How to get multiple cgminer instances working on same rig

[mxc0bbn]

Ok, so I've got 3 7950s and a 7970 on the same machine.  A while back someone suggested the best way to run this was to have one cgminer instance for the 7950s and a separate instance for the 7970.

After trying a bunch of different things without success I'd definitely appreciate some feedback on what the best way to do this is (if at all possible)...

I would love to just stick to one type/manufacturer of card per rig, but until I can buy 2 more 7970s I'd like to have this working in this one rig...

Any suggestions?

Thanks,
Mike

[os2sam]

You can  try reading the CGMiner top post or the readme.

[os2sam]

Couple hints

--remove-disabled   Remove disabled devices entirely, as if they didn't exist

--device|-d <arg>   Select device to use, (Use repeat -d for multiple devices, default: all)

So put in one instance

-d 0 --remove-disabled

And in the other CGMiner instance

-d 1 --remove-disabled

[mxc0bbn]

Although I am aware of the --remove-disabled and -d for each device I'll take the feedback, thank you.  You can keep the sarcasm though...

I've already read through the readme file (several times in fact) but if I'm posting it's because when I do exactly what you suggested I get instability and problems with the GPUs on both instances and eventually they start to go SICK/DEAD. 

My question was geared towards finding out if there's a better way to do this than the way I've tried and failed.

Thanks,
Mike
 

[philips]

So why running both 7950 and 7970 on the same rig would be a problem?

[os2sam]

I am aware of the --remove-disabled and -d for each device

when I do exactly what you suggested I get instability and problems with the GPUs on both instances and eventually they start to go SICK/DEAD.  

My question was geared towards finding out if there's a better way to do this than the way I've tried and failed.

Thanks,
Mike
  

So you had already tried using "-d 0 --remove-disabled" and it failed?  I didn't see that in your original post.  Or you tried my suggestion and it didn't work for you?

Anyway I have used it in the past with no problems running two CGMiner instances, one for each of my GPU's in one machine.  Don't know why it cause yours to get sick and die.  But that is the way to get CGMiner to ingnore GPU's in one instance so that it can be used in a different instance.

I would say you have some other problem but I have no clue what it is.

[os2sam]

Although I am aware of the --remove-disabled and -d for each device I'll take the feedback, thank you.  You can keep the sarcasm though... 

As for my "sarcasm" it's always best to point folks to the plethora of information that ckolivas has take the time to create for us.  So I posted that suggestion while I refreshed my memory and found the references which had worked for me in the past.  If you had already found those same references and tried them and they failed then, maybe, that should have been in your original post so that someone else could have, maybe, given you suggestions on our actual problem which should probably be phrased something like:

Why do my GPU's go sick and die when I use the --remove-disabled command line argument to disable them so that I can run multiple instances of CGMiner with my multiple GPU's?

Then we could have all done without my obviously undeserved sarcasm and I could have used my time doing something else besides duplicating work that you had already done and just didn't bother to say.

And people wonder why they don't get decent help in these threads.

The only other thing I might suggest, if its not too sarcastic for your delicate ears, maybe post in the CGMiner thread where the authors actually read peoples problems and make for informed suggestions and take bug reports.

Good Luck,
Sam

[os2sam]

So why running both 7950 and 7970 on the same rig would be a problem?

An excellent question.  I wonder if his GPU's were already getting sick before trying the separate instances?

[Eastwind]

So why running both 7950 and 7970 on the same rig would be a problem?

My 7970 likes -g 2
My 7950 likes -g 1
You can  only use either -g 1 or -g 2 in one instance

[os2sam]

So why running both 7950 and 7970 on the same rig would be a problem?

My 7970 likes -g 2
My 7950 likes -g 1
You can  only use either -g 1 or -g 2 in one instance

Really?  Just out of curiosity what intensity are you using for each?

And I think you should be able to set different number of threads per GPU in the same instance.
Sam

[philips]

My 7970 likes -g 2
My 7950 likes -g 1

What happens with 7970 at -g 1 ? Lower hashrate?

I presume you are mining scrypt, because for my 7970 cards -g 1 or -g 2 makes no difference.
(or if there is one I am unable to see it after one hour of testing)

Better post your config here.

[Tailung]

I know exactly what you mean, and I'm doing it for my rig to.
This is how I did it.

Copy the directory of cgminer to a dedicated directory for every gpu you own.

Then do something like this for every gpu. Make a script for running your tasks
This one is for gpu0:
Change the bold/red to your needs

#!/bin/sh
cd /home/miners/cgminer-gpu0  #point to the directory of you dedicated instance of cgminer
rm *.bin
export DISPLAY=:0
export GPU_USE_SYNC_OBJECTS=1
./cgminer -d 0 -o stratum+tcp://mint.bitminter.com:3333 -u xxxx -p XXX -I 9 --gpu-engine 1160 --gpu-memdiff -150 --temp-target 60 --temp-overheat 70 --temp-hysteresis 3 --temp-cutoff 80 --auto-fan --gpu-fan 25-85 --vectors 1 --worksize 256 --failover-only --gpu-threads 2 --no-submit-stale --remove-disabled


Run the script.
Now you have a instance of cgminer for every gpu
This is a linux script but if your running windows you could make a .bat script.

[mxc0bbn]

gave up on this...just running same model on one rig now...thanks for all the advice...

[ssateneth]

scrypt mining, theres your problem