Page 1 of 1
Linux: Use OnboardVGA for screen instead of GPU for more PPD
Posted: Wed Dec 18, 2013 8:37 am
by mxplm
Hi everyone,
I have a system with an Intel Core i3 550 and an NVIDIA GTX 650 Ti crunching for f@h. Currently I'm using the GTX for screen output as well. I tried using the integrated Intel graphics chip (iGPU) for that, hoping to get some extra PPD. To do this, I changed the BIOS setting for the Onboard VGA from "only activate when there is no external GPU detected" to "always active". The iGPU worked but, I ran into problems because the NVIDIA driver wasn't loaded. Consequently, the F@H GPU task wasn't processed.
First question: Does it pay to set this up?
Follow-up question: How can I load the NVIDIA driver when I'm using the iGPU for screen output?
Machine specs:
Intel Core i3 550 @ OC 3.84 GHz, 2 cores + HT enabled, (Noctua NH-C12P-SE14 cooler)
8 GB RAM
Gigabyte H55M-USB3 Board
NVIDIA GTX 650 Ti
Samsung 840 Pro SSD
BeQuiet! 430W power supply
50MBit/s DSL internet connection
Software:
Ubuntu Linux 13.10 amd64
NVIDIA Driver 319.32 for Linux 64 Bit
FAHClient 7.3.6 amd64
- CPU core a4 currently running project 7048
- GPU core 17 currently running project 8900
FAHControl 7.3.6-1 all
FAHView 7.3.6 amd64
Regards, mxplm
Re: Using Onboard VGA for screen instead of GPU to increase
Posted: Wed Dec 18, 2013 10:27 am
by P5-133XL
Intel onboard GPU's are not supported for folding. The only real benefit on having the onboard running the screen is lack of lag (The screen gets refreshed immediately rather than needing to wait for folding to finish doing its current work). Having two video cards on a system does significantly complicate configuration though. I will bow out to someone with much more experience with Linux in helping you with Linux driver and multiple video card issues. Even beyond Linux, there may be GPU indexing issues within the folding configuration causing problems.
It could be quite useful if you supplied a copy of the current folding log, specifically including the system and config portions (Click refresh before copy & pasting).
Re: Using Onboard VGA for screen instead of GPU to increase
Posted: Wed Dec 18, 2013 10:47 am
by mxplm
Thanks for your thoughts. I will supply the logs when I'm back home. Meanwhile, I just wanted to say I don't plan to crunch with my onboard VGA
What I was hoping for was getting more PPD from my GTX by moving the "burden" of rendering the screen to somewhere else. That way the GTX could be fully dedicated to F@H, also reducing scheduling and memory overhead. "Lack of lags" would be another benefit, of course
Re: Using Onboard VGA for screen instead of GPU to increase
Posted: Wed Dec 18, 2013 11:24 am
by HaloJones
Yes. it will make a difference. It's easy to setup in Windows but I have no idea if that's true in Linux, sorry.
Re: Using Onboard VGA for screen instead of GPU to increase
Posted: Wed Dec 18, 2013 11:54 am
by ChristianVirtual
I managed that with 90% success rate :
IGPU for UI and dedicated 660 Ti for folding. Needed to play with xorg.conf
Jesse_V started a thread; I added my config there.
viewtopic.php?nomobile=1&f=89&t=25075&p=252474#p252474
Only downside: the menus are not visible and I need to start a shell window via keyboard or ssh.
But else it folding incl fan control. Enough for me and as dedicated rig
Re: Using Onboard VGA for screen instead of GPU to increase
Posted: Wed Dec 18, 2013 12:03 pm
by mxplm
Ok, so I don't do this for nothing, thank you
Here are the logs I promised:
1.) Default configuration: GTX crunching and connected to screen, Onboard VGA disabled, no problems:
Code: Select all
*********************** Log Started 2013-12-18T11:44:36Z ***********************
11:44:36:Started thread 1 on PID 992
11:44:36:************************* Folding@home Client *************************
11:44:36: Website: http://folding.stanford.edu/
11:44:36: Copyright: (c) 2009-2013 Stanford University
11:44:36: Author: Joseph Coffland <joseph@cauldrondevelopment.com>
11:44:36: Args: --child --lifeline 940 /etc/fahclient/config.xml --run-as fahclient
11:44:36: --pid-file=/var/run/fahclient.pid --daemon
11:44:36: Config: /etc/fahclient/config.xml
11:44:36:******************************** Build ********************************
11:44:36: Version: 7.3.6
11:44:36: Date: Feb 18 2013
11:44:36: Time: 07:24:08
11:44:36: SVN Rev: 3923
11:44:36: Branch: fah/trunk/client
11:44:36: Compiler: GNU 4.4.7
11:44:36: Options: -std=gnu++98 -O3 -funroll-loops -mfpmath=sse -ffast-math
11:44:36: -fno-unsafe-math-optimizations -msse2
11:44:36: Platform: linux2 3.2.0-1-amd64
11:44:36: Bits: 64
11:44:36: Mode: Release
11:44:36:******************************* System ********************************
11:44:36: CPU: Intel(R) Core(TM) i3 CPU 550 @ 3.20GHz
11:44:36: CPU ID: GenuineIntel Family 6 Model 37 Stepping 5
11:44:36: CPUs: 4
11:44:36: Memory: 7.73GiB
11:44:36:Free Memory: 7.24GiB
11:44:36: Threads: POSIX_THREADS
11:44:36:Has Battery: false
11:44:36: On Battery: false
11:44:36: UTC offset: 1
11:44:36: PID: 992
11:44:36: CWD: /var/lib/fahclient
11:44:36: OS: Linux 3.11.0-14-generic x86_64
11:44:36: OS Arch: AMD64
11:44:36: GPUs: 1
11:44:36: GPU 0: NVIDIA:3 GK106 [GeForce GTX 650 Ti]
11:44:36: CUDA: Not detected
11:44:36:***********************************************************************
11:44:36:<config>
11:44:36: <!-- Client Control -->
11:44:36: <client-threads v='4'/>
11:44:36: <cycle-rate v='4'/>
11:44:36: <cycles v='-1'/>
11:44:36: <data-directory v='.'/>
11:44:36: <disable-sleep-when-active v='true'/>
11:44:36: <exec-directory v='/usr/bin'/>
11:44:36: <exit-when-done v='false'/>
11:44:36: <fold-anon v='false'/>
11:44:36: <idle-seconds v='300'/>
11:44:36: <open-web-control v='false'/>
11:44:36:
11:44:36: <!-- Configuration -->
11:44:36: <config-rotate v='true'/>
11:44:36: <config-rotate-dir v='configs'/>
11:44:36: <config-rotate-max v='16'/>
11:44:36:
11:44:36: <!-- Debugging -->
11:44:36: <assignment-servers>
11:44:36: assign3.stanford.edu:8080 assign4.stanford.edu:80
11:44:36: </assignment-servers>
11:44:36: <capture-directory v='capture'/>
11:44:36: <capture-on-error v='false'/>
11:44:36: <capture-packets v='false'/>
11:44:36: <capture-requests v='false'/>
11:44:36: <capture-responses v='false'/>
11:44:36: <capture-sockets v='false'/>
11:44:36: <debug-sockets v='false'/>
11:44:36: <exception-locations v='true'/>
11:44:36: <gpu-assignment-servers>
11:44:36: assign-GPU.stanford.edu:80 assign-GPU.stanford.edu:8080
11:44:36: </gpu-assignment-servers>
11:44:36: <stack-traces v='false'/>
11:44:36:
11:44:36: <!-- Error Handling -->
11:44:36: <max-slot-errors v='5'/>
11:44:36: <max-unit-errors v='5'/>
11:44:36:
11:44:36: <!-- Folding Core -->
11:44:36: <checkpoint v='3'/>
11:44:36: <core-dir v='cores'/>
11:44:36: <core-priority v='idle'/>
11:44:36: <cpu-affinity v='true'/>
11:44:36: <cpu-usage v='100'/>
11:44:36: <gpu-usage v='100'/>
11:44:36: <no-assembly v='false'/>
11:44:36:
11:44:36: <!-- Folding Slot Configuration -->
11:44:36: <cause v='ANY'/>
11:44:36: <client-subtype v='LINUX'/>
11:44:36: <client-type v='normal'/>
11:44:36: <cpu-species v='X86_PENTIUM_II'/>
11:44:36: <cpu-type v='AMD64'/>
11:44:36: <cpus v='-1'/>
11:44:36: <cuda-index v='0'/>
11:44:36: <gpu v='false'/>
11:44:36: <max-packet-size v='normal'/>
11:44:36: <opencl-index v='0'/>
11:44:36: <os-species v='UNKNOWN'/>
11:44:36: <os-type v='LINUX'/>
11:44:36: <power v='off'/>
11:44:36: <project-key v='0'/>
11:44:36: <smp v='true'/>
11:44:36:
11:44:36: <!-- HTTP Server -->
11:44:36: <allow v='127.0.0.1'/>
11:44:36: <connection-timeout v='60'/>
11:44:36: <deny v='0/0'/>
11:44:36: <http-addresses v='0:7396'/>
11:44:36: <https-addresses v=''/>
11:44:36: <max-connect-time v='900'/>
11:44:36: <max-connections v='800'/>
11:44:36: <max-request-length v='52428800'/>
11:44:36: <min-connect-time v='300'/>
11:44:36: <threads v='4'/>
11:44:36:
11:44:36: <!-- Logging -->
11:44:36: <log v='log.txt'/>
11:44:36: <log-color v='true'/>
11:44:36: <log-crlf v='false'/>
11:44:36: <log-date v='false'/>
11:44:36: <log-date-periodically v='21600'/>
11:44:36: <log-debug v='true'/>
11:44:36: <log-domain v='false'/>
11:44:36: <log-header v='true'/>
11:44:36: <log-level v='true'/>
11:44:36: <log-no-info-header v='true'/>
11:44:36: <log-redirect v='false'/>
11:44:36: <log-rotate v='true'/>
11:44:36: <log-rotate-dir v='logs'/>
11:44:36: <log-rotate-max v='16'/>
11:44:36: <log-short-level v='false'/>
11:44:36: <log-simple-domains v='true'/>
11:44:36: <log-thread-id v='false'/>
11:44:36: <log-thread-prefix v='true'/>
11:44:36: <log-time v='true'/>
11:44:36: <log-to-screen v='true'/>
11:44:36: <log-truncate v='false'/>
11:44:36: <verbosity v='5'/>
11:44:36:
11:44:36: <!-- Network -->
11:44:36: <proxy v=':8080'/>
11:44:36: <proxy-enable v='false'/>
11:44:36: <proxy-pass v=''/>
11:44:36: <proxy-user v=''/>
11:44:36:
11:44:36: <!-- Process Control -->
11:44:36: <child v='true'/>
11:44:36: <daemon v='true'/>
11:44:36: <pid v='false'/>
11:44:36: <pid-file v='/var/run/fahclient.pid'/>
11:44:36: <respawn v='false'/>
11:44:36: <service v='false'/>
11:44:36:
11:44:36: <!-- Remote Command Server -->
11:44:36: <command-address v='0.0.0.0'/>
11:44:36: <command-allow-no-pass v='127.0.0.1'/>
11:44:36: <command-deny-no-pass v='0/0'/>
11:44:36: <command-port v='36330'/>
11:44:36:
11:44:36: <!-- Slot Control -->
11:44:36: <idle v='false'/>
11:44:36: <max-shutdown-wait v='60'/>
11:44:36: <pause-on-battery v='false'/>
11:44:36: <pause-on-start v='false'/>
11:44:36:
11:44:36: <!-- User Information -->
11:44:36: <machine-id v='0'/>
11:44:36: <passkey v='********************************'/>
11:44:36: <team v='3'/>
11:44:36: <user v='mxplm'/>
11:44:36:
11:44:36: <!-- Web Server -->
11:44:36: <session-timeout v='3600'/>
11:44:36: <web-allow v='127.0.0.1'/>
11:44:36: <web-deny v='0/0'/>
11:44:36:
11:44:36: <!-- Work Unit Control -->
11:44:36: <dump-after-deadline v='true'/>
11:44:36: <max-queue v='16'/>
11:44:36: <max-units v='0'/>
11:44:36: <next-unit-percentage v='99'/>
11:44:36:
11:44:36: <!-- Folding Slots -->
11:44:36: <slot id='0' type='CPU'>
11:44:36: <cpus v='1'/>
11:44:36: </slot>
11:44:36: <slot id='1' type='GPU'>
11:44:36: <client-type v='advanced'/>
11:44:36: </slot>
11:44:36:</config>
11:44:36:Switching to user fahclient
11:44:36:Trying to access database...
11:44:36:Successfully acquired database lock
11:44:36:Enabled folding slot 00: PAUSED cpu:1 (paused)
11:44:36:Enabled folding slot 01: PAUSED gpu:0:GK106 [GeForce GTX 650 Ti] (paused)
11:44:36:Started thread 5 on PID 992
11:44:36:Started thread 6 on PID 992
11:44:36:Started thread 7 on PID 992
11:44:36:Started thread 8 on PID 992
11:45:28:Started thread 9 on PID 992
11:45:34:WU02:FS01:Starting
11:45:34:WU02:FS01:Running FahCore: /usr/bin/FAHCoreWrapper /var/lib/fahclient/cores/www.stanford.edu/~pande/Linux/AMD64/NVIDIA/Fermi/Core_17.fah/FahCore_17 -dir 02 -suffix 01 -version 703 -lifeline 992 -checkpoint 3 -gpu 0 -gpu-vendor nvidia
11:45:34:WU02:FS01:Started FahCore on PID 3652
11:45:34:Started thread 10 on PID 992
11:45:34:WU02:FS01:Core PID:3656
11:45:34:WU02:FS01:FahCore 0x17 started
11:45:34:WU01:FS00:Starting
11:45:34:WU01:FS00:Running FahCore: /usr/bin/FAHCoreWrapper /var/lib/fahclient/cores/www.stanford.edu/~pande/Linux/AMD64/Core_a4.fah/FahCore_a4 -dir 01 -suffix 01 -version 703 -lifeline 992 -checkpoint 3
11:45:34:WU01:FS00:Started FahCore on PID 3657
11:45:34:Started thread 11 on PID 992
11:45:34:WU01:FS00:Core PID:3661
11:45:34:WU01:FS00:FahCore 0xa4 started
11:45:35:WU02:FS01:0x17:*********************** Log Started 2013-12-18T11:45:34Z ***********************
11:45:35:WU02:FS01:0x17:Project: 8900 (Run 812, Clone 4, Gen 32)
11:45:35:WU02:FS01:0x17:Unit: 0x00000030028c126651a6db57d693b6d9
11:45:35:WU02:FS01:0x17:CPU: 0x00000000000000000000000000000000
11:45:35:WU02:FS01:0x17:Machine: 1
11:45:35:WU02:FS01:0x17:Digital signatures verified
11:45:35:WU02:FS01:0x17: Found a checkpoint file
11:45:35:WU01:FS00:0xa4:
11:45:35:WU01:FS00:0xa4:*------------------------------*
11:45:35:WU01:FS00:0xa4:Folding@Home Gromacs GB Core
11:45:35:WU01:FS00:0xa4:Version 2.27 (Dec. 15, 2010)
11:45:35:WU01:FS00:0xa4:
11:45:35:WU01:FS00:0xa4:Preparing to commence simulation
11:45:35:WU01:FS00:0xa4:- Looking at optimizations...
11:45:35:WU01:FS00:0xa4:- Files status OK
11:45:35:WU01:FS00:0xa4:- Expanded 135714 -> 306920 (decompressed 226.1 percent)
11:45:35:WU01:FS00:0xa4:Called DecompressByteArray: compressed_data_size=135714 data_size=306920, decompressed_data_size=306920 diff=0
11:45:35:WU01:FS00:0xa4:- Digital signature verified
11:45:35:WU01:FS00:0xa4:
11:45:35:WU01:FS00:0xa4:Project: 7048 (Run 0, Clone 716, Gen 4)
11:45:35:WU01:FS00:0xa4:
11:45:35:WU01:FS00:0xa4:Assembly optimizations on if available.
11:45:35:WU01:FS00:0xa4:Entering M.D.
11:45:41:WU01:FS00:0xa4:Using Gromacs checkpoints
11:45:41:WU01:FS00:0xa4:Resuming from checkpoint
11:45:41:WU01:FS00:0xa4:Verified 01/wudata_01.log
11:45:41:WU01:FS00:0xa4:Verified 01/wudata_01.trr
11:45:41:WU01:FS00:0xa4:Verified 01/wudata_01.xtc
11:45:41:WU01:FS00:0xa4:Verified 01/wudata_01.edr
11:45:41:WU01:FS00:0xa4:Completed 1221955 out of 25000000 steps (4%)
11:46:35:Removing old file 'configs/config-20131217-124611.xml'
11:46:35:Saving configuration to /etc/fahclient/config.xml
11:46:35:<config>
11:46:35: <!-- Folding Core -->
11:46:35: <checkpoint v='3'/>
11:46:35: <cpu-affinity v='true'/>
11:46:35:
11:46:35: <!-- Folding Slot Configuration -->
11:46:35: <power v='full'/>
11:46:35:
11:46:35: <!-- Logging -->
11:46:35: <verbosity v='5'/>
11:46:35:
11:46:35: <!-- Network -->
11:46:35: <proxy v=':8080'/>
11:46:35:
11:46:35: <!-- Slot Control -->
11:46:35: <pause-on-battery v='false'/>
11:46:35:
11:46:35: <!-- User Information -->
11:46:35: <passkey v='********************************'/>
11:46:35: <team v='3'/>
11:46:35: <user v='mxplm'/>
11:46:35:
11:46:35: <!-- Folding Slots -->
11:46:35: <slot id='0' type='CPU'>
11:46:35: <cpus v='1'/>
11:46:35: </slot>
11:46:35: <slot id='1' type='GPU'>
11:46:35: <client-type v='advanced'/>
11:46:35: </slot>
11:46:35:</config>
11:47:17:WU02:FS01:0x17:Completed 1300000 out of 2500000 steps (52%)
2.) New configuration: Onboard VGA enabled and connected to screen, GTX should be crunching but FahCore returns error code: UNKNOWN_ENUM (127 = 0x7f)
Code: Select all
*********************** Log Started 2013-12-18T11:40:43Z ***********************
11:40:43:Started thread 1 on PID 1123
11:40:43:************************* Folding@home Client *************************
11:40:43: Website: http://folding.stanford.edu/
11:40:43: Copyright: (c) 2009-2013 Stanford University
11:40:43: Author: Joseph Coffland <joseph@cauldrondevelopment.com>
11:40:43: Args: --child --lifeline 916 /etc/fahclient/config.xml --run-as fahclient
11:40:43: --pid-file=/var/run/fahclient.pid --daemon
11:40:43: Config: /etc/fahclient/config.xml
11:40:43:******************************** Build ********************************
11:40:43: Version: 7.3.6
11:40:43: Date: Feb 18 2013
11:40:43: Time: 07:24:08
11:40:43: SVN Rev: 3923
11:40:43: Branch: fah/trunk/client
11:40:43: Compiler: GNU 4.4.7
11:40:43: Options: -std=gnu++98 -O3 -funroll-loops -mfpmath=sse -ffast-math
11:40:43: -fno-unsafe-math-optimizations -msse2
11:40:43: Platform: linux2 3.2.0-1-amd64
11:40:43: Bits: 64
11:40:43: Mode: Release
11:40:43:******************************* System ********************************
11:40:43: CPU: Intel(R) Core(TM) i3 CPU 550 @ 3.20GHz
11:40:43: CPU ID: GenuineIntel Family 6 Model 37 Stepping 5
11:40:43: CPUs: 4
11:40:43: Memory: 7.60GiB
11:40:43:Free Memory: 7.04GiB
11:40:43: Threads: POSIX_THREADS
11:40:43:Has Battery: false
11:40:43: On Battery: false
11:40:43: UTC offset: 1
11:40:43: PID: 1123
11:40:43: CWD: /var/lib/fahclient
11:40:43: OS: Linux 3.11.0-14-generic x86_64
11:40:43: OS Arch: AMD64
11:40:43: GPUs: 1
11:40:43: GPU 0: NVIDIA:3 GK106 [GeForce GTX 650 Ti]
11:40:43: CUDA: Not detected
11:40:43:***********************************************************************
11:40:43:<config>
11:40:43: <!-- Client Control -->
11:40:43: <client-threads v='4'/>
11:40:43: <cycle-rate v='4'/>
11:40:43: <cycles v='-1'/>
11:40:43: <data-directory v='.'/>
11:40:43: <disable-sleep-when-active v='true'/>
11:40:43: <exec-directory v='/usr/bin'/>
11:40:43: <exit-when-done v='false'/>
11:40:43: <fold-anon v='false'/>
11:40:43: <idle-seconds v='300'/>
11:40:43: <open-web-control v='false'/>
11:40:43:
11:40:43: <!-- Configuration -->
11:40:43: <config-rotate v='true'/>
11:40:43: <config-rotate-dir v='configs'/>
11:40:43: <config-rotate-max v='16'/>
11:40:43:
11:40:43: <!-- Debugging -->
11:40:43: <assignment-servers>
11:40:43: assign3.stanford.edu:8080 assign4.stanford.edu:80
11:40:43: </assignment-servers>
11:40:43: <capture-directory v='capture'/>
11:40:43: <capture-on-error v='false'/>
11:40:43: <capture-packets v='false'/>
11:40:43: <capture-requests v='false'/>
11:40:43: <capture-responses v='false'/>
11:40:43: <capture-sockets v='false'/>
11:40:43: <debug-sockets v='false'/>
11:40:43: <exception-locations v='true'/>
11:40:43: <gpu-assignment-servers>
11:40:43: assign-GPU.stanford.edu:80 assign-GPU.stanford.edu:8080
11:40:43: </gpu-assignment-servers>
11:40:43: <stack-traces v='false'/>
11:40:43:
11:40:43: <!-- Error Handling -->
11:40:43: <max-slot-errors v='5'/>
11:40:43: <max-unit-errors v='5'/>
11:40:43:
11:40:43: <!-- Folding Core -->
11:40:43: <checkpoint v='3'/>
11:40:43: <core-dir v='cores'/>
11:40:43: <core-priority v='idle'/>
11:40:43: <cpu-affinity v='true'/>
11:40:43: <cpu-usage v='100'/>
11:40:43: <gpu-usage v='100'/>
11:40:43: <no-assembly v='false'/>
11:40:43:
11:40:43: <!-- Folding Slot Configuration -->
11:40:43: <cause v='ANY'/>
11:40:43: <client-subtype v='LINUX'/>
11:40:43: <client-type v='normal'/>
11:40:43: <cpu-species v='X86_PENTIUM_II'/>
11:40:43: <cpu-type v='AMD64'/>
11:40:43: <cpus v='-1'/>
11:40:43: <cuda-index v='0'/>
11:40:43: <gpu v='false'/>
11:40:43: <max-packet-size v='normal'/>
11:40:43: <opencl-index v='0'/>
11:40:43: <os-species v='UNKNOWN'/>
11:40:43: <os-type v='LINUX'/>
11:40:43: <power v='off'/>
11:40:43: <project-key v='0'/>
11:40:43: <smp v='true'/>
11:40:43:
11:40:43: <!-- HTTP Server -->
11:40:43: <allow v='127.0.0.1'/>
11:40:43: <connection-timeout v='60'/>
11:40:43: <deny v='0/0'/>
11:40:43: <http-addresses v='0:7396'/>
11:40:43: <https-addresses v=''/>
11:40:43: <max-connect-time v='900'/>
11:40:43: <max-connections v='800'/>
11:40:43: <max-request-length v='52428800'/>
11:40:43: <min-connect-time v='300'/>
11:40:43: <threads v='4'/>
11:40:43:
11:40:43: <!-- Logging -->
11:40:43: <log v='log.txt'/>
11:40:43: <log-color v='true'/>
11:40:43: <log-crlf v='false'/>
11:40:43: <log-date v='false'/>
11:40:43: <log-date-periodically v='21600'/>
11:40:43: <log-debug v='true'/>
11:40:43: <log-domain v='false'/>
11:40:43: <log-header v='true'/>
11:40:43: <log-level v='true'/>
11:40:43: <log-no-info-header v='true'/>
11:40:43: <log-redirect v='false'/>
11:40:43: <log-rotate v='true'/>
11:40:43: <log-rotate-dir v='logs'/>
11:40:43: <log-rotate-max v='16'/>
11:40:43: <log-short-level v='false'/>
11:40:43: <log-simple-domains v='true'/>
11:40:43: <log-thread-id v='false'/>
11:40:43: <log-thread-prefix v='true'/>
11:40:43: <log-time v='true'/>
11:40:43: <log-to-screen v='true'/>
11:40:43: <log-truncate v='false'/>
11:40:43: <verbosity v='5'/>
11:40:43:
11:40:43: <!-- Network -->
11:40:43: <proxy v=':8080'/>
11:40:43: <proxy-enable v='false'/>
11:40:43: <proxy-pass v=''/>
11:40:43: <proxy-user v=''/>
11:40:43:
11:40:43: <!-- Process Control -->
11:40:43: <child v='true'/>
11:40:43: <daemon v='true'/>
11:40:43: <pid v='false'/>
11:40:43: <pid-file v='/var/run/fahclient.pid'/>
11:40:43: <respawn v='false'/>
11:40:43: <service v='false'/>
11:40:43:
11:40:43: <!-- Remote Command Server -->
11:40:43: <command-address v='0.0.0.0'/>
11:40:43: <command-allow-no-pass v='127.0.0.1'/>
11:40:43: <command-deny-no-pass v='0/0'/>
11:40:43: <command-port v='36330'/>
11:40:43:
11:40:43: <!-- Slot Control -->
11:40:43: <idle v='false'/>
11:40:43: <max-shutdown-wait v='60'/>
11:40:43: <pause-on-battery v='false'/>
11:40:43: <pause-on-start v='false'/>
11:40:43:
11:40:43: <!-- User Information -->
11:40:43: <machine-id v='0'/>
11:40:43: <passkey v='********************************'/>
11:40:43: <team v='3'/>
11:40:43: <user v='mxplm'/>
11:40:43:
11:40:43: <!-- Web Server -->
11:40:43: <session-timeout v='3600'/>
11:40:43: <web-allow v='127.0.0.1'/>
11:40:43: <web-deny v='0/0'/>
11:40:43:
11:40:43: <!-- Work Unit Control -->
11:40:43: <dump-after-deadline v='true'/>
11:40:43: <max-queue v='16'/>
11:40:43: <max-units v='0'/>
11:40:43: <next-unit-percentage v='99'/>
11:40:43:
11:40:43: <!-- Folding Slots -->
11:40:43: <slot id='0' type='CPU'>
11:40:43: <cpus v='1'/>
11:40:43: </slot>
11:40:43: <slot id='1' type='GPU'>
11:40:43: <client-type v='advanced'/>
11:40:43: </slot>
11:40:43:</config>
11:40:43:Switching to user fahclient
11:40:43:Trying to access database...
11:40:43:Successfully acquired database lock
11:40:43:Enabled folding slot 00: PAUSED cpu:1 (paused)
11:40:43:Enabled folding slot 01: PAUSED gpu:0:GK106 [GeForce GTX 650 Ti] (paused)
11:40:43:Started thread 6 on PID 1123
11:40:43:Started thread 5 on PID 1123
11:40:43:Started thread 7 on PID 1123
11:40:43:Started thread 8 on PID 1123
11:40:57:Started thread 9 on PID 1123
11:41:23:WU02:FS01:Starting
11:41:23:WU02:FS01:Running FahCore: /usr/bin/FAHCoreWrapper /var/lib/fahclient/cores/www.stanford.edu/~pande/Linux/AMD64/NVIDIA/Fermi/Core_17.fah/FahCore_17 -dir 02 -suffix 01 -version 703 -lifeline 1123 -checkpoint 3 -gpu 0 -gpu-vendor nvidia
11:41:23:WU02:FS01:Started FahCore on PID 3644
11:41:23:Started thread 10 on PID 1123
11:41:23:WU02:FS01:Core PID:3648
11:41:23:WU02:FS01:FahCore 0x17 started
11:41:23:WU01:FS00:Starting
11:41:23:WU01:FS00:Running FahCore: /usr/bin/FAHCoreWrapper /var/lib/fahclient/cores/www.stanford.edu/~pande/Linux/AMD64/Core_a4.fah/FahCore_a4 -dir 01 -suffix 01 -version 703 -lifeline 1123 -checkpoint 3
11:41:23:WU01:FS00:Started FahCore on PID 3649
11:41:23:Started thread 11 on PID 1123
11:41:23:WU01:FS00:Core PID:3653
11:41:23:WU01:FS00:FahCore 0xa4 started
11:41:24:WARNING:WU02:FS01:FahCore returned: UNKNOWN_ENUM (127 = 0x7f)
11:41:24:WU01:FS00:0xa4:
11:41:24:WU01:FS00:0xa4:*------------------------------*
11:41:24:WU01:FS00:0xa4:Folding@Home Gromacs GB Core
11:41:24:WU01:FS00:0xa4:Version 2.27 (Dec. 15, 2010)
11:41:24:WU01:FS00:0xa4:
11:41:24:WU01:FS00:0xa4:Preparing to commence simulation
11:41:24:WU01:FS00:0xa4:- Looking at optimizations...
11:41:24:WU01:FS00:0xa4:- Files status OK
11:41:24:WU01:FS00:0xa4:- Expanded 135714 -> 306920 (decompressed 226.1 percent)
11:41:24:WU01:FS00:0xa4:Called DecompressByteArray: compressed_data_size=135714 data_size=306920, decompressed_data_size=306920 diff=0
11:41:24:WU01:FS00:0xa4:- Digital signature verified
11:41:24:WU01:FS00:0xa4:
11:41:24:WU01:FS00:0xa4:Project: 7048 (Run 0, Clone 716, Gen 4)
11:41:24:WU01:FS00:0xa4:
11:41:24:WU01:FS00:0xa4:Assembly optimizations on if available.
11:41:24:WU01:FS00:0xa4:Entering M.D.
11:41:24:WU02:FS01:Starting
11:41:24:WU02:FS01:Running FahCore: /usr/bin/FAHCoreWrapper /var/lib/fahclient/cores/www.stanford.edu/~pande/Linux/AMD64/NVIDIA/Fermi/Core_17.fah/FahCore_17 -dir 02 -suffix 01 -version 703 -lifeline 1123 -checkpoint 3 -gpu 0 -gpu-vendor nvidia
11:41:24:WU02:FS01:Started FahCore on PID 3657
11:41:24:Started thread 12 on PID 1123
11:41:24:WU02:FS01:Core PID:3661
11:41:24:WU02:FS01:FahCore 0x17 started
11:41:25:WARNING:WU02:FS01:FahCore returned: UNKNOWN_ENUM (127 = 0x7f)
11:41:30:WU01:FS00:0xa4:Using Gromacs checkpoints
11:41:30:WU01:FS00:0xa4:Resuming from checkpoint
11:41:30:WU01:FS00:0xa4:Verified 01/wudata_01.log
11:41:30:WU01:FS00:0xa4:Verified 01/wudata_01.trr
11:41:30:WU01:FS00:0xa4:Verified 01/wudata_01.xtc
11:41:30:WU01:FS00:0xa4:Verified 01/wudata_01.edr
11:41:30:WU01:FS00:0xa4:Completed 1221955 out of 25000000 steps (4%)
As you can see, in both cases the GTX (and only the GTX) is recorgnized as GPU. I still assume the problem is that drivers aren't loaded correctly. The nvidia-settings tool that comes with the driver can't access the cards information in the second config. So how can check if the correct driver (module???) is loaded?
Re: Using Onboard VGA for screen instead of GPU to increase
Posted: Wed Dec 18, 2013 12:16 pm
by PantherX
Welcome to the F@H Forum mxplm,
Please note that using the default verbosity level of 3 is recommended instead of 5. The reason is with level 3, only the non-default values are printed in the log which makes it easy to identify what has changed. With level 5, everything is printed which can make it difficult to figure out what is default and what isn't.
Regarding the simultaneous use of iGPU and dGPU, on Windows, that can be done by ensuring that you have installed Intel and Nvidia Drivers on the system. I am using it on my laptop and it works fine. I would assume that maybe you need Intel drivers on Linux too to ensure that both GPUs are fully functional.
Also, since FahCore_17 requires a single CPU per GPU, you may want to ensure that your CPU Slot is assigned 3 CPUs and not 4 which would over-subscribe the CPU and cause performance loss. Is there any reason to change the checkpoint time from the default of 15 minutes to 3 minutes? Do note that FahCore_17 will write a checkpoint every 2% and different FahCores may or may not use the default checkpoint value.
Re: Using Onboard VGA for screen instead of GPU to increase
Posted: Wed Dec 18, 2013 1:31 pm
by mxplm
PantherX wrote:Please note that using the default verbosity level of 3 is recommended instead of 5.
Noted
PantherX wrote:I would assume that maybe you need Intel drivers on Linux too to ensure that both GPUs are fully functional.
The Intel graphics drivers are open source and come with most Linux distros, so you generally don't have to do anything. I did a small test to confirm this. The Intel driver is present and starts working as soon as I activate the Intel chip in the BIOS. However, at this point my GTX stops working as described above.
PantherX wrote:Also, since FahCore_17 requires a single CPU per GPU, you may want to ensure that your CPU Slot is assigned 3 CPUs and not 4 which would over-subscribe the CPU and cause performance loss.
I currently have 1 core configured, since I only have 2 (real) cores.
Code: Select all
11:44:36: <!-- Folding Slots -->
11:44:36: <slot id='0' type='CPU'>
11:44:36: <cpus v='1'/>
11:44:36: </slot>
So 1 core is used up by GPU tasks and is left 1 for CPU tasks. The HT-feature, i.e. the two extra threads can be used by my regular non-crunching processes (OS, firefox, etc.) to quickly jump in and do their if they have to. In case the f@h application spends a notable amount of time on I/O-wait, then 2 extra threads could make sense. Any experiences/recommendations?
PantherX wrote:Is there any reason to change the checkpoint time from the default of 15 minutes to 3 minutes?
That is just for now, while I fiddle with the drivers and have to reboot a lot.
Re: Using Onboard VGA for screen instead of GPU to increase
Posted: Wed Dec 18, 2013 2:54 pm
by billford
mxplm wrote:
PantherX wrote:Is there any reason to change the checkpoint time from the default of 15 minutes to 3 minutes?
That is just for now, while I fiddle with the drivers and have to reboot a lot.
If I'm rebooting a lot I set
pause-on-start to
true and don't start the client again until I'm reasonably happy it'll stay running for a while.
edit- just realised that's a bit pointless if you're checking whether the client is running properly
Re: Using Onboard VGA for screen instead of GPU to increase
Posted: Wed Dec 18, 2013 2:56 pm
by PantherX
mxplm wrote:...So 1 core is used up by GPU tasks and is left 1 for CPU tasks. The HT-feature, i.e. the two extra threads can be used by my regular non-crunching processes (OS, firefox, etc.) to quickly jump in and do their if they have to. In case the f@h application spends a notable amount of time on I/O-wait, then 2 extra threads could make sense. Any experiences/recommendations?...
Since you have 2 Cores and 4 CPUs, the GPU can use the Virtual Core and you can change the CPU from 1 to 2 since it will use the 2 Cores (FPUs). Thus, you will have 3 CPUs for F@H (2 for SMP and 1 for GPU) and 1 CPU for the OS. If you are using common daily applications, 1 CPU is sufficient for them. Thus, in idle, your CPU Usage would be 75% and if you are using the system, it can be higher. To give you a rough idea, I get about 2K PPD on my non-dedicated Pentium dual cores at 2.7 GHz.
Re: Linux: Use OnboardVGA for screen instead of GPU for more
Posted: Wed Dec 18, 2013 3:29 pm
by mxplm
PantherX wrote:Since you have 2 Cores and 4 CPUs, the GPU can use the Virtual Core and you can change the CPU from 1 to 2 since it will use the 2 Cores (FPUs). Thus, you will have 3 CPUs for F@H (2 for SMP and 1 for GPU) and 1 CPU for the OS.
Thanks, I will give it a shot after I figured out a stable setup for my GPU. Any ideas on that? My basic question at this point: Does anyone know how to load NVIDIA drivers if the kernel didn't do it?
Re: Linux: Use OnboardVGA for screen instead of GPU for more
Posted: Wed Dec 18, 2013 3:56 pm
by 7im
It's not a driver install issue, because it clearly works already when just the GTX is active.
Did you look at the info posted by ChristianVirtual?
Re: Using Onboard VGA for screen instead of GPU to increase
Posted: Wed Dec 18, 2013 6:44 pm
by bruce
P5-133XL wrote:... Having two video cards on a system does significantly complicate configuration though....
One bad example:
I was successfully running a single Kepler GPU plus SMP on Ubuntu. The mb has no iGPU, so I decided to add a low-performance GPU to drive my display. It happened to be an ATI HD5xxx so I tried to configure both NV and ATI drivers. I failed, and now I have no UI display, nor does either GPU fold.
Until I get time to rebuild the whole system, it's now a SMP-only system, managed remotely via FAHControl.
I'm not surprised that I failed. I'm just really disappointed. I am absolutely not a Linux expert ... and I know from my Windows history, that getting both ATI and NV drivers to work together has been a challenge.
Re: Linux: Use OnboardVGA for screen instead of GPU for more
Posted: Thu Dec 19, 2013 12:44 pm
by mxplm
7im wrote:It's not a driver install issue, because it clearly works already when just the GTX is active.
I'm not saying the driver doesn't work, my guess is it's just not loaded by the kernel when the iGPU is active.
7im wrote:Did you look at the info posted by ChristianVirtual?
Until now I didn't, I had overlooked it because it was posted while I wrote on the following post myself. Thanks for the hint.
I tried the xorg.conf from the link (only changed GTX 660 TI to 650), but it didn't work. I still got the UNKNOWN_ENUM (127 = 0x7f) error and additionally an X11 crash report. However, the GUI seemed to work fine and I had all my (GNOME) menus.
At this point I think about just letting it go. I don't know much about the X server, so all I can do is just copying someone's config, but I don't understand it. I don't want to end up where bruce is because this is not a dedicated rig, I need the machine to work.