FAHClient V7.1.38 released (4th Open-Beta)

[brucomela]

forgive me, but, can i run Folding 7.1 GPU on my ubuntu with nvidia driver?

thanks

[Zagen30]

There's no native support for GPU slots in Linux at this time.

EDIT: Wasn't paying much attention to what I was saying, meant to say there's no Linux GPU core. Last time I checked Linux v7 can easily set up a GPU slot, but with no core/projects nothing will happen other than endless error messages about missing stuff.

[jcoffland]

Another interesting feature for v7 would be some kind of update alert feature. Some other software I have does this, it has a little popup that says "A new version is available, would you like to download it?" and then you press Yes and and it does. That simplifies things for the user I think, since even the best software gets updates every now and then, and I don't think v7 will be an exception. Even some kind of self-update thing upon launch (like Minecraft does) would be useful. That way after v7 goes public, cores/fixes/whitelist/etc updates can still be easily applied and everyone benefits even if they aren't following Dr. Pande's blog all the time.

We've talked about this idea. It would be nice. One thing though is that we don't want any popups. Popups are rather rude. I was quite annoyed when twice recently, while I was in the middle of some work a screen popped in my face offering an upgrade to the next version of Ubuntu. We would probably just place a message somewhere obvious on F@H control.

Also, I think the way to go would be to have the AS tell the client there was a new version when it requested a WS assignment. This would avoid any extra server polling.

Here's the ticket: #748

There's no native support for GPU slots in Linux at this time.

I just bought a GeForce GT430 and I work almost exclusively in Linux so I will be pushing for a Linux GPU core too. It's up to higher powers than me though.

[jcoffland]

6904 on Linux v7 / 7.1.33 As you can see the problem is not present in Linux

Thanks. That's what I thought. I suspect that non-blocking writes trigger a quirk in the Windows socket's implementation of the Nagle algorithm. (That's geek speak for, "Blame it on Microsoft.")

[Grandpa_01]

6904 on Linux v7 / 7.1.33 As you can see the problem is not present in Linux

Thanks. That's what I thought. I suspect that non-blocking writes trigger a quirk in the Windows socket's implementation of the Nagle algorithm. (That's geek speak for, "Blame it on Microsoft.")

So I am assuming it is something you can maybe fix since it is not present in Windows v6 or Linux v7. Apparently Windows v7 of F@H is triggering a problem in Windows 7 viewtopic.php?f=18&t=18609&start=30#p186886

[davidcoton]

Further to the upload speed problems:

Note: b=bit B=Byte. Read with care -- I hope I've got it right!

My ADSL line was upgraded today, the raw upload speed has more than doubled to just over 1Mb/s. However, the first two uploads (both from Linux, one a small v6 GPU and one 10MB v7 SMP) still only achieved 5KB/s -- no improvement, and nowhere near the limit for the ADSL link. No Windows uploads yet for comparison, but used to be in the same ballpark.

Tests of bandwidth to the Stanford NDT server indicate I can achieve around 500Kb/s upload on both Windows and Linux -- half the theoretical bandwidth but ten times the actual upload speed.

This is not a specific v7 problem, but it does limit my ability to process large units (because the upload will exceed the 45min timeout on the server). I don't know if there is anything that can be done to help in the v7 client -- I suspect the servers need greater ability to detect an active but slow upload and avoid the timeout.

Regards,

David

[Napoleon]

Those lines indicate that a connection to either FAHControl or FAHViewer was broken. There are many reasons this can happen and it's generally nothing to worry about. However, the last beta had connection dropping problems so if you are using either an FAHClient or FAHControl from a previous beta that could be the reason for the messages.

Had the broken connection happen again. No other strange things in the log this time, as far as I can tell. Here's the full log in case I've missed something. I haven't mixed FAHClient or FAHControl versions, and I don't I use FAHViewer. Uninstalled v7.1.33 keeping the data, then installed v7.1.38 as such.

