[1200 TH] EMC: 0 Fee DGM. Anonymous PPS. US & EU servers. No Registration!

[gnar1ta$]

So what miner should I use to mine with a GPU (6950)?

I have 6 6950's using cgminer and eclipse.  No problems.  Check your config.

[1714491845]

1714491845

[1714491845]

1714491845

[kano]

...
cgminer
...
but the lack of ability to split GPUs by miner
...

?
I presume you don't mean the -d option ...

[Inaba]

No, I mean the ability to assign different workers to different GPUs.  I think it's crazy to aggregate all GPUs into one worker.  Then you never know if a GPU is having issues.

[[Tycho]]

No, I mean the ability to assign different workers to different GPUs.  I think it's crazy to aggregate all GPUs into one worker.  Then you never know if a GPU is having issues.

Today's problem is experienced with GUIminer/poclbm, not CGminer

[Inaba]

Oh, whoops, I guess I misread that completely. 

What version of GUIMiner?  Its working with v2011-08-24 for me?

[Kohta]

Oh, whoops, I guess I misread that completely. 

What version of GUIMiner?  Its working with v2011-08-24 for me?

I had it with 20110824,

It seems to only happen when you stop/start, though you might want to double check to see that you are still sending shares just to be safe.

[ocminer]

You can do exactly the same with cgminer since it has the API feature:

http://www.imagebanana.com/view/r3v6r5vi/Bildschirmfoto20120205um09.07.29.png

You can watch every single GPU, it even gets highlighted in red if something is "out of range/limits" using the ANUBIS webinterface for it.

But back to the pool:
When will you do something against all these invalid blocks in the last time ?

[kano]

No, I mean the ability to assign different workers to different GPUs.  I think it's crazy to aggregate all GPUs into one worker.  Then you never know if a GPU is having issues.

Again ... I presume you don't mean the -d option ...

[The00Dustin]

No, I mean the ability to assign different workers to different GPUs.  I think it's crazy to aggregate all GPUs into one worker.  Then you never know if a GPU is having issues.

Again ... I presume you don't mean the -d option ...

Pretty sure he doesn't mean the -d option.  If I am correct he is at least the fourth person to want to have multiple workers in one instance (as opposed to running multiple instances with -d).  Either some other software does that, or cgminer isn't "better enough" (not sure which).  Doesn't really affect me with my lowly one device.

[kano]

No, I mean the ability to assign different workers to different GPUs.  I think it's crazy to aggregate all GPUs into one worker.  Then you never know if a GPU is having issues.

Again ... I presume you don't mean the -d option ...

Pretty sure he doesn't mean the -d option.  If I am correct he is at least the fourth person to want to have multiple workers in one instance (as opposed to running multiple instances with -d).  Either some other software does that, or cgminer isn't "better enough" (not sure which).  Doesn't really affect me with my lowly one device.

Well yeah it's quite simple to do that and saying it can't is quite incorrect ...
Anyway, it makes no difference if you run multiple cgminers
(I'd even expect that if you are using windows and using the 'wrong' ATI drivers with the CPU bug you'd just set their affinity all to the same CPU and not lose anything from that either - no idea of course but anyone want to verify that who knows what they are talking about?)
And with the API (that I wrote ) I'd expect that many people are now using some front end to monitor their mining with cgminer so running 1 or 5 of them would make no difference.

Lastly ... using a pool to decide if something is up with your mining seems a pretty bad hack IMO

[Inaba]

To those of you having problems with Poclbm - the error has been found and you need to update your poblbm.  This applies to raw poclbm AND GUIMiner.  It will get rid of the "Unexpected error" you've been experiencing. 

[Inaba]

Kano -

Yes, I could run multiple instances, but cgminer is too heavy as far as a mining client goes to run multiple instances on some of my stripped down rigs (they already struggle with a plain poclbm and nothing else).  Additionally, one of the advantages of cgminer is suppose to be the multiple GPU support built in, so not being able to split up the GPUs defeats that purpose.

