Topic: OFFICIAL CGMINER mining software thread for linux/win/osx/mips/arm/r-pi 4.11.0 - page 337. (Read 5806004 times)

Scrypt mining uses way more bandwidth than BTC mining.
Thus if you are reducing your PCI-e bandwidth dramatically (16x to 1x) then expect a performance drop or related problems.
Scrypt mining is affect by many, things including RAM, GPU RAM and Motherboard (which includes PCI-e)
- which is why you can't expect to be able to copy someone else's config and expect to get the same results.

I am having an issue with CGMiner. I've been mining just fine with 2 X 7950s @ around 1.1MH/s for BTB. I recently received risers in the mail -1x to 16x. I have not received my 16x to 16x risers yet, so I am only running 3 GPUs. All Gigabyte 7950s. All new. Haven't had any problems with them except some SICK/DEAD statuses if I let them run too hot for too long. I've realized impatience waiting for all the risers is not worth burning up $1000 worth of GPUs so I am keeping them under clocked for the moment.

I use Team Viewer to monitor the rig while I am away from my home. It usually works flawlessly. The android app is nowhere short of amazing.

I was running Catalyst 13.4 / APP SDK 2.8 / CGMiner 3.1.0. I had been mining for 5 days straight with minimal problems. My only problem was cooling. I have since redesigned my Crate Rig to be more efficient for cooling. I will have to wait until I can put all 4 X 7950s on risers though until I can run at full force and at a manageable temperature.

Now, my CGMiner issue.

With my original drivers stated above, I was having my KH/s range with a -20% +20% range. I closed out of CGMiner to give the rig a moment to rest(again, not trying to burn out my hardware). When I started CGMiner back up, I would get different errors every time. 'Could not connect to pool', 'pool 0 not providing work fast enough' 'Could not connect, make sure the URL is correct' and every other error I know about that deals with trouble connecting. The pool would connect rarely with a slow to nothing connection. I then began getting multiple HW errors for each card. After a couple of minutes the rig was submitting shares. This didn't last long as the cards were heating up so I wanted to fix the connection and HW errors for good.

 I was having issues with my ISP so I tethered my Nexus 4 with non throttled HPSA+, but the connection errors are still being shown. Is there something I am missing? Does the currency you intend to mine, have to have the corresponding wallet installed to the PC with the block chain fully caught up? I didn't think so because my BTB wallet was having trouble syncing with a time error. I was still able to mine BTB through binarycoins.eu . I also emailed support for binarycoins.eu and he said that his pool was submitting work and others had no problem connecting. I also tried notroll.in for LTC just to see if it was a computer error or remote server error. All help and suggestions are highly appreciated.

I am running Catalyst 12.8 / APP SDK 2.7 / While trying CGMiner 2.8.0, 2.8.2, 3.1.0, and 3.1.1. Also, last night I installed a fresh copy of Windows 7 64x just to see if a Windows update caused the error. Before trying a fresh install, I tried rolling the system restore back and installing all of the above (recommended for best performance) drivers.

These drivers were in a last ditch effort to fix my pool connection issue. They actually blew me away with the capability of the cards. With the one card mounted to the outter PCI-E slot and the one on a 1X to 16X riser mounted to the top of the crate, both run cool. The main slot does not. In a short time of testing even though the pools were not connecting, the cards would start hashing. I was able to push STEADY 700+KH/s for each RADEON 7950!! I thought these cards were only capable of 630KH/s max for litecoin based coins! I can't wait to get this connectivity issue fixed and have my rig running a solid 3KH/s !!  

If you guys need to know the rest of my HW:

Seasonic Gold 80+ 1250W Modular
Corsair 4GB DDR 3 1666MHZ
ASUS M5A97 R2.0
500 GB Laptop Drive (soon to be replaced by a 60GB SSD OCZ)
Sempron 145 O/C'd to 3.6 GHz stable. MOBO has built in auto O/C features.
MICRO USB EDIMAX WIFI N

StartMining.bat is pretty generic
cgminer --scrypt -o http://binarycoins.eu:9333/ -u MyMiner.1 -p x
with a few settings added(cant connect to PC at the moment as my ISP COMCAST has their heads up their arses. But I do know the settings I have can hash just fine at good and consistent rates, my main issue is the pool connectivity.

If I missed any details, let me know and I can provide them.

The question was mostly rhetorical. Windows does the driver reset dance sometimes and I have a small related entry in the FAQ about it. I've had linux mining on the same cgminer instance for 6 months once. I'm led to believe that's impossible on windows.

Mine is.  Op didn't say.

M

Windows?

I have the same issue.  I remember reading the AMD drivers go kaput after so many days of continually use. 

