OFFICIAL CGMINER mining software thread for linux/win/osx/mips/arm/r-pi 4.11.1

[The00Dustin]

Hi,

Just installed CGMINER 2.08 on a windows rig with a 5870 overclocked to 990 MHZ, i was getting 400 Mhash/s.
Quite low i thought, so i tried 2.05.
Now i'm getting 440 MHash/s.

Did some code changed from 2.05 to 2.08 that can get this difference in performance?

I am not sure about 10%, but I have seen stopping and restarting cgminer on F15 net a pretty significant MH/s difference.  Perhaps switching to 2.0.5 was just A red herring.

[-ck]

Hi,

Just installed CGMINER 2.08 on a windows rig with a 5870 overclocked to 990 MHZ, i was getting 400 Mhash/s.
Quite low i thought, so i tried 2.05.
Now i'm getting 440 MHash/s.

Did some code changed from 2.05 to 2.08 that can get this difference in performance?

No

[Proofer]

I have done a clean install of Ubuntu 11.10 as part of a conservative strategy to eliminate as many variables as possible in order to get cgminer 2.0.8 to run successfully.  Anyone unfamiliar with my problems thus far who's interested can see my previous posts here and here.

I have 5970 GPUs. How shall I install the ATI driver/utility/dev, or in other words, which version?  The packages version offered by Ubuntu 11.10 (not to be confused with AMD Catalyst 11.10) is 8.881, or with the post-release updates, 8.901.  The version in the current Catalyst 11.12 download from AMD is 8.920.  Or if I should be seeking some even earlier version, where?

I chose to install Catalyst 11.12 and AMD-APP-SDK 2.6, which was released today.  No joy.  I can't think of anything else to try, so I guess I'll put cgminer on the shelf for now.

[P4man]

Hi,

Just installed CGMINER 2.08 on a windows rig with a 5870 overclocked to 990 MHZ, i was getting 400 Mhash/s.
Quite low i thought, so i tried 2.05.
Now i'm getting 440 MHash/s.

FYI, Im using a 5870 @ 1 GHz and Im getting 457 MH using cgminer 2.0.8 on linuxcoin.

[P4man]

Is it possible cgminer doesnt handle DHCP public IP address changes properly?

I am having this problem once in a while, like once or twice per week: on both my machines (one windows, one linux), on both configured pools (failover) cgminer stalls complaining it cant connect to the pool, too slow to respond, switches to the other pool, then it will sometimes connect briefly and fail again over and over. Hashrate is +-0 with only very short moments of normal speeds.

Internet connection is fine. Restarting the router does nothing. Restarting cgminer solves it on both machines.

Im puzzled. THe only thing I can think off that both machines have in common is the NAT  router and therefore my public IP. I dont know when or how often it changes (ill start logging it now) but could there be a correlation, something with long polling perhaps?

Anything else anyone can think off?

[Proofer]

I chose to install Catalyst 11.12 and AMD-APP-SDK 2.6, which was released today.  No joy.  I can't think of anything else to try, so I guess I'll put cgminer on the shelf for now.

Yeah, right.  Like I'm ever going to sleep again if I don't get or make a fix.   So...

The notorious ATI 100% CPU bug:  I could use a little more info.  Is this bug consistent with what I'm experiencing, namely:  after about 15 seconds, i.e., after cgminer appears to have initialized all 6 GPU threads (I have three 5970s) and after the sixth output of

Code:

[2011-12-20 08:08:31] X-Roll-Ntime found
[2011-12-20 08:08:31] Pushing work to requesting thread
[2011-12-20 08:08:31] Pushing work to getwork queue
[2011-12-20 08:08:31] Popping work to stage thread

there is no more output (-D and --verbose are on), single-core CPU usage is 100% (16% x 6), and there is no response to keypress commands (g q, etc.); but there is a response to Ctrl-C.

Is that merely the "100% CPU" bug, which is platform and application independent, or is it a different problem more specific to my setup?  I think it's the latter, but I'm not positive.

[ancow]

Is that merely the "100% CPU" bug, which is platform and application independent, or is it a different problem more specific to my setup?  I think it's the latter, but I'm not positive.

The 100% bug is simply that any usage of the driver's 3D capabilities results in an immediate jump in CPU usage, which doesn't really affect mining except for the increased heat (unless you're also CPU mining, which would become even more useless).

What you describe seems more like an endless loop that causes the program to hang. I find it kind of curious that there's no more output, though. To me, that suggests that the problem probably isn't in cgminer.

[DeathAndTaxes]

The notorious ATI 100% CPU bug:  I could use a little more info.  Is this bug consistent with what I'm experiencing, namely:  after about 15 seconds, i.e., after cgminer appears to have initialized all 6 GPU threads (I have three 5970s) and after the sixth output of

Code:

[2011-12-20 08:08:31] X-Roll-Ntime found
[2011-12-20 08:08:31] Pushing work to requesting thread
[2011-12-20 08:08:31] Pushing work to getwork queue
[2011-12-20 08:08:31] Popping work to stage thread

there is no more output (-D and --verbose are on), single-core CPU usage is 100% (16% x 6), and there is no response to keypress commands (g q, etc.); but there is a response to Ctrl-C.

Is that merely the "100% CPU" bug, which is platform and application independent, or is it a different problem more specific to my setup?  I think it's the latter, but I'm not positive.

That is 100% not the "100% CPU" bug.  CPU bug is simply a condition where you waste power because of a flaw in AMD driver which pegs CPU at 100% usage when running OpenCL code.  What you describe is something else and something I haven't seen before.   One thing you could try is running it w/ option "real-quiet" and check for output on the remote server.   If still nothing then likely you got some config/system issue which is putting cgminer into a race condition.

[Proofer]

Is that merely the "100% CPU" bug, which is platform and application independent, or is it a different problem more specific to my setup?  I think it's the latter, but I'm not positive.

The 100% bug is simply that any usage of the driver's 3D capabilities results in an immediate jump in CPU usage, which doesn't really affect mining except for the increased heat (unless you're also CPU mining, which would become even more useless).

What you describe seems more like an endless loop that causes the program to hang. I find it kind of curious that there's no more output, though. To me, that suggests that the problem probably isn't in cgminer.

My plan is to start by downgrading the fglrx driver to the version that's in the Ubuntu 11.10 repository (currently the Catalyst 11.12 version is installed), which I think is approximately equivalent to Catalyst 11.8.

I hope I don't have to end by becoming proficient in GDB, but if that's what it takes...

Meanwhile, any other ideas are welcome!

[Proofer]

[...]  One thing you could try is running it w/ option "real-quiet" and check for output on the remote server.   If still nothing then likely you got some config/system issue which is putting cgminer into a race condition.

Sorry, don't know what you mean by the remote server.  I access the miner (Ubuntu) rig via ssh from my Mac; both are on my LAN.  A couple of times I've tried running it from the local (miner rig) terminal window, but then when the output stops I'm completely locked out -- no keyboard, no mouse -- and to regain control I have to go back to my Mac, where via ssh I "sudo reboot".  (I always reboot, to reduce the GPU fans' speed.)  I just tried this again, and noticed that the output stopped earlier -- in the initializing of the 2nd or 3rd (of 6) GPU cores -- but that could be because that the hang/freeze occurred before all the terminal output was flushed.

[ancow]

My plan is to start by downgrading the fglrx driver to the version that's in the Ubuntu 11.10 repository (currently the Catalyst 11.12 version is installed), which I think is approximately equivalent to Catalyst 11.8.

Versions 11.7 through 11.10 had the 100% bug and versions 11.11+ can behave strange (sometimes displaying the buggy behaviour, sometimes not), so the safe bet is to go with 11.6.

I hope I don't have to end by becoming proficient in GDB, but if that's what it takes...

Don't! Using gdb to debug cgminer while it's executing OpenCL code on the card connected to your display *will* lead to a complete X server freeze. (If you're still using VNC to access an X server not using the fglrx driver, you might be fine.)

[Proofer]

[Catalyst] Versions 11.7 through 11.10 had the 100% bug and versions 11.11+ can behave strange (sometimes displaying the buggy behaviour, sometimes not), so the safe bet is to go with 11.6.

Is 11.6 compatible with Linux kernel 3 (Ubuntu 11.10)?  I thought I'd seen somewhere that it's not.

I hope I don't have to end by becoming proficient in GDB, but if that's what it takes...

Don't! Using gdb to debug cgminer while it's executing OpenCL code on the card connected to your display *will* lead to a complete X server freeze. (If you're still using VNC to access an X server not using the fglrx driver, you might be fine.)

No more VNC: I'm using ssh -X, which not only gives me remote access via a terminal but lets me run GUI apps (e.g., synaptic) as well.

[Proofer]

I use Gentoo linux on the x86_64 platform, I'm currently running kernel 3.1.4 with the Con Kolivas patches. Either the ATI driver 11.6 or the recently released 11.11 driver work fine for me *without* the 100% cpu use bug. Neither is the current "production" gentoo version of the driver, which is 11.8 .

As an aside, I do have an older system that dual boots to Windows XP Pro 32 bit. I tried the 11.11 driver there and the 100% cpu bug was present. 11.6 works fine. I have not tried a 32 bit linux platform.

Video driver compatibility has most to do with your linux kernel version and options. It's likely any recent ATI driver should at least mostly work for any kernel that you are likely to have in use.

(Emphasis added)

What Con Kolivas patches?

[DeathAndTaxes]

Sorry, don't know what you mean by the remote server.  I access the miner (Ubuntu) rig via ssh from my Mac; both are on my LAN.  A couple of times I've tried running it from the local (miner rig) terminal window, but then when the output stops I'm completely locked out -- no keyboard, no mouse -- and to regain control I have to go back to my Mac, where via ssh I "sudo reboot".  (I always reboot, to reduce the GPU fans' speed.)  I just tried this again, and noticed that the output stopped earlier -- in the initializing of the 2nd or 3rd (of 6) GPU cores -- but that could be because that the hang/freeze occurred before all the terminal output was flushed.

I mean check for shares on pool server.  I was wondering if it was just some display glitch but if you are getting system lock outside of cgminer it is far beyond that.  I have never have cgminer lockup a linux system like that.  Very weird behavior. 

[Proofer]

Anyone tried this?

The ATI drivers after 11.6 have a bug that makes them consume 100% of CPU.
But drivers before 11.7 not work with kernel 3.0+
So I have found a patch:
http://eduard-x.blogspot.com/2011/05/amd-catalyst-115-2639-debian.html
and using 3.1 kernel and 11.6 drivers with this patch on Ubuntu oneiric.
Everything works fine (also spent full day to found this solution).

The direct source of the patch, in English:
http://www.mindwerks.net/2011/04/ubuntu-11-04-natty-with-fglrx-and-2-6-39/

[Proofer]

I mean check for shares on pool server.  I was wondering if it was just some display glitch but if you are getting system lock outside of cgminer it is far beyond that.  I have never have cgminer lockup a linux system like that.  Very weird behavior.  

According to the pool server, my last activity was 15,328 days ago, i.e., at the beginning of time (in the Unix universe).  Cgminer reported getting some work from it, but that might not count as "activity."

I fell back to fglrx 8.881, the version w/o post-release update, that 11.10 installs from its repositories.  The only difference is that now running cgminer locks up the system so tight that I can't get in via ssh; with the Catalyst 11.12 driver, I was getting in via ssh.

Very weird behavior, indeed.  Two other anomalies that seem unrelated but which I'll mention in case any bells ring:  with fglrx installed, (1) Gnome Classic comes up with only the desktop, no top panel and hence no way to do anything with the GUI; (2) there's no way (that I've found) to dismiss the screen blanker in Unity other than Ctrl-Alt-F2 to get a shell login prompt followed by Ctrl-Alt-F7 to return to the GUI.  Without fglrx, these two anomalies are not present.  But without fglrx, cgminer exits immediately  

Other than the cgminer problem and those two 11.10 anomalies the system seems normal to me, although I'm a newbie w/r Linux.  For completeness, I mention that this is a homebrew system which first had power applied about a week ago.

I'll bet that if a solution is found, it's going to be of the forehead-slapping, "you did [or didn't do] WHAT!?" variety.

[BkkCoins]

Anyone tried this?

The ATI drivers after 11.6 have a bug that makes them consume 100% of CPU.
But drivers before 11.7 not work with kernel 3.0+
So I have found a patch:
http://eduard-x.blogspot.com/2011/05/amd-catalyst-115-2639-debian.html
and using 3.1 kernel and 11.6 drivers with this patch on Ubuntu oneiric.
Everything works fine (also spent full day to found this solution).

The direct source of the patch, in English:
http://www.mindwerks.net/2011/04/ubuntu-11-04-natty-with-fglrx-and-2-6-39/

I'm running Ubuntu 11.04 (32 bit) Desktop with ADK 2.4 and drivers 11.6 and cgminer. No problems at all. I'd suggest starting from fresh and following closely the Ubuntu headless tutorial here,

https://bitcointalk.org/index.php?topic=9239.0

(in my case I used 32 bit versions where suitable, and I ran Phoenix for a while before trying and prefering cgminer) This combo works great for me and I wouldn't suggest bothering with newer driver versions as there are generally cpu problems and no advantages afaik.

[Proofer]

I'll bet that if a solution is found, it's going to be of the forehead-slapping, "you did [or didn't do] WHAT!?" variety.

WOOHOO!  cgminer is running!  The solution?  Don't log in on the miner -- leave it at the login screen.  For now, for me, this is a head-scratcher rather than a forehead-slapper.

However, not out of the woods completely:

[2011-12-20 15:09:22] ADL Initialisation Error!
[2011-12-20 15:09:22] 0 gpu miner threads started
[2011-12-20 15:09:22] 1 cpu miner threads started, using SHA256 'sse2_64' algorithm.

Edit:  Maybe it's running only because it couldn't access the GPUs 

[DeathAndTaxes]

I'll bet that if a solution is found, it's going to be of the forehead-slapping, "you did [or didn't do] WHAT!?" variety.

WOOHOO!  cgminer is running!  The solution?  Don't log in on the miner -- leave it at the login screen.  For now, for me, this is a head-scratcher rather than a forehead-slapper.

However, not out of the woods completely:

[2011-12-20 15:09:22] ADL Initialisation Error!
[2011-12-20 15:09:22] 0 gpu miner threads started
[2011-12-20 15:09:22] 1 cpu miner threads started, using SHA256 'sse2_64' algorithm.

Edit:  Maybe it's running only because it couldn't access the GPUs 

Wait what?  What login screen?

Also just curious what happens if you try "cgminer -n"?

[BkkCoins]

I'll bet that if a solution is found, it's going to be of the forehead-slapping, "you did [or didn't do] WHAT!?" variety.

WOOHOO!  cgminer is running!  The solution?  Don't log in on the miner -- leave it at the login screen.  For now, for me, this is a head-scratcher rather than a forehead-slapper.

However, not out of the woods completely:

[2011-12-20 15:09:22] ADL Initialisation Error!
[2011-12-20 15:09:22] 0 gpu miner threads started
[2011-12-20 15:09:22] 1 cpu miner threads started, using SHA256 'sse2_64' algorithm.

Edit:  Maybe it's running only because it couldn't access the GPUs 

Make sure you don't skip these commands any time you change GPU slots/combinations:

sudo aticonfig -f --initial --adapter=all
sudo reboot

Make sure cards are recognized correctly and listed with,

DISPLAY=:0 sudo fglrxinfo

Also see what you get listed from the SDK utility "clinfo". It should list any cards and CPU. If not then some initialization hasn't occurred.