go very high, mine is at +113 clock and has a LOT of headroom, they let you drag the clock really high, and the same on the power

[lagur]

Unlocked bios?

the MSI cards naturally go very high, mine is at +113 clock and has a LOT of headroom, they let you drag the clock really high, and the same on the power

Well I'm using non-MSI which can only go +175 clock but anything higher will cause miner to crash or gpu to idle.

[bigjme]

Which make? someone else may be able to help.
I know the MSI ones will go to 1400MHz without even increasing the power rating, and that's i believe +213MHz on the clock
haven't tried it myself, but i know someone who did. Right now i am getting 39MH/s on my 3 750Ti's with the clock at +113MHz
so the clock rate doesn't really need to go very high to achieve 13MH/s a card

[@msteramned]

Which make? someone else may be able to help.
I know the MSI ones will go to 1400MHz without even increasing the power rating, and that's i believe +213MHz on the clock
haven't tried it myself, but i know someone who did. Right now i am getting 39MH/s on my 3 750Ti's with the clock at +113MHz
so the clock rate doesn't really need to go very high to achieve 13MH/s a card

Some of them are overclocked by the factory so have less headroom for further overclocking I imagine.

I have a Gainward and a Gigabyte and the Gainward can't be dialed up as high as the Gigabyte. Gigabyte is at 1301mhz (+125 overclock) and the Gainward is at 1317mhz (+89mhz).

[bigjme]

I think mine are the overclock edition yeh. At stock their at 1201MHz which is pretty high

[@msteramned]

How do I run two instances? Just two separate .bat files and drop the -t 2 command from each or is something else needed?

My bat files look like this
ccminer-v0.5\x86\ccminer35.exe -a heavy -q -d 1 -o stratum+tcp://hvcpool.1gh.com:5333 -u wallet -p x -v 512

the -d tells it which card to use, and i just run the bat twice. My cards are all on usb risers, and you can see a rough idea of performance increase in my above post, obviously it is on HVC but gives some idea of increases

Hey thats great thanks man, got 2 x 12mh/s now so I'm already up to 2mh/s! So does it show up as two miners or as one? My pool doesn't have any individual miner stats so can't tell...

[zelante]

Afterburner
Core Clock: +250, Memory: -502, Fan speed: fixed 40%
Open case -> gpu temperature: 40-43C

what clock are your 750Ti's sitting at? the 1400MHz mark?

GPU-Z -> Sensors -> GPU Core Clock: 1438,9 MHz

[cbuchner1]

no, 2 separate instances. -d 0,0 reduces performance.

first and foremost -d 0,0 breaks currently because it operates both threads in the same CUDA context,
and you will get some non-validating hashes (the threads will step on each other's toes). This is fixable,
but only with an update to ccminer.

With two GPUs, I'd recommend to run a bat file with -d 0,1 twice on the same PC.  The total hash rate
of both instances would be more than the hash rate of a single instance running - mainly because of
higher overall GPU utilization.

[bigjme]

Right now, i have 2 instances for each card to get the best performance (6 instances in total)
If i run all 3 devices per instance i lose around 4MH/s

[@msteramned]

no, 2 separate instances. -d 0,0 reduces performance.

first and foremost -d 0,0 breaks currently because it operates both threads in the same CUDA context,
and you will get some non-validating hashes (the threads will step on each other's toes). This is fixable,
but only with an update to ccminer.

With two GPUs, I'd recommend to run a bat file with -d 0,1 twice on the same PC.  The total hash rate
of both instances would be more than the hash rate of a single instance running - mainly because of
higher overall GPU utilization.

Thanks for replying. So I shouldn't run one bat with -d 0 and the other with -d 1 but rather two bats with -d 0,1 in each...so each is identical?

[bigjme]

If you have 2 devices then you use
-d 0,1

And run that bat twice.
I personally find i get better performance from running all my cards individually, so for 2 cards you have 4 ccminers open. But it is a lot more cpu heavy. My system has 36% cpu usage from 6 instances, and its a 4 core, 8 thread

[zelante]

I have two same msi 750ti.
For each of them i start one bat file for start ccminer.
For restart one of them (not all by devcon64 disable *dev_1380) I find solution.

1. Get full string of device

Code:

C:\Mining\ccminer-v0.5-advanced-startscript-1.0\x86>devcon64 status *1380
   PCI\VEN_10DE&DEV_1380&SUBSYS_31021462&REV_A2\4&C9C687&0&0009
    Name: NVIDIA GeForce GTX 750 Ti
    Driver is running.
   PCI\VEN_10DE&DEV_1380&SUBSYS_31021462&REV_A2\4&20BDBE28&0&00E0
    Name: NVIDIA GeForce GTX 750 Ti
    Driver is running.
2 matching device(s) found.

2. For select first 750ti card i use : devcon64 disable @pci\*1380*\*0009
   For select second 750ti card i use : devcon64 disable @pci\*1380*\*00E0

3. Now you can run one bat for 1 card:

ccminer35.exe -d 0 (i'm use for 660ti)

For first 750ti card:
ccminer35.exe -d 1
devcon64 disable @pci\*1380*\*0009
devcon64 enable @pci\*1380*\*0009

For second 750ti card:
ccminer35.exe -d 2
devcon64 disable @pci\*1380*\*00E0
devcon64 enable @pci\*1380*\*00E0


So 1 instance of ccminer for 1 card. If 1 instance crashed it's restarting and all others still working.

[atc1]

@cbuchner1

Is there a current project to optimize cuda for the X11 algo? Most new coins are run on it and it increasingly looks to be the alternative to scrypt. As such,AMD holds a strong advantage in mining power and outperforms Nvidia by a massive margin. I believe there's a bounty for such a miner as well.

[djm34]

@cbuchner1

Is there a current project to optimize cuda for the X11 algo? Most new coins are run on it and it increasingly looks to be the alternative to scrypt. As such,AMD holds a strong advantage in mining power and outperforms Nvidia by a massive margin. I believe there's a bounty for such a miner as well.

most new coin... I see two coins using it and a nice propaganda so that it gets used by more coins...

[atc1]

@cbuchner1

Is there a current project to optimize cuda for the X11 algo? Most new coins are run on it and it increasingly looks to be the alternative to scrypt. As such,AMD holds a strong advantage in mining power and outperforms Nvidia by a massive margin. I believe there's a bounty for such a miner as well.

most new coin... I see two coins using it and a nice propaganda so that it gets used by more coins...

Check up on the new pre-anns. Way more than 2 coins. Agreed that they're new and will prolly be worthless,but still it's usage is gaining ground and getting popular.And they were planning on forking Litecoin to it,and not any other algo,so there's that...

[cbuchner1]

@cbuchner1

Is there a current project to optimize cuda for the X11 algo? Most new coins are run on it and it increasingly looks to be the alternative to scrypt. As such,AMD holds a strong advantage in mining power and outperforms Nvidia by a massive margin. I believe there's a bounty for such a miner as well.

on the roadmap, but it's not the current project...

Note that X11 is not ASIC resistant because every single one of its constituent hashes was submitted to the NIST SHA-3
competition which made efficient hardware-implementability a requirement.

So when many altcoins switch to X11 (a scenario I consider unlikely), expect companies to produce X11 ASICs in no time: history repeating.

[liomojo1]

Guys I have a serious problem, my 6 cards ti 750 is restarting often and Bluscreens apear. If I get it work the chrome trick doesnot work and I don't know what is happening , it was working well for a week and now, crashes. Stock bios all not over locked. Reinstalled windows and again not good. All on USB risers , asrock pro btc , psu 700w cooler master.

I saw now that sometimes 1card stops spinning the fan , when I reconnect the card , this one starts and another one stops. What black magic is this? I don't believe that 5v of my coolermaster can't manage the 6 cards. What do you think could be the issue? 

[bigjme]

Well your gpus will be pulling around 300w. All on molex, is the power supply struggling? Ideally you need as little gpus per molex strand. As they only support 75w per molex strand, so 1 750ti is getting close to it at 50w per card.

You may find the psu cant handle the load and its dropping power, would cause 1 card to stop, which would shut off the system as it would consider it a hardware failure and give a bsod

[liomojo1]

They are 6 evga oc edition with no external power supply molex. The cards are connected with USB risers that are powered by the5v hdd line   I have 3 on one line and 3 on another. I saw that USB risers don't take energy from the pci-e and they power the cards only from the connected 5v line. Could this be the problem?