Pages:
Author

Topic: BAMT version 0.5 - Easy USB based mining Linux with farm wide management tools - page 75. (Read 324176 times)

hero member
Activity: 910
Merit: 1000
Items flashing here available at btctrinkets.com
I'd like to report a stability issue in bamt 0.5a:
Using the exact same settings as I did in the latest 0.4 version I cant get the only rig I have tried Bamt 0.5 in to remain stable, as Im writing this im writing the 0.4 Image back to my usb-stick.

The rig contains one 5870, the clocks Im running are core 970, stock voltage, mem clock at 300. Im using phoenix miner, kernel params BFI_INT VECTORS FASTLOOP=false AGGRESSION=8 (8 because im pointing it at P2pool). Every time (twice so far) that I head out to work or to sleep I come back to find the rig has hung up, wont respond to anything or cant get any imange on it's display.

What other relevant information can I post to help identify the issue ?


I really don't know.

0.5 is the same mining software, same overclocking software, same drivers, same SDK....

I have no idea why or how it would behave any differently in 0.5 than it did in 0.4.
It is the same software.




Also on a sidenote the exactly same rig, with the exact same setup Is providing a slightly higher Mhs (according to GPUmon) in bamt 0.4. Were talking 1% here, 2-5mhs more, GPUmon has always shown slight variance between ticks to me. Currently it's hashing away at 432.25-439.9 Mhs.

Edit: It just occured to me that theres one difference in the setup. In 0.4 I use phatk, where as in 0.5 I used the reccomended phatk2. Any toughts on this ?


*sigh...*

a note to anyone reading this

PLEASE

do not tell me "the exact same thing that works on X doesn't work on Y" if you aren't doing the exact same thing.  I cannot help you without accurate information.  Even worse, inaccurate information leads to bad conclusions that can effect the support other people get.

So far, the only repeatable and verified "issues" with 0.5 have been:

2 cases of people not realizing that the pound sign is a comment in Linux

2 cases of typos

1 case of unknown reason but that went away after installing fix #1.

(and a verified broken fixer, which is indeed a pretty silly bug to have)

(PS - This is out of hundreds of downloads)

I understand that it is frustrating when things don't work, but I just cannot guess at what you may have changed and not mentioned.  

Anyway.. more to the point, of course if you change kernels you will very often see a variance in hash rate and very possibly cause instability.  Changing kernels can be a drastic move when you are teetering on the edge of crashing your cards anyway.  Each kernel will have a different heat to performance to stability ratio.  phatk2 is generally accepted as the best kernel to use for most GPUs.  However, it may not be the best for every combination of GPU and clock rates.  Also, you may have to adjust your clock rates to get the best performance out of a new kernel.  You cannot just drop a new kernel in and expect it to have the same characteristics or work very well without tuning your GPUs for it!





I understand it's frustrating getting mislead with innaccurate reports and do apologize for mine. I assure you I triple checked the configs in comparison before posting, but sadly it was only while sitting on the toilet that the difference of configurations occured to me. That phatk2 kernerl was the only real incentive I had to swap out from 0.4, guess Im going to stick with it for the time being.

Imho you need to understand or appreciate the fact that a fair amount of the ppl using or even considering using Bamt migth be next-to illiterate when it comes to linux, myself for one. Up to until fairly recentley I have found bamt to be slightly easyer to use than P2pool for example ,which I managed to eventually do. For your software to be more-wideley used and your support donations to grow alongside it you propably should make attempts at making setting things up and using it easyer, here are my two bitcents on the issue:
-We like dropbox style menus, we like mouse tooltips giving us hints, we'd propably love invalid syntaxes in configs to "magically" turn red and start blinking or something.
-We (and you) might benefit from being discouraged on installing new versions among the first users.
hero member
Activity: 616
Merit: 506
I'd like to report a stability issue in bamt 0.5a:
Using the exact same settings as I did in the latest 0.4 version I cant get the only rig I have tried Bamt 0.5 in to remain stable, as Im writing this im writing the 0.4 Image back to my usb-stick.

The rig contains one 5870, the clocks Im running are core 970, stock voltage, mem clock at 300. Im using phoenix miner, kernel params BFI_INT VECTORS FASTLOOP=false AGGRESSION=8 (8 because im pointing it at P2pool). Every time (twice so far) that I head out to work or to sleep I come back to find the rig has hung up, wont respond to anything or cant get any imange on it's display.

What other relevant information can I post to help identify the issue ?


I really don't know.

0.5 is the same mining software, same overclocking software, same drivers, same SDK....

I have no idea why or how it would behave any differently in 0.5 than it did in 0.4.
It is the same software.




Also on a sidenote the exactly same rig, with the exact same setup Is providing a slightly higher Mhs (according to GPUmon) in bamt 0.4. Were talking 1% here, 2-5mhs more, GPUmon has always shown slight variance between ticks to me. Currently it's hashing away at 432.25-439.9 Mhs.

Edit: It just occured to me that theres one difference in the setup. In 0.4 I use phatk, where as in 0.5 I used the reccomended phatk2. Any toughts on this ?


*sigh...*

a note to anyone reading this

PLEASE

do not tell me "the exact same thing that works on X doesn't work on Y" if you aren't doing the exact same thing.  I cannot help you without accurate information.  Even worse, inaccurate information leads to bad conclusions that can effect the support other people get.

So far, the only repeatable and verified "issues" with 0.5 have been:

2 cases of people not realizing that the pound sign is a comment in Linux

2 cases of typos

1 case of unknown reason but that went away after installing fix #1.

(and a verified broken fixer, which is indeed a pretty silly bug to have)

(PS - This is out of hundreds of downloads)

I understand that it is frustrating when things don't work, but I just cannot guess at what you may have changed and not mentioned.  

Anyway.. more to the point, of course if you change kernels you will very often see a variance in hash rate and very possibly cause instability.  Changing kernels can be a drastic move when you are teetering on the edge of crashing your cards anyway.  Each kernel will have a different heat to performance to stability ratio.  phatk2 is generally accepted as the best kernel to use for most GPUs.  However, it may not be the best for every combination of GPU and clock rates.  Also, you may have to adjust your clock rates to get the best performance out of a new kernel.  You cannot just drop a new kernel in and expect it to have the same characteristics or work very well without tuning your GPUs for it!



hero member
Activity: 910
Merit: 1000
Items flashing here available at btctrinkets.com
I'd like to report a stability issue in bamt 0.5a:
Using the exact same settings as I did in the latest 0.4 version I cant get the only rig I have tried Bamt 0.5 in to remain stable, as Im writing this im writing the 0.4 Image back to my usb-stick.

The rig contains one 5870, the clocks Im running are core 970, stock voltage, mem clock at 300. Im using phoenix miner, kernel params BFI_INT VECTORS FASTLOOP=false AGGRESSION=8 (8 because im pointing it at P2pool). Every time (twice so far) that I head out to work or to sleep I come back to find the rig has hung up, wont respond to anything or cant get any imange on it's display.

What other relevant information can I post to help identify the issue ?


I really don't know.

0.5 is the same mining software, same overclocking software, same drivers, same SDK....

I have no idea why or how it would behave any differently in 0.5 than it did in 0.4.
It is the same software.




Also on a sidenote the exactly same rig, with the exact same setup Is providing a slightly higher Mhs (according to GPUmon) in bamt 0.4. Were talking 1% here, 2-5mhs more, GPUmon has always shown slight variance between ticks to me. Currently it's hashing away at 432.25-439.9 Mhs.

Edit: It just occured to me that theres one difference in the setup. In 0.4 I use phatk, where as in 0.5 I used the reccomended phatk2. Any toughts on this ?
hero member
Activity: 616
Merit: 506
I'd like to report a stability issue in bamt 0.5a:
Using the exact same settings as I did in the latest 0.4 version I cant get the only rig I have tried Bamt 0.5 in to remain stable, as Im writing this im writing the 0.4 Image back to my usb-stick.

The rig contains one 5870, the clocks Im running are core 970, stock voltage, mem clock at 300. Im using phoenix miner, kernel params BFI_INT VECTORS FASTLOOP=false AGGRESSION=8 (8 because im pointing it at P2pool). Every time (twice so far) that I head out to work or to sleep I come back to find the rig has hung up, wont respond to anything or cant get any imange on it's display.

What other relevant information can I post to help identify the issue ?


I really don't know.

0.5 is the same mining software, same overclocking software, same drivers, same SDK....

I have no idea why or how it would behave any differently in 0.5 than it did in 0.4.
It is the same software.


hero member
Activity: 910
Merit: 1000
Items flashing here available at btctrinkets.com
I'd like to report a stability issue in bamt 0.5a:
Using the exact same settings as I did in the latest 0.4 version I cant get the only rig I have tried Bamt 0.5 in to remain stable, as Im writing this im writing the 0.4 Image back to my usb-stick.

The rig contains one 5870, the clocks Im running are core 970, stock voltage, mem clock at 300. Im using phoenix miner, kernel params BFI_INT VECTORS FASTLOOP=false AGGRESSION=8 (8 because im pointing it at P2pool). Every time (twice so far) that I head out to work or to sleep I come back to find the rig has hung up, wont respond to anything or cant get any imange on it's display.

What other relevant information can I post to help identify the issue ?
hero member
Activity: 504
Merit: 500
sr. member
Activity: 271
Merit: 250
I have a few small questions about my BAMT, one is about BAMT v0.4b

Is it OK to ask questions about the last version in heree, or are you trying to keep them seperate? or
hero member
Activity: 626
Merit: 500
Mining since May 2011.
hero member
Activity: 626
Merit: 500
Mining since May 2011.
hero member
Activity: 616
Merit: 506
hero member
Activity: 626
Merit: 500
Mining since May 2011.


If you are speaking of the core_speed: 950 & mem_speed: 300 vs. core_speed_2: 950 & mem_speed_2: 300 it does not matter. My 0.4 USB stick for that rig is stock 0.4 bamt.conf, edited with what you see there, been running for weeks.

0.5 (even when testing the beta x64) seemed to have introduced the core_speed_X:, mem_speed_X:, etc (3 profiles) I have never been able to get the 5850s OC working in 0.5 any build (x86/x64).

No, the 3 profile, whatever_0 etc type settings have been in 0.4 for months.  Either style works fine in either version.  if you don't specify the number with _X, it just sets profile 2.  Always been that way.

interesting that you mention the x64 test build did not work either.  it was exactly the same code as 0.4!  only difference between x64 test and 0.4 was the 64bit kernel.  I shouldn't have called that "0.5" at all, that was a mistake.

I have even tried taking my working 0.4 bamt.conf over to 0.5 verbatim, does not work. Just an FYI as I mentioned before, this is not just one 5850, I have tested 4 separate rigs now with the same results. Some are Sapphire, one is Gigabyte. All different model MB, 3 ASUS, 1 Gigabyte. The only constant here is the 5850s, my 5830s are mining without issue, same exact setting going from 0.4 to 0.5. I will start in on testing 69xx after some sleep.

There is something weird going on, but it is not the "difference" between 0.4 and 0.5, since you saw this problem on x64, which was 0.4.   Time to look at other possibilities.


Here are the results of some further testing today with 3 more rigs:

miner-04 - running 0.5
GPU0: Sapphire 5830 - 314 Mhash/s - OC Working
GPU1: Diamond 6970 - 420 Mhash/s - OC Working
Same exact results in 0.4!  Cool

miner-06 - running 0.5
GPU0: Gigabyte 5870 - 406 Mhash/s - OC Working
GPU1: Sapphire 5830 - 313 Mhash/s - OC Working
Same exact results in 0.4!  Cool

miner-02 - running 0.5
GPU0: HIS 6950 - 341 Mhash/s - OC NOT Working
GPU1: Gigabyte 5850 - 306 Mhash/s - OC NOT Working
queue SadTrombone.com  Cry
And yes I am copying my exact settings from 0.4 to 0.5.

That same rig running 0.4
GPU0: HIS 6950 - 373 Mhash/s - OC Working
GPU1: Gigabyte 5850 - 387 Mhash/s - OC Working

Results of debug under 0.5:
Code:
root@miner-02:~# mine restart
Stopping mining processes...: mine..3481.gpu1..3448.gpu0..wrapper-1.pid..wrapper-0.pid...
Starting mining processes...: mine
(config sync: 'bamt.conf' in live dir is updated, copying to offline)
..munin..GPU 0..fan 0

--[ Debug info for O/C on GPU 0 ]------------------------------------------------

GPU is enabled, overclocking is enabled

OC command - profile 0: DISPLAY=:0.0 /usr/local/bin/atitweak -P 0 -A 0 -e 300 -m 300 -v 1.125

Results:
Setting performance level 0 on adapter 0: engine clock 300MHz, memory clock 300MHz, core voltage 1.125VDC
ADL_Overdrive5_ODPerformanceLevels_Set failed.