M

Thank you. So besides of these options, the icarus usb devices plus a ztex board would be automatically recognized and worked on? That'd be great.

I have a 7950 in one of my rigs that is a bit "wonky", and every other week or so, it just stops hashing and appears as DEAD in CGMiner.

If I quit and restart the miner (q, up-arrow, enter), up comes the wonky card again for at least another week with no problems.

Thought it might be worth reporting as a curiosity.

Not really, there is some magic going on that you're not aware of. When GPUs talk to the operating system, they use pinned memory if possible to transfer information from GPU to system, even if cgminer itself uses hardly anything (I designed it to be very lean internally). There is no doubt that having enough ram matters, but what that equation is, no one has been able to accurately figure out. I think you just need more system RAM than the largest GPU RAM. You probably don't need more system RAM than the total GPU RAM.

Thanks for cgminer. It's an awesome tool. That's some pretty weird programming, for scrypt, with some weird magic sweet spot ratio between GPU clock and memory clock speed. I can't even imagine what the heck you're doing. But it works!

You say to make sure you have enough system RAM, more than your combined GPU RAM. I also can't make any sense of that. In Linux (Ubuntu-64bit), my cgminer process reserves just over 1GB VIRT memory, and only uses 60-70MB resident. I have 4GB system RAM, and "top" claims 2.5GB free. Yet my two GPUs add up to 6 GB of GDDR5. Everything seems to work well, but I can't hit some of the peaks I read about, like 700 KH/sec on a 7970 w/3GB.

I tried doubling my system RAM from 4GB to 8GB and absolutely nothing changed.

Any suggestions or comments?

Run a separate instance of cgminer for each coin, and make a script that uses the API to enable GPUs on some instance, and disable them on all the others.

Third line of readme...

You're welcome.

Let me know if this is unsupported, but I think I found a bug (or my misunderstanding of the capabilities) in cgminer.  I'm running cgminer on BAMT with a GPU mining bitcoins and everything works just fine 24/7. 

I wanted to check out some FreiCoin (sha256d), so I added a second pool to my list of pools in cgminer.conf and then set rotate to 5 minutes.  My expectation is that every five minutes I would alternate between mining BitCoins and FreiCoins.  Everything seemed to work the first day, but the second day the BitCoin pool had some connection problems (stratum connection interrupted) and it kept switching between the pools.  Unfortunately it seemed that it would submit BitCoin work to the FreiCoin pool and vice-versa.  I got quite a few strange errors, including 'H not zero' replies from the pools.  Sorry, I don't have a good log right now.

Is this supposed to work?  Is there a better way to achieve this goal of mining two types of coins on the same hardware?

-E

P.S.  Thanks for the great software.

Tks for the reply, it is a very good tip for those who use several GPUs, i would be happy if it was my case but unfortunately i only use one. I tryed it just in case anyway but it didnt maked any change.

I will try look into Power saving configs. when the screen its turned off to see if there is smth i can change, meanwhile if someone got any idea or know what can be causing the problem it would be welcome some tips

Yes - I do already.
 https://bitcointalksearch.org/topic/m.2129102
and
 https://bitcointalksearch.org/topic/m.2137231

However at the moment it also requires options of --icarus-options 115200:1:1 --icarus-timing 2.9761=100

I'm now updating the code to automatically detect all Icarus devices (the AMU - AsicminerUSB is an Icarus device) and it will have similar to those options by default for the AMU and default options for Lancelot and Cairnsmore1 also

If you are using cards that support ULPS, disable it. Latest MSI Afterburner has a convenient checkbox for doing so. That's exactly what happens when the display is powered off on an ULPS-enabled card.

Also, with a 2x7970 Crossfire setup, I had to disable monitor power saving totally, just disabling ULPS didn't help there.

Tks for the tip, i haven't tryed that but looks easy and i got the resistances, but it will make a little energy drop there.

Im still wondering if there is some solution on software conf. that dont allow the GPU (ATI) to idle and drop near half of the hash power.

Tks for the sugestions /help

Short question: Is it possible to mine with several different fpgas at the same time? For instance a ztex board plus a few of these: https://bitcointalksearch.org/topic/m.2025318

Works that "out of the box" (start cgminer and that's it) or would that be somehow tricky?

Looks like the GPU is idling for some peculiar reason.
Did you try dummy vga plugs?

I recently moved to cgminer and i noticed that hash rate drop when screen power its turn off, is there a way to avoid that?

Shouldnt be the opposite as GPU no longer need to refresh the screen?


Most likely its some software config or any linux/windows configuration that i'm falling. Sry if its double post i tryed to search and found noothing related with this issue.

Tks in advance for the reply