I'm not saying cgminer is bad... and that web interface is nice, but it doesn't really help unless you watch it like a hawk. I would probably use it as my primary miner if it would just let me split the GPUs up by miner in once instance. I can get by without native SMS alerts

But, even then, lets say it encountered a bug like the poclbm bug that just happened... the client wouldn't even know there is a problem - it was busy hashing away at full speed, yet unable to submit shares.  The pool knew, of course, and will alert if no more shares are coming in.  So there are instances where the pool is better at alerting and instances where it's better for the mining client.  But in either case, there's no SMS alerting built in to cgminer, sadly   The webinterface is also kind of useless for people behind firewalls wanting to monitor multiple machines.  Sure you can setup a bunch of firewall and port forwarding rules, but then you have to monitor each machine individually, which is impractical for large setups.

[wknight]

Hehe.. I join the pool.. and find a block

Fear my skills! NOT

https://blockchain.info/block-height/165489

[kano]

Firstly, I'm not trying to convince you to do anything, I'm pointing out that it is quite possible to do EXACTLY what you want to do using cgminer simply coz you said it can't do it.

Kano -

Yes, I could run multiple instances, but cgminer is too heavy as far as a mining client goes to run multiple instances on some of my stripped down rigs (they already struggle with a plain poclbm and nothing else).  Additionally, one of the advantages of cgminer is suppose to be the multiple GPU support built in, so not being able to split up the GPUs defeats that purpose.

I'm not saying cgminer is bad... and that web interface is nice, but it doesn't really help unless you watch it like a hawk. I would probably use it as my primary miner if it would just let me split the GPUs up by miner in once instance. I can get by without native SMS alerts

Do any of them have native status SMS/email?

But, even then, lets say it encountered a bug like the poclbm bug that just happened... the client wouldn't even know there is a problem - it was busy hashing away at full speed, yet unable to submit shares.  The pool knew, of course, and will alert if no more shares are coming in.  So there are instances where the pool is better at alerting and instances where it's better for the mining client.  But in either case, there's no SMS alerting built in to cgminer, sadly  The webinterface is also kind of useless for people behind firewalls wanting to monitor multiple machines.  Sure you can setup a bunch of firewall and port forwarding rules, but then you have to monitor each machine individually, which is impractical for large setups.

The web interface is simply what some people have written.

It is quite simple to ask cgminer for it's status of each GPU ... ... ...
 GPU 0 status on localhost even in a linux shell:
  echo -n "gpu|0" | nc -4 localhost 4028 | tr "|," "\n" | grep -a "^Status=" | cut -d= -f2
Then if it's not "Alive" send out a notification
 (or use a different grep for hash rate of that GPU ... or even the date/time of the last accepted share for that GPU)

Hell, for kicks I even wrote a script that will hop using the API ... yes it works
It's really not hard once you have an interface that doesn't require buckets of code for basic access ...

Any script will also know if cgminer is dead coz cgminer won't reply ...

So what this means is of course you DON'T need to run multiple instances you simply check each GPU on the single instance of cgminer.

Your requirement (which I said needs multiple instances) is to suit using the pool to notify you (hack )
... and if a pool doesn't have that or you use P2Pool ...

The other outage issue is the internet going down and well firstly that means you can't mine anyway ... until it comes back up.
And that really isn't the miners problem, that's the networks problem for how to deal with that ...

Edit: though I should point out the obvious that the 'devs' commands returns details of all GPUs using one command
(that you can then divide up as JSON or plain text depending on the request format being JSON or plain text)

[The00Dustin]

I am guessing that either of you could write something to monitor the API and send e-mails when a GPU stops hashing, and I haven't looked at the API myself, but it may not be difficult to set up a timer that resets per GPU when shares are submitted in order to catch problems that the mining software itself doesn't know about.  Of course, I am not a programmer, so I could be 100% wrong on either count.

[kano]

I am guessing that either of you could write something to monitor the API and send e-mails when a GPU stops hashing, and I haven't looked at the API myself, but it may not be difficult to set up a timer that resets per GPU when shares are submitted in order to catch problems that the mining software itself doesn't know about.  Of course, I am not a programmer, so I could be 100% wrong on either count.

or 100% right.
With the current 2.2.1 (and much older versions) you can see share count per GPU so yeah you could do it that way also ... but it would be better to also check other things like the 'Status' being 'Alive' (or not 'Dead' if you are only looking for 'Dead' GPUs) and the Hash Rate and maybe even the Activity

Here's the latest pull request output from one of my GPUs (I added line breaks):

Code:

Text reply:
STATUS=S,Code=17,Msg=GPU0,Description=Subaru|GPU=0,Enabled=Y,Status=Alive,
Temperature=75.00,Fan Speed=3380,Fan Percent=85,GPU Clock=870,Memory Clock=745,
GPU Voltage=1.175,GPU Activity=99,Powertune=0,MHS av=358.14,MHS 5s=358.59,
Accepted=2195,Rejected=3,Hardware Errors=0,Utility=5.02,Intensity=9,Last Share Pool=0,
Last Share Time=1328460687|

JSON reply example:
{"STATUS":[{"STATUS":"S","Code":17,"Msg":"GPU0","Description":"Subaru"}],
"GPU":[{"GPU":0,"Enabled":"Y","Status":"Alive","Temperature":75.00,"Fan Speed":3371,
"Fan Percent":85,"GPU Clock":870,"Memory Clock":745,"GPU Voltage":1.175,"GPU Activity":99,
"Powertune":0,"MHS av":358.10,"MHS 5s":358.66,"Accepted":2244,"Rejected":3,
"Hardware Errors":0,"Utility":5.03,"Intensity":"9","Last Share Pool":0,"Last Share Time":1328461198}],"id":1}

You can even check 'Last Share Time' if that pull request ever goes into cgminer ...

[village.idiot]

Well it was a nice day until I shut down everything to change out a fan.
Now I'm getting this...

us3.eclipsemc.com:8337 05/02/2012 12:17:12, Unexpected error:                   
Traceback (most recent call last):
  File "/home/mickey/poclbm/HttpTransport.py", line 45, in loop
    self.queue_work(work)
  File "/home/mickey/poclbm/Transport.py", line 117, in queue_work
    self.process(work)
  File "/home/mickey/poclbm/Transport.py", line 76, in process
    self.set_difficulty(work.difficulty)
  File "/home/mickey/poclbm/Transport.py", line 71, in set_difficulty
    self.true_target = np.array(unpack('IIIIIIII', true_target.decode('hex')), dtype=np.uint32)
error: unpack requires a string argument of length 32
us3.eclipsemc.com:8337 05/02/2012 12:17:12, LP connected to us3.eclipsemc.com:8337
us3.eclipsemc.com:8337 05/02/2012 12:17:30, Unexpected error:                   
Traceback (most recent call last):
  File "/home/mickey/poclbm/HttpTransport.py", line 50, in loop
    rv = self.send(result)
  File "/home/mickey/poclbm/Transport.py", line 88, in send
    is_block = belowOrEquals(h[:7], self.true_target[:7])
TypeError: 'NoneType' object is not subscriptable
us3.eclipsemc.com:8337 [202.547 MH/s (~0 MH/s)] [Rej: 0/0 (0.00%)]^C

Can someone tell me what this means?

[Inaba]

Your poclbm is out of date and needs to be updated.

[village.idiot]

How difficult is that.
I don't know how other than to start from scratch following the guide

[Inaba]

Which guide?  If you used my guide, you can go into the poclbm directory and issue "git pull" to update.

so:

cd poclbm
git pull


done.