Code: Select all

*********************** Log Started 2011-10-21T12:57:50 ************************
12:57:50:************************* Folding@home Client *************************
12:57:50:      Website: http://folding.stanford.edu/
12:57:50:    Copyright: (c) 2009-2011 Stanford University
12:57:50:       Author: Joseph Coffland <joseph@cauldrondevelopment.com>
12:57:50:         Args: --lifeline 7548 --command-port=36330
12:57:50:       Config: C:/Users/********/FAHv7/Data/config.xml
12:57:50:******************************** Build ********************************
12:57:50:      Version: 7.1.38
12:57:50:         Date: Oct 6 2011
12:57:50:         Time: 19:57:04
12:57:50:      SVN Rev: 3080
12:57:50:       Branch: fah/trunk/client
12:57:50:     Compiler: Intel(R) C++ MSVC 1500 mode 1200
12:57:50:      Options: /TP /nologo /EHa /Qdiag-disable:4297,4103,1786,279 /Ox -arch:SSE
12:57:50:               /QaxSSE2,SSE3,SSSE3,SSE4.1,SSE4.2 /Qopenmp /Qrestrict /MT
12:57:50:     Platform: win32 XP
12:57:50:         Bits: 32
12:57:50:         Mode: Release
12:57:50:******************************* System ********************************
12:57:50:          CPU: Intel(R) Atom(TM) CPU 330 @ 1.60GHz
12:57:50:       CPU ID: GenuineIntel Family 6 Model 28 Stepping 2
12:57:50:         CPUs: 4
12:57:50:       Memory: 2.75GiB
12:57:50:  Free Memory: 1.15GiB
12:57:50:      Threads: WINDOWS_THREADS
12:57:50:   On Battery: false
12:57:50:   UTC offset: 3
12:57:50:          PID: 1720
12:57:50:          CWD: C:/Users/JuhanOma/FAHv7/Data
12:57:50:           OS: Windows 7 Ultimate
12:57:50:      OS Arch: AMD64
12:57:50:         GPUs: 2
12:57:50:        GPU 0: NVIDIA:1 ION VGA
12:57:50:        GPU 1: FERMI:1 GF108 [GeForce GT 430]
12:57:50:         CUDA: 2.1
12:57:50:  CUDA Driver: 4000
12:57:50:Win32 Service: false
12:57:50:***********************************************************************
12:57:50:<config>
12:57:50:  <service-description v='Folding@home Client'/>
12:57:50:  <service-restart v='true'/>
12:57:50:  <service-restart-delay v='5000'/>
12:57:50:
12:57:50:  <!-- Client Control -->
12:57:50:  <cycle-rate v='4'/>
12:57:50:  <cycles v='-1'/>
12:57:50:  <data-directory v='.'/>
12:57:50:  <disable-project-lookup v='false'/>
12:57:50:  <exec-directory v='C:\Users\JuhanOma\FAHv7\Client'/>
12:57:50:  <exit-when-done v='false'/>
12:57:50:  <threads v='4'/>
12:57:50:
12:57:50:  <!-- Configuration -->
12:57:50:  <config-rotate v='true'/>
12:57:50:  <config-rotate-dir v='configs'/>
12:57:50:  <config-rotate-max v='16'/>
12:57:50:
12:57:50:  <!-- Debugging -->
12:57:50:  <assignment-servers>
12:57:50:    assign3.stanford.edu:8080 assign4.stanford.edu:80
12:57:50:  </assignment-servers>
12:57:50:  <capture-directory v='capture'/>
12:57:50:  <capture-sockets v='false'/>
12:57:50:  <debug-sockets v='false'/>
12:57:50:  <exception-locations v='true'/>
12:57:50:  <gpu-assignment-servers>
12:57:50:    assign-GPU.stanford.edu:80 assign-GPU.stanford.edu:8080
12:57:50:  </gpu-assignment-servers>
12:57:50:  <stack-traces v='false'/>
12:57:50:
12:57:50:  <!-- Error Handling -->
12:57:50:  <max-slot-errors v='5'/>
12:57:50:  <max-unit-errors v='5'/>
12:57:50:
12:57:50:  <!-- FahCore Control -->
12:57:50:  <checkpoint v='30'/>
12:57:50:  <core-dir v='cores'/>
12:57:50:  <core-priority v='idle'/>
12:57:50:  <cpu-affinity v='false'/>
12:57:50:  <cpu-usage v='100'/>
12:57:50:  <no-assembly v='false'/>
12:57:50:
12:57:50:  <!-- Folding Slot Configuration -->
12:57:50:  <client-subtype v='STDCLI'/>
12:57:50:  <client-type v='beta'/>
12:57:50:  <cpu-species v='X86_PENTIUM_II'/>
12:57:50:  <cpu-type v='AMD64'/>
12:57:50:  <cpus v='-1'/>
12:57:50:  <cuda-index v='0'/>
12:57:50:  <gpu v='false'/>
12:57:50:  <gpu-usage v='100'/>
12:57:50:  <max-packet-size v='big'/>
12:57:50:  <opencl-index v='0'/>
12:57:50:  <os-species v='UNKNOWN'/>
12:57:50:  <os-type v='WIN32'/>
12:57:50:  <project-key v='0'/>
12:57:50:  <smp v='true'/>
12:57:50:
12:57:50:  <!-- Logging -->
12:57:50:  <log v='log.txt'/>
12:57:50:  <log-color v='false'/>
12:57:50:  <log-crlf v='true'/>
12:57:50:  <log-date v='false'/>
12:57:50:  <log-debug v='true'/>
12:57:50:  <log-domain v='false'/>
12:57:50:  <log-header v='true'/>
12:57:50:  <log-level v='true'/>
12:57:50:  <log-no-info-header v='true'/>
12:57:50:  <log-redirect v='false'/>
12:57:50:  <log-rotate v='true'/>
12:57:50:  <log-rotate-dir v='logs'/>
12:57:50:  <log-rotate-max v='16'/>
12:57:50:  <log-short-level v='false'/>
12:57:50:  <log-simple-domains v='true'/>
12:57:50:  <log-thread-id v='false'/>
12:57:50:  <log-time v='true'/>
12:57:50:  <log-to-screen v='true'/>
12:57:50:  <log-truncate v='false'/>
12:57:50:  <verbosity v='5'/>
12:57:50:
12:57:50:  <!-- Network -->
12:57:50:  <proxy v=':8080'/>
12:57:50:  <proxy-enable v='false'/>
12:57:50:  <proxy-pass v=''/>
12:57:50:  <proxy-user v=''/>
12:57:50:
12:57:50:  <!-- Process Control -->
12:57:50:  <child v='false'/>
12:57:50:  <daemon v='false'/>
12:57:50:  <pid v='false'/>
12:57:50:  <pid-file v='Folding@home Client.pid'/>
12:57:50:  <respawn v='false'/>
12:57:50:  <service v='false'/>
12:57:50:
12:57:50:  <!-- Remote Command Server -->
12:57:50:  <command-address v='0.0.0.0'/>
12:57:50:  <command-allow v='127.0.0.1'/>
12:57:50:  <command-allow-no-pass v='127.0.0.1'/>
12:57:50:  <command-deny v='0.0.0.0/0'/>
12:57:50:  <command-deny-no-pass v='0.0.0.0/0'/>
12:57:50:  <command-port v='36330'/>
12:57:50:  <password v=''/>
12:57:50:
12:57:50:  <!-- Slot Control -->
12:57:50:  <max-shutdown-wait v='60'/>
12:57:50:  <pause-on-battery v='false'/>
12:57:50:  <pause-on-start v='false'/>
12:57:50:
12:57:50:  <!-- User Information -->
12:57:50:  <machine-id v='0'/>
12:57:50:  <passkey v='********************************'/>
12:57:50:  <team v='191980'/>
12:57:50:  <user v='Napoleon'/>
12:57:50:
12:57:50:  <!-- Work Unit Control -->
12:57:50:  <dump-after-deadline v='true'/>
12:57:50:  <max-queue v='16'/>
12:57:50:  <max-units v='0'/>
12:57:50:  <next-unit-percentage v='100'/>
12:57:50:
12:57:50:  <!-- Folding Slots -->
12:57:50:  <slot id='0' type='GPU'>
12:57:50:    <cuda-index v='0'/>
12:57:50:    <extra-core-args v='-forceasm -verbose'/>
12:57:50:    <gpu-index v='1'/>
12:57:50:    <passkey v='********************************'/>
12:57:50:    <pause-on-start v='true'/>
12:57:50:    <user v='Zotac430'/>
12:57:50:  </slot>
12:57:50:  <slot id='1' type='UNIPROCESSOR'>
12:57:50:    <extra-core-args v='-forceasm -verbose -np 2'/>
12:57:50:    <pause-on-start v='true'/>
12:57:50:  </slot>
12:57:50:  <slot id='2' type='UNIPROCESSOR'>
12:57:50:    <extra-core-args v='-forceasm -verbose -np 2'/>
12:57:50:    <pause-on-start v='true'/>
12:57:50:  </slot>
12:57:50:  <slot id='3' type='UNIPROCESSOR'>
12:57:50:    <extra-core-args v='-forceasm -verbose'/>
12:57:50:    <pause-on-start v='true'/>
12:57:50:  </slot>
12:57:50:  <slot id='4' type='UNIPROCESSOR'>
12:57:50:    <extra-core-args v='-forceasm -verbose'/>
12:57:50:    <pause-on-start v='true'/>
12:57:50:  </slot>
12:57:50:  <slot id='5' type='SMP'>
12:57:50:    <extra-core-args v='-forceasm -verbose'/>
12:57:50:    <pause-on-start v='true'/>
12:57:50:  </slot>
12:57:50:  <slot id='6' type='GPU'>
12:57:50:    <extra-core-args v='-forceasm -verbose'/>
12:57:50:    <gpu-index v='0'/>
12:57:50:    <opencl-index v='1'/>
12:57:50:    <passkey v='********************************'/>
12:57:50:    <pause-on-start v='true'/>
12:57:50:    <user v='ION'/>
12:57:50:  </slot>
12:57:50:</config>
12:57:52:Trying to access database...
12:57:52:Successfully acquired database lock
12:57:52:Enabled folding slot 00: PAUSED gpu:1:"GF108 [GeForce GT 430]"
12:57:52:Enabled folding slot 01: PAUSED uniprocessor
12:57:52:Enabled folding slot 02: PAUSED uniprocessor
12:57:52:Enabled folding slot 03: PAUSED uniprocessor
12:57:52:Enabled folding slot 04: PAUSED uniprocessor
12:57:52:Enabled folding slot 05: PAUSED smp:4
12:57:52:Enabled folding slot 06: PAUSED gpu:0:"ION VGA"
12:57:52:Started thread 1 on PID 1720
12:57:52:Started thread 5 on PID 1720
12:57:52:Started thread 4 on PID 1720
12:57:52:Started thread 3 on PID 1720
12:57:52:Started thread 6 on PID 1720
12:57:52:Server connection id=1 on 0.0.0.0:36330 from 127.0.0.1
12:57:52:Started thread 7 on PID 1720
12:58:14:Slot 06 unpaused
12:58:14:Starting Unit 01
12:58:14:Running core: C:/Users/JuhanOma/FAHv7/Data/cores/www.stanford.edu/~pande/Win32/AMD64/NVIDIA/G80/Core_11.fah/FahCore_11.exe -dir 01 -suffix 01 -lifeline 1720 -version 701 -checkpoint 30 -gpu 1 -forceasm -verbose
12:58:14:Started core on PID 772
12:58:14:FahCore 0x11 started
12:58:14:Started thread 8 on PID 1720
12:58:15:Unit 01:
12:58:15:Unit 01:*------------------------------*
12:58:15:Unit 01:Folding@Home GPU Core
12:58:15:Unit 01:Version                1.33 (Fri Jan 28 12:22:04 PST 2011)
12:58:15:Unit 01:Compiler               Microsoft (R) 32-bit C/C++ Optimizing Compiler Version 14.00.50727.42 for 80x86
12:58:15:Unit 01:Build host             amoeba
12:58:15:Unit 01:Board Type             NVIDIA/CUDA
12:58:15:Unit 01:Core                   11
12:58:15:Unit 01:GPU device id          1
12:58:15:Unit 01:
12:58:15:Unit 01:Window's signal control handler registered.
12:58:15:Unit 01:Preparing to commence simulation
12:58:15:Unit 01:- Ensuring status. Please wait.
12:58:24:Unit 01:- Assembly optimizations manually forced on.
12:58:24:Unit 01:- Not checking prior termination.
12:58:24:Unit 01:sizeof(CORE_PACKET_HDR) = 512 file=<>
12:58:24:Unit 01:- Expanded 45455 -> 251112 (decompressed 552.4 percent)
12:58:24:Unit 01:Called DecompressByteArray: compressed_data_size=45455 data_size=251112, decompressed_data_size=251112 diff=0
12:58:24:Unit 01:- Digital signature verified
12:58:24:Unit 01:
12:58:24:Unit 01:Project: 5769 (Run 8, Clone 147, Gen 6921)
12:58:24:Unit 01:
12:58:24:Unit 01:Assembly optimizations on if available.
12:58:24:Unit 01:Entering M.D.
12:58:26:Unit 01:Will resume from checkpoint file 01/wudata_01.ckp
12:58:26:Unit 01:Tpr hash 01/wudata_01.tpr:  2252805937 525691267 1733756084 4272005091 4280062583
12:58:26:Unit 01:calling fah_main gpuDeviceId=1
12:58:27:Unit 01:Working on Protein
12:58:29:Unit 01:Client config unavailable.
12:58:29:Unit 01:Setting checkpoint frequency: 150000
12:58:29:Unit 01:Completed         0 out of 15000000 steps (0%).
12:58:29:Unit 01:Starting GUI Server
12:58:29:Unit 01:Resuming from checkpoint
12:58:29:Unit 01:fcCheckPointResume: retreived and current tpr file hash:
12:58:29:Unit 01:   0   2252805937   2252805937
12:58:29:Unit 01:   1    525691267    525691267
12:58:29:Unit 01:   2   1733756084   1733756084
12:58:29:Unit 01:   3   4272005091   4272005091
12:58:29:Unit 01:   4   4280062583   4280062583
12:58:29:Unit 01:fcCheckPointResume: file hashes same.
12:58:29:Unit 01:fcCheckPointResume: state restored.
12:58:29:Unit 01:fcCheckPointResume: name 01/wudata_01.log Verified 01/wudata_01.log
12:58:29:Unit 01:fcCheckPointResume: name 01/wudata_01.edr Verified 01/wudