BAMT version 0.5 - Easy USB based mining Linux with farm wide management tools

[rjk]

Also, you may want to reconsider 100% fan...unless you enjoy replacing fans every few months.

No.

Running fan at 100% is fine.  Changing its speed a lot can in fact lead to premature failure.  Best to set it at some large value and leave alone.

My aching and dying fans have told me otherwise in the past.

I suppose you did an exhaustive study with a large control group of fans not set to 100, compared to a large set of samples with fan set to 100?
Because I did..  and found the fans most likely to die were those on "auto" where they varied in speed.  In fact none of the fans set to 100, 85 or 50 had any failures at all.

No... but 3 5870s at 100% all failed in 2 1.5 months, where 6 5870s with autofan have yet to fail in more than 4 months. I should have a larger set to choose from, true.

[lodcrappo]

Also, you may want to reconsider 100% fan...unless you enjoy replacing fans every few months.

No.

Running fan at 100% is fine.  Changing its speed a lot can in fact lead to premature failure.  Best to set it at some large value and leave alone.

My aching and dying fans have told me otherwise in the past.

I suppose you did an exhaustive study with a large control group of fans not set to 100, compared to a large set of samples with fan set to 100?
Because I did..  and found the fans most likely to die were those on "auto" where they varied in speed.  In fact none of the fans set to 100, 85 or 50 had any failures at all.

No... but 3 5870s at 100% all failed in 2 1.5 months, where 6 5870s with autofan have yet to fail in more than 4 months. I should have a larger set to choose from, true.

The quality of the fans in various samples probably has more to do with fan life than the way that you use them.  However, I do have several models from several manufacturers that have run for nearly 1 year at 100% without any issue.

[mmartoccia]

Hey guys, looking to further optimize my 5970 rig.  I was completely stable prior to the 0.5 release and now having to reboot my box twice daily to keep the cards from faulting.  I was pulling about 330 MHx2 and was hoping someone else who is using a similar card could share an idea or two on what I could to to push up the limit some.   I've also noticed that my second GPU's Mem is clocking at 1000 Mhz when I've told it to stay at 175.  Am I doing something stupid?

Here's my current configuration:

Code:

gpu0:
  disabled: 0
  # optional command to execute prior to overclocking
  pre_oc_cmd: /etc/init.d/munin-node restart

  #overclocking.. all optional 955/830
  fan_speed: 100
  core_speed_0: 730
  core_speed_1: 730
  core_speed_2: 730
  mem_speed_0: 175
  mem_speed_1: 175
  mem_speed_2: 175
  # core_voltage_0: 1.125
  # core_voltage_1: 1.125
  # core_voltage_2: 1.125000

  # optional command to execute after overclocking
  post_oc_cmd:

  # kernel to use.. currently one of:  phatk phatk2 phatk-bamt04 phatk-svn phatk-1.50 poclbm
  kernel: phatk2
  # phoenix style kernel params.  do not include DEVICE=x
  kernel_params: BFI_INT VECTORS FASTLOOP=false AGGRESSION=13 WORKSIZE=128

  # file containing pool URLs
  pool_file: /etc/bamt/pools
  # default time to allow phoenix to run without finding a share.. exceed this and we move to next pool in poolfile
  pool_timeout: 180

  # monitoring values, used by email alerts and gpumon
  monitor_temp_lo: 45
  monitor_temp_hi: 80
  monitor_load_lo: 80
  monitor_hash_lo: 125
  monitor_fan_lo: 2000
  monitor_reject_hi: 2

gpu1:
  disabled: 0
  # optional command to execute prior to overclocking
  pre_oc_cmd:

  #overclocking.. all optional
  # fan_speed: 80
  core_speed_0: 730
  core_speed_1: 730
  core_speed_2: 730
  mem_speed_0: 175
  mem_speed_1: 175
  mem_speed_2: 175
  # core_voltage_0: 1.088
  # core_voltage_1: 1.088
  # core_voltage_2: 1.088

  # optional command to execute after overclocking
  post_oc_cmd:

  # kernel to use.. currently one of:  phatk phatk2 phatk-bamt04 phatk-svn phatk-1.50 poclbm
  kernel: phatk2
  # phoenix style kernel params.  do not include DEVICE=x
  kernel_params: BFI_INT VECTORS FASTLOOP=false AGGRESSION=13 WORKSIZE=128

  # file containing pool URLs
  pool_file: /etc/bamt/pools
  # default time to allow phoenix to run without finding a share.. exceed this and we move to next pool in poolfile
  pool_timeout: 180

  # monitoring values, used by email alerts and gpumon
  monitor_temp_lo: 45
  monitor_temp_hi: 80
  monitor_load_lo: 80
  monitor_hash_lo: 125
  monitor_fan_lo: 2000
  monitor_reject_hi: 2

Im propably the biggest noob who dares give advice here so you might want to wait for confirmation but to me it seems that your feeding stock voltage to a highly underclocked card. It may or may not be the source of your problems, but either way I'd lower the core voltage.

As mentioned multiple times in this thread, do not change profiles 0 and 1 unless you need to.   You can cause instability by OCing the lower profiles.[/size] Profile 2 is what is used during mining.

Loud and clear Lodcrappo, thanks.  - So... can I just comment out profiles 0 and 1 and let them run stock with settings for 2 at what I have them at?  Is that the preferred method?  Thanks again everyone! 

[Red Emerald]

Hey guys, looking to further optimize my 5970 rig.  I was completely stable prior to the 0.5 release and now having to reboot my box twice daily to keep the cards from faulting.  I was pulling about 330 MHx2 and was hoping someone else who is using a similar card could share an idea or two on what I could to to push up the limit some.   I've also noticed that my second GPU's Mem is clocking at 1000 Mhz when I've told it to stay at 175.  Am I doing something stupid?

Here's my current configuration:

Code:

gpu0:
  disabled: 0
  # optional command to execute prior to overclocking
  pre_oc_cmd: /etc/init.d/munin-node restart

  #overclocking.. all optional 955/830
  fan_speed: 100
  core_speed_0: 730
  core_speed_1: 730
  core_speed_2: 730
  mem_speed_0: 175
  mem_speed_1: 175
  mem_speed_2: 175
  # core_voltage_0: 1.125
  # core_voltage_1: 1.125
  # core_voltage_2: 1.125000

  # optional command to execute after overclocking
  post_oc_cmd:

  # kernel to use.. currently one of:  phatk phatk2 phatk-bamt04 phatk-svn phatk-1.50 poclbm
  kernel: phatk2
  # phoenix style kernel params.  do not include DEVICE=x
  kernel_params: BFI_INT VECTORS FASTLOOP=false AGGRESSION=13 WORKSIZE=128

  # file containing pool URLs
  pool_file: /etc/bamt/pools
  # default time to allow phoenix to run without finding a share.. exceed this and we move to next pool in poolfile
  pool_timeout: 180

  # monitoring values, used by email alerts and gpumon
  monitor_temp_lo: 45
  monitor_temp_hi: 80
  monitor_load_lo: 80
  monitor_hash_lo: 125
  monitor_fan_lo: 2000
  monitor_reject_hi: 2

gpu1:
  disabled: 0
  # optional command to execute prior to overclocking
  pre_oc_cmd:

  #overclocking.. all optional
  # fan_speed: 80
  core_speed_0: 730
  core_speed_1: 730
  core_speed_2: 730
  mem_speed_0: 175
  mem_speed_1: 175
  mem_speed_2: 175
  # core_voltage_0: 1.088
  # core_voltage_1: 1.088
  # core_voltage_2: 1.088

  # optional command to execute after overclocking
  post_oc_cmd:

  # kernel to use.. currently one of:  phatk phatk2 phatk-bamt04 phatk-svn phatk-1.50 poclbm
  kernel: phatk2
  # phoenix style kernel params.  do not include DEVICE=x
  kernel_params: BFI_INT VECTORS FASTLOOP=false AGGRESSION=13 WORKSIZE=128

  # file containing pool URLs
  pool_file: /etc/bamt/pools
  # default time to allow phoenix to run without finding a share.. exceed this and we move to next pool in poolfile
  pool_timeout: 180

  # monitoring values, used by email alerts and gpumon
  monitor_temp_lo: 45
  monitor_temp_hi: 80
  monitor_load_lo: 80
  monitor_hash_lo: 125
  monitor_fan_lo: 2000
  monitor_reject_hi: 2

Im propably the biggest noob who dares give advice here so you might want to wait for confirmation but to me it seems that your feeding stock voltage to a highly underclocked card. It may or may not be the source of your problems, but either way I'd lower the core voltage.

As mentioned multiple times in this thread, do not change profiles 0 and 1 unless you need to.   You can cause instability by OCing the lower profiles.[/size] Profile 2 is what is used during mining.

Loud and clear Lodcrappo, thanks.  - So... can I just comment out profiles 0 and 1 and let them run stock with settings for 2 at what I have them at?  Is that the preferred method?  Thanks again everyone! 

Leave

Code:

  mem_speed_0: 175
  mem_speed_1: 175
  mem_speed_2: 175

But you don't need to set the clocks or voltage for profiles 0 and 1

[Coinoisseur]

Side topic comment: I leave my fans at 86% fixed, haven't had the same trouble as some others with the 5830 Sapphire (BitcoinXtreme versions lol). Small sample size but judging from what I've read from datacenter studies, Google has published several, reliability isn't that different between low and high for most computer parts. Main exception if I remember right was hard drives which both Google and Microsoft have found are the least resilient to high temperature and constant use.

[dizzy1]

So I have a couple machines mining with bamt, but on machine will not report mining stats to gpumon. The api ports, address and passwords match. Where should I look to see why its not reporting? Also is there a way in bamt.conf to set intensity for cgminer?

bamt.conf

Code:

settings:
  miner_id: oldhda
  miner_loc: ...
  do_monitor: 1
  do_bcast_status: 1

  smtp_host: ...
  smtp_to: ...
  smtp_auth_user: ...
  smtp_auth_pass: ....
  smtp_tls: 1
  smtp_ssl: 1
  do_mgpumon: 0
  do_autoconf_client: 1
  detect_defunct: 1

  phoenix2: 1
  phoenix2_config: /etc/bamt/phoenix2.conf
  phoenix2_port: 7789
  phoenix2_pass: bamt


gpu0:
  disabled: 0
  phoenix2: 1

  # core_speed_0: 300
  # core_speed_1: 800
  core_speed_2: 900

  mem_speed_0: 175
  mem_speed_1: 175
  mem_speed_2: 175

  fan_speed: 60

  kernel: phatk2
  kernel_params: BFI_INT VECTORS FASTLOOP=false AGGRESSION=9

  pool_file: /etc/bamt/pools

  pool_timeout: 180
  monitor_temp_lo: 45
  monitor_temp_hi: 80
  monitor_load_lo: 80
  monitor_hash_lo: 350
  monitor_fan_lo: 2000
  monitor_reject_hi: 2

phoenix.conf

Code:

[general]
    autodetect = +cl -cpu #Use autodetect for all OpenCL devices, except those which are CPUs.
    verbose = True #Enable verbose logging?
    backend = http://...@pit.deepbit.net:8332 #The primary backend.
    # backups = http://user2:password2@server2.com:8332 http://bitcoin:bitcoin@localhost:8332 #A space seperated list of backup servers.
    failback = 600 #Seconds between attempts to reconnect to primary backend when using backups. (0 to disable)
    queuesize = 3 #Target/maximum size of the queue
    queuedelay = 5 #Seconds before work expires to request more work (WARNING: don't change this unless you know what you are doing!)
    statusinterval = 1 #Seconds between statusbar updates
    ratesamples = 10 #Number of samples to average for hashrate reporting
    # logfile =  #Set this option to log to a file.
[web]
    disabled = False #Disable the RPC server?
    bind = 127.0.0.1 #IP to bind the RPC server to
    port = 7789 #RPC port MUST MATCH SETTING IN BAMT.CONF
    password = bamt #RPC password  MUST MATCH SETTING IN BAMT.CONF
    root = /var/www/phoenix #Root directory for the web server
    logbuffer = 1000 #How many logs to remember in the getlogs() RPC call

# Hey, you.  yeah, you. You must fix the stuff below to match your GPUs!

[cl:0:0]
    autoconfigure = False #Automatically configure this device?
    kernel = phatk2 #The kernel to use for the device
    name = GPU 0 #The name to display the device as (default is device ID if not specified)
    start_undetected = False #Do not start the kernel if the device is not present
    disabled = False #Disable this device?
    worksize = 128 #Work group size, tweaking this option may improve performance
    vectors = True #Use unit2 vectors? (enable this or vectors4, not both)
    vectors4 = False #Use uint4 vectors? (enable this or vectors, not both)
    bfi_int = True #Use BFI_INT instruction on ATI VLIW GPUs?
    goffset = True #Use OpenCL 1.1 global offset?
    fastloop = False #Use fast internal loop? (ideal for low aggression)
    aggression = 9 #Number of nonces to test per kernel execution (lower value = less desktop lag, higher value = higher hashrate)

[boozer]

No don't wait a week for the 7990's to come out.. much more worth the money. 1.3Ghash per card cost 849.99

Thats why I am selling of a lot of my 5 series and 7970's

I assume you mean "wait a week" instead of "don't wait a week"   

Crap... wish i would have researched better... 7990. 1.3Ghash per card... $849.99.... thats pretty sweet.  Of course, i wouldn't have had 6 7970's mining for a couple months, so I still came out alright by buying them earlier.... I'll wait for official specs/hashrate/power/temp of the 7990's, but I might very well upgrade as well here pretty soon.

[bulanula]

I don't quite see why running 100% would be better than autofan setting

100% gets the bearings hotter -> oil evaporates at high temps -> bearing damaged -> fan fails / rattles

Autofan can only prolong life of fans I think ...

[Joshwaa]

No don't wait a week for the 7990's to come out.. much more worth the money. 1.3Ghash per card cost 849.99

Thats why I am selling of a lot of my 5 series and 7970's

I assume you mean "wait a week" instead of "don't wait a week"   

Crap... wish i would have researched better... 7990. 1.3Ghash per card... $849.99.... thats pretty sweet.  Of course, i wouldn't have had 6 7970's mining for a couple months, so I still came out alright by buying them earlier.... I'll wait for official specs/hashrate/power/temp of the 7990's, but I might very well upgrade as well here pretty soon.

Yea meant wait a week. I have 6 7970's also funny. I will be first in line for the 7990s!

[01BTC10]

I don't quite see why running 100% would be better than autofan setting

100% gets the bearings hotter -> oil evaporates at high temps -> bearing damaged -> fan fails / rattles

Autofan can only prolong life of fans I think ...

Maybe it's like a car. It is harder for mechanic if you live in a city when driving at irregular speed compared to rural area where you spend more time on the highway.

[DeathAndTaxes]

Crap... wish i would have researched better... 7990. 1.3Ghash per card... $849.99.... thats pretty sweet.  Of course, i wouldn't have had 6 7970's mining for a couple months, so I still came out alright by buying them earlier.... I'll wait for official specs/hashrate/power/temp of the 7990's, but I might very well upgrade as well here pretty soon.

Not sure why people think a 7990 will get 1.3 GH/s?

A 5970 doesn't get 2x the hashrate of a 5870   
A 6990 doesn't get 2x the hashrate of a 6970

An overclocked 5870 can push 450 MH/s.  I wish my 4x5970 rigs pushed 3.6 GH/s.  They push about 3GH/s or ~85% of what a 8x5870 would push.

The 7990 should get >1GH/s.  Maybe 1.1GH/s. I doubt it goes much higher on air.

[lodcrappo]

I don't quite see why running 100% would be better than autofan setting

100% gets the bearings hotter -> oil evaporates at high temps -> bearing damaged -> fan fails / rattles

Autofan can only prolong life of fans I think ...

this is questionable theory.  fan turning faster means more airflow to cool bearings.  its possible that bearing temp is much higher at 30 or 50% than 100.
i don't think anyone has done enough actual research to make authoritative claims.  I can only offer the information found in my own use, which involved 30 some GPUs from mixed vendors and models over the course of nearly one year.  failure rate on fans using variable speed was nearly 50%.  no failures in the fans using fixed (high) speeds.  whether this was coincidental, or due to the slower speeds the variable fans used, or simply the fact that they ran at many different speeds, or some other factor, who knows.
 

[bulanula]

I don't quite see why running 100% would be better than autofan setting

100% gets the bearings hotter -> oil evaporates at high temps -> bearing damaged -> fan fails / rattles

Autofan can only prolong life of fans I think ...

this is questionable theory.  fan turning faster means more airflow to cool bearings.  its possible that bearing temp is much higher at 30 or 50% than 100.
i don't think anyone has done enough actual research to make authoritative claims.  I can only offer the information found in my own use, which involved 30 some GPUs from mixed vendors and models over the course of nearly one year.  failure rate on fans using variable speed was nearly 50%.  no failures in the fans using fixed (high) speeds.  whether this was coincidental, or due to the slower speeds the variable fans used, or simply the fact that they ran at many different speeds, or some other factor, who knows.

I would like to see D&T's opinion on this.

So you would suggest leaving fans set to whatever speed if autofan / variable is prone to failures ? 60% all the time OR automatic ATI setting ?

Thanks !

[lodcrappo]

I don't quite see why running 100% would be better than autofan setting

100% gets the bearings hotter -> oil evaporates at high temps -> bearing damaged -> fan fails / rattles

Autofan can only prolong life of fans I think ...

this is questionable theory.  fan turning faster means more airflow to cool bearings.  its possible that bearing temp is much higher at 30 or 50% than 100.
i don't think anyone has done enough actual research to make authoritative claims.  I can only offer the information found in my own use, which involved 30 some GPUs from mixed vendors and models over the course of nearly one year.  failure rate on fans using variable speed was nearly 50%.  no failures in the fans using fixed (high) speeds.  whether this was coincidental, or due to the slower speeds the variable fans used, or simply the fact that they ran at many different speeds, or some other factor, who knows.

I would like to see D&T's opinion on this.

So you would suggest leaving fans set to whatever speed if autofan / variable is prone to failures ? 60% all the time OR automatic ATI setting ?

Thanks !

I've had great luck simply setting the fans to 100% and not thinking about it again.  Whether this works as well for you, I cannot predict.

For me, using software to control the fan speed has always seemed risky.  When GPUs are on the edge of a lock up, they tend to ignore software control.  This means you lose the ability to increase the fan at exactly the worst time.    If you must have "auto fan" I'd highly recommend using a bios editor to set an appropriate temp/speed curve right on the card and letting it deal with the fan directly.

[DeathAndTaxes]

So you would suggest leaving fans set to whatever speed if autofan / variable is prone to failures ? 60% all the time OR automatic ATI setting ?

Honestly I don't know.  Most of the "data" we have is anecdotal.

Rapid changes in fan speed are likely damaging but GPU temps tend to rise slowly.  So with a temp-hysteresis of 2+ the fan shouldn't change speed that often or that rapidly.  Fans running "too slow" will likely cause unnecessary wear but who knows what too slow is.  One thing I do know is that on 5970 the ATI/AMD fan slope is too "weak".  AMD seems to be optimizing for noise not temps so I wouldn't recommend that.

Given all the variables of fan speed, ambient temp, fan quality, normal variance, air quality (dirt) I honestly don't know.

I set all my fans at 80% fixed (65% in winter) because they are in the garage and I can't hear the noise.  I use auto-gpu (throttle clocks not fan to keep temps below target) because until recently 5970s didn't play nice with "auto-fan".  Of 24 GPUs I have had 3 fan failures but they were all on used GPUs (and one was less 1 month after purchase) so who knows how they were abused before I got them.  I see no reason to change my setup since it works.

[lodcrappo]

So you would suggest leaving fans set to whatever speed if autofan / variable is prone to failures ? 60% all the time OR automatic ATI setting ?

Honestly I don't know.  Most of the "data" we have is anecdotal.

Rapid changes in fan speed are likely damaging but GPU temps tend to rise slowly.  So with a temp-hysteresis of 2+ the fan shouldn't change speed that often or that rapidly.  Fans running "too slow" will likely cause unnecessary wear but who knows what too slow is.  One thing I do know is that on 5970 the ATI/AMD fan slope is too "weak".  AMD seems to be optimizing for noise not temps so I wouldn't recommend that.

Given all the variables of fan speed, ambient temp, fan quality, normal variance, air quality (dirt) I honestly don't know.

I set all my fans at 80% fixed (65% in winter) because they are in the garage and I can't hear the noise.  I use auto-gpu (throttle clocks not fan to keep temps below target) because until recently 5970s didn't play nice with "auto-fan".  Of 24 GPUs I have had 3 fan failures but they were all on used GPUs (and one was less 1 month after purchase) so who knows how they were abused before I got them.  I see no reason to change my setup since it works.

agreed, all we really have is anecdotal evidence.  can't hurt to share it though. patterns may emerge if enough experiences are known.

It's not just 5970s.. the built in fan slope on every GPU I've tested is far too low to be useful in mining.  However, you can easily edit this slope with a bios editor.  If someone wants auto fan speed, I think this would be the best approach.

[Joshwaa]

Crap... wish i would have researched better... 7990. 1.3Ghash per card... $849.99.... thats pretty sweet.  Of course, i wouldn't have had 6 7970's mining for a couple months, so I still came out alright by buying them earlier.... I'll wait for official specs/hashrate/power/temp of the 7990's, but I might very well upgrade as well here pretty soon.

Not sure why people think a 7990 will get 1.3 GH/s?

A 5970 doesn't get 2x the hashrate of a 5870   
A 6990 doesn't get 2x the hashrate of a 6970

An overclocked 5870 can push 450 MH/s.  I wish my 4x5970 rigs pushed 3.6 GH/s.  They push about 3GH/s or ~85% of what a 8x5870 would push.

The 7990 should get >1GH/s.  Maybe 1.1GH/s. I doubt it goes much higher on air.

Well I have 6 7970's all of witch get over 700MH/s so yes 2 x 700 = 1400 (1.4 GH/s)
 So you are saying that a 7990 will not get 1.3? Would you care to place a wager? I know that my 7970's at 1GHZ get about 630MH/s-650MH/s. So a stock clocked @ 1GHZ 7990 will not do 1.3 when overclocked? hmmmm... Sounds like a bet.

[DeathAndTaxes]

No a 7990 @ 1 Ghz will do 2x what a 7970 @ 1 Ghz will do however I don't think a 7990 will do 1 Ghz (based on track record of 5970 and 6990).

Given stock clock on a 7970 is 925 Mhz and it achieves 1 GHz using 4 phase power it is unlikely that with the limits of 3 phase power and double GPU cooling the 7970 will be able to achieve a 1 Ghz clock (mining stable 24/7 for weeks).

I have no interest in betting.  We will find out soon enough.  To avoid this thread going off topic I will leave it at that.  The likely performance of 7990 can be debated in another thread (new or existing).

[pocketartillery]

I have been testing BAMT for a while on my rigs and I am getting some odd behavior from some of them.

All the rigs are identical in that they all use the same motherboard, processor, power supplies, ram, flash drives and cards.

Problem #1
1 out of 9 of the rigs will not overclock any of the cards.  All 4 Sapphire 5830's are at 246Mhps.  There is nothing in the BAMT/live/control/active directory to clear.
Problem #2
1 out of the 9 rigs will not overclock any cards except for card 0. There is nothing in the BAMT/live/control/active directory to clear.
Problem #3
This has only happened twice and a quick reboot fixed it I just thought it was odd.  Two different rigs began mining with 1 less card than was shown in the upper right corner of the desktop.  Ie if 4 cards were detected by the OS it would only mine with 3 and didn't show the other in GPUmon etc.

Thanks for your help!

[bobers]

how can i manualy switch between primary and backup pool in gpumon?