OC command - profile 1: DISPLAY=:0.0 /usr/local/bin/atitweak -P 1 -A 0 -e 800 -m 300 -v 1.125

Results:
Setting performance level 1 on adapter 0: engine clock 800MHz, memory clock 300MHz, core voltage 1.125VDC
ADL_Overdrive5_ODPerformanceLevels_Set failed.

OC command - profile 2: DISPLAY=:0.0 /usr/local/bin/atitweak -P 2 -A 0 -e 920 -m 300 -v 1.125000

Results:
Setting performance level 2 on adapter 0: engine clock 920MHz, memory clock 300MHz, core voltage 1.125VDC
ADL_Overdrive5_ODPerformanceLevels_Set failed.

-------------------------------------------------------------------------------

..GPU 1..fan 1

--[ Debug info for O/C on GPU 1 ]------------------------------------------------

GPU is enabled, overclocking is enabled

OC command - profile 0: DISPLAY=:0.0 /usr/local/bin/atitweak -P 0 -A 1 -e 300 -m 300 -v 1.125

Results:
Setting performance level 0 on adapter 1: engine clock 300MHz, memory clock 300MHz, core voltage 1.125VDC
ADL_Overdrive5_ODPerformanceLevels_Set failed.

OC command - profile 1: DISPLAY=:0.0 /usr/local/bin/atitweak -P 1 -A 1 -e 800 -m 300

Results:
Setting performance level 1 on adapter 1: engine clock 800MHz, memory clock 300MHz
ADL_Overdrive5_ODPerformanceLevels_Set failed.

OC command - profile 2: DISPLAY=:0.0 /usr/local/bin/atitweak -P 2 -A 1 -e 930 -m 300

Results:
Setting performance level 2 on adapter 1: engine clock 930MHz, memory clock 300MHz
ADL_Overdrive5_ODPerformanceLevels_Set failed.

-------------------------------------------------------------------------------

So as it sits, it looks as if this narrows it down to 5850's from Gigabyte/Sapphire and 6950 from HIS. (from the GPUs in my farm)
I can test a 5970 and some 6870s in a couple days, as they are remote and working without issue under 0.4 so I kind of want to leave them alone. I'm starting to get into my "if it works don't fix it mode". (but what fun is that right?)

I'm providing all of this as feedback, I know it will take time to research and fix. I just want other people to be aware in case they are have my specific issues with those GPUs. (if they actually page through the forum topic) For now I will stick with 0.4 on the rigs with the OC issues and 0.5 on the ones that don't.
hero member
Activity: 626
Merit: 500
Mining since May 2011.
I can confirm that the workaround to apply the patch works and that the patch resolves the Oc issue.

If you don't mind saying, what flavor of card(s) are you running?
hero member
Activity: 910
Merit: 1000
Items flashing here available at btctrinkets.com
I can confirm that the workaround to apply the patch works and that the patch resolves the Oc issue.
hero member
Activity: 616
Merit: 506

When I run /opt/bamt/fixer I get a message "One of the BAMT tools has suffered a fatal error. Not a GLOB reference at /usr/share/perl/5.10/IO/Zlib.pm line 559


https://bitcointalksearch.org/topic/m.767818

Just found it myslef, I really dislike the way this forum works Tongue

yeah it kind of sucks.  well, not this forum anymore than every forum, afaict.

no real way to notify everybody: hey - here is the answer to that thing 50 of you are going to post about.
hero member
Activity: 910
Merit: 1000
Items flashing here available at btctrinkets.com

When I run /opt/bamt/fixer I get a message "One of the BAMT tools has suffered a fatal error. Not a GLOB reference at /usr/share/perl/5.10/IO/Zlib.pm line 559


https://bitcointalksearch.org/topic/m.767818

Just found it myslef, I really dislike the way this forum works Tongue
hero member
Activity: 616
Merit: 506

When I run /opt/bamt/fixer I get a message "One of the BAMT tools has suffered a fatal error. Not a GLOB reference at /usr/share/perl/5.10/IO/Zlib.pm line 559


https://bitcointalksearch.org/topic/m.767818
hero member
Activity: 910
Merit: 1000
Items flashing here available at btctrinkets.com
Im back to square one: the usb-stick problem is gone.

bamt.conf wont apply overclocking, the values set are known stable clocks for this particular card. The card in question is a 5870, core voltage 1.16 core clock 970.
My workaround to this is to stop mining and run:
                                                                  atitweak --set-core-voltage=1.16
                                                                  atitweak --set engine-clock=970
And then restart mining.

Would you mind doing this to confirm bamt is understanding your config:

while mining, in a root shell or from ssh, type:

pcontrol

press d
press f
press enter

you will see a screen like this:



Make sure the O/C info (shaded purple in shot above, but won't be in real life)  matches what you set in bamt.conf

use up down arrow to check all your GPUs

let me know


On my test rig I run pcontrol, hit d and then f and it bombs out like this.
Code:
 BAMT                                                                    miner1Exception in thread Thread-1:
 Display   GPUs   Pools   CoTraceback (most recent call last):
                                                                File "/usr/lib/python2.6/threading.py", linCurrent Pool: pit.deepbit.net
                                                          self.run()
  Hash: 305.334 Mh/s         Pools File: /etc/bamt/pools              File "/opt/bamt/pcontrolADLT.py", liPool Strategy: priority failover
   A/R: 98/0                     Kernel: phatk2f.scr.addstr(top + x,rt, ("%" + s   Rej: 0.00% "s") % " ") Kernel Params: BFI_INT VECTORS FASTLOOP=false AGG..
  Stat: got              error: addstr() returned ERR
   Que: 0/0
  Dlay: 164ms                                                            miner1
   Eff: 33.3%

pcontrol is still very alpha, not intended for normal use at all. 

try making window bigger, sometimes it gets confused.

Sorry for the delayed reply, I was at work.
When I run /opt/bamt/fixer I get a message "One of the BAMT tools has suffered a fatal error. Not a GLOB reference at /usr/share/perl/5.10/IO/Zlib.pm line 559
donator
Activity: 1218
Merit: 1079
Gerald Davis
I have been trying to overclock my cards all night. If I set my voltage to the stock setting it accepts the overclock. Not much of a solution though. Even in ati, it only lets the voltage raise to 1.049 which is stock on the 5970.


I don't think Linux supports overvolting the 5970.  There is a tool called raedonvolt that supposedly can be hacked to sort of work, but it doesn't seem very robust or well supported.  it bypasses ATI's api and mucks around in the card directly.

for example..
http://allmybase.com/2012/01/25/overvolting-the-ati-radeon-5970-in-linux-with-radeonvolt/


Hard to believe with all the people on the forum acting like everyone is crazy who uses windows on dedicated miners.

Due to exponential increase in power combined with heat load on a 3x5970 or 4x5970 rig most people don't overvolt 5970s.  Still this is one area where windows is superior.  You can always do a custom bios flash.
sr. member
Activity: 392
Merit: 250
I have been trying to overclock my cards all night. If I set my voltage to the stock setting it accepts the overclock. Not much of a solution though. Even in ati, it only lets the voltage raise to 1.049 which is stock on the 5970.


I don't think Linux supports overvolting the 5970.  There is a tool called raedonvolt that supposedly can be hacked to sort of work, but it doesn't seem very robust or well supported.  it bypasses ATI's api and mucks around in the card directly.

for example..
http://allmybase.com/2012/01/25/overvolting-the-ati-radeon-5970-in-linux-with-radeonvolt/


Hard to believe with all the people on the forum acting like everyone is crazy who uses windows on dedicated miners.
hero member
Activity: 616
Merit: 506
I have been trying to overclock my cards all night. If I set my voltage to the stock setting it accepts the overclock. Not much of a solution though. Even in ati, it only lets the voltage raise to 1.049 which is stock on the 5970.


I don't think Linux supports overvolting the 5970.  There is a tool called raedonvolt that supposedly can be hacked to sort of work, but it doesn't seem very robust or well supported.  it bypasses ATI's api and mucks around in the card directly.

for example..
http://allmybase.com/2012/01/25/overvolting-the-ati-radeon-5970-in-linux-with-radeonvolt/

Pages:
Jump to: