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

[manifold]

yes...  know, but I prefer the default kernel (is it diabolo?) since I tested all of them a while ago and diabolo was the fastest. Therefore I prefer to work with the older git version...   but why did the hashrate drop?

The kernels have radically changed and I need to choose the right ones. I'm asking you because I need your feedback, i'm not telling you a workaround. You are running very experimental code by trying the git tree.

Ok, of course, sorry:
Current Git:
diablo: 175  MH/s
poclbm: 157 MH/s
phatk: 168 MH/s
poclbm: 160 MH/s

Git from 3/3/2012:
default/diablo: 190 MH/s


Everything on Ubuntu 11.10 64bit, with HD5770 , Intensity 9, no overclocking/underclocking ,
ADL SDK 3.0
APP SDK 2.6  RC3
Cataclyst 12.2

Should I test all the kernels with the older git, too?

[-ck]

yes...  know, but I prefer the default kernel (is it diabolo?) since I tested all of them a while ago and diabolo was the fastest. Therefore I prefer to work with the older git version...   but why did the hashrate drop?

The kernels have radically changed and I need to choose the right ones. I'm asking you because I need your feedback, i'm not telling you a workaround. You are running very experimental code by trying the git tree.

Ok, of course, sorry:
Current Git:
diablo: 175  MH/s
poclbm: 157 MH/s
phatk: 168 MH/s
poclbm: 160 MH/s

Git from 3/3/2012:
default/diablo: 190 MH/s


Everything on Ubuntu 11.10 64bit, with HD5770 , Intensity 9, no overclocking/underclocking ,
ADL SDK 3.0
APP SDK 2.6  RC3
Cataclyst 12.2

Should I test all the kernels with the older git, too?

No, that's cool thanks very much. What a pain. Fixing diablo on 7970 for the earlier 2.6 SDK seems to ruin vliw5. Oh well I guess I can just leave diablo kernel as is and just work with poclbm for 7970.

[tnkflx]

Con,

ckvolias:using latest cgminer 2.3.1-2 win7 64x
after awhile I loose temps and fan speed displays both on screen and from API - a cgminer reboot fixes the issue

Windows driver ATI Display Library fsckage. I can't do anything about it. Does restarting cgminer fix it or rebooting the machine fix it? I could probably try working around the former by detecting "losing adl support" and re-initialising it.

Restarting cgminer always fixes it, no need to restart the whole box.

Is this still something you consider adding to cgminer?

[-ck]

Con,

ckvolias:using latest cgminer 2.3.1-2 win7 64x
after awhile I loose temps and fan speed displays both on screen and from API - a cgminer reboot fixes the issue

Windows driver ATI Display Library fsckage. I can't do anything about it. Does restarting cgminer fix it or rebooting the machine fix it? I could probably try working around the former by detecting "losing adl support" and re-initialising it.

Restarting cgminer always fixes it, no need to restart the whole box.

Is this still something you consider adding to cgminer?

Yes.

[-ck]

yes...  know, but I prefer the default kernel (is it diabolo?) since I tested all of them a while ago and diabolo was the fastest. Therefore I prefer to work with the older git version...   but why did the hashrate drop?

The kernels have radically changed and I need to choose the right ones. I'm asking you because I need your feedback, i'm not telling you a workaround. You are running very experimental code by trying the git tree.

Ok, of course, sorry:
Current Git:
diablo: 175  MH/s
poclbm: 157 MH/s
phatk: 168 MH/s
poclbm: 160 MH/s

Git from 3/3/2012:
default/diablo: 190 MH/s


Everything on Ubuntu 11.10 64bit, with HD5770 , Intensity 9, no overclocking/underclocking ,
ADL SDK 3.0
APP SDK 2.6  RC3
Cataclyst 12.2

Should I test all the kernels with the older git, too?

Try again please with the updated git now. (only diablo kernel has been changed this time).

[manifold]

Try again please with the updated git now. (only diablo kernel has been changed this time).

Correction: it is on average 190 MH/s.
Therefore: Thanks! It works!

[-ck]

Try again please with the updated git now. (only diablo kernel has been changed this time).

no problem:
between 188 and 190 MH/s

Thanks! it works!

But in average it seems to be 2 MH/s lower...

It's the same kernel. Limitations of the goddamn annoying hashmeter over a small timeframe (hence why I introduced the benchmark parameter). Thanks for testing.

[Krak]

Just started getting this today after an update told me to restart.

Code:

[2012-03-28 12:40:03] Started cgminer 2.3.1X Error of failed request:  BadRequest (invalid request code or no such operation)
                                               Major opcode of failed request:  136 ()
        Minor opcode of failed request:  19
                                             Serial number of failed request:  8
  Current serial number in output stream:  8

Using cgminer 2.3.1-2 on Ubuntu 11.10.

[-ck]

Just started getting this today after an update told me to restart.

Code:

[2012-03-28 12:40:03] Started cgminer 2.3.1X Error of failed request:  BadRequest (invalid request code or no such operation)
                                               Major opcode of failed request:  136 ()
        Minor opcode of failed request:  19
                                             Serial number of failed request:  8
  Current serial number in output stream:  8

Using cgminer 2.3.1-2 on Ubuntu 11.10.

AMD  fail. Dodgy driver/SDK combo for your hardware. Downgrade your driver to a less bad one.

[Krak]

AMD  fail. Dodgy driver/SDK combo for your hardware. Downgrade your driver to a less bad one.

I would, but I haven't found a decent guide anywhere to downgrade properly. What was the optimal driver/SDK combo for 6000 series cards again anyway?

[-ck]

AMD  fail. Dodgy driver/SDK combo for your hardware. Downgrade your driver to a less bad one.

I would, but I haven't found a decent guide anywhere to downgrade properly. What was the optimal driver/SDK combo for 6000 series cards again anyway?

12.1 and sdk 2.5 is the least worst -recent- driver sdk combo. 11.6/2.5 is good but wont install on new linux versions.

[Krak]

12.1 and sdk 2.5 is the least worst -recent- driver sdk combo. 11.6/2.5 is good but wont install on new linux versions.

Alright, I'll have to look into it. I dunno why it would start messing up after a random system update though. Seems weird to me.

[-ck]

12.1 and sdk 2.5 is the least worst -recent- driver sdk combo. 11.6/2.5 is good but wont install on new linux versions.

Alright, I'll have to look into it. I dunno why it would start messing up after a random system update though. Seems weird to me.

That's because you're thinking "update" means better... I wish the world was like that. Likely xorg had some update which made the AMD driver weirdness become even weirder, or you got yourself a new AMD driver in the process.

[Krak]

Well, I tried installing 12.1/2.5, but now my computer won't boot past the Ubuntu loading screen. Kinda considering just going back to Windows now.

[-ck]

Well, I tried installing 12.1/2.5, but now my computer won't boot past the Ubuntu loading screen. Kinda considering just going back to Windows now.

Welcome to the AMD fail wagon 

[Krak]

Welcome to the AMD fail wagon 

Yeah. Second time this has happened since I switched to Ubuntu. So it's either boot fails on Ubuntu or registry corruption BSoDs on Windows. Why can't OS X be more efficient at mining? Ugh.

[jake262144]

Yeah. Second time this has happened since I switched to Ubuntu. So it's either boot fails on Ubuntu or registry corruption BSoDs on Windows. Why can't OS X be more efficient at mining? Ugh.

If Windows falls flat on its face and Ubuntu dies in pain, what makes you think that the faildrivers won't take down OS X just as easily?

You need to learn some basic sysadmin tactics: never-ever deploy untested updates on production machines.
Your miners are your production machines, leave good enough alone - find a working driver/sdk combo and do yourself the favor of never touching it without a very good reason.

I've once seen my Debian systems "lose" the AMD driver following a slew of system updates. Fglrx module could no longer be found, manual reinstallation of the driver package helped.

[Krak]

If Windows falls flat on its face and Ubuntu dies in pain, what makes you think that the faildrivers won't take down OS X just as easily?

You need to learn some basic sysadmin tactics - never-ever deploy untested updates on production machines.
Your miners are your production machines.
Leave good enough alone - find a working driver/sdk combo and do yourself the favor of never touching it without a very good reason.

I've once seen my Debian systems "lose" the AMD driver following a slew of system updates. Fglrx module could no longer be found, manual reinstallation of the driver package helped.

OS X has always been great for me. It's just that I lose about 100 MH/s per card when I try to mine with it.

I just have the one miner. I try to keep it up-to-date just to satisfy my own paranoia.

[-ck]

GPU mining sure can be exciting. I had a reference fan on one of my sapphire 6970s fail in spectacular fashion last night. About 1/8th of the blades came flying off into the GPU and then it seized. The GPU still works but obviously won't cool so I can't mine with it. It happened on start up when the fans spin up abruptly to 85% so I've reluctantly relinquished that default speed now to a much gentler 50% on autofan startup and autofan management should be fast enough to raise it if necessary. It's still under warranty, and at least the rest of the machine is still running so nothing like the PSU drama I had and the mobo one before it. Sheesh...

[jake262144]

GPU mining sure can be exciting...

Con, you might be interested in P4man's solution. Fastening the fan with duct tape won't void your warranty.

BTW, whatever did happen to that PSU of yours? Feel like sharing some data points? What make and model was it? What load did it typically pull?
Oftentimes, glitches at the production plant like solder points with inadequate amount of solder won't be severe enough to render the device completely inoperable but they will make themselves known (perhaps even with a bang and some fireworks) as the load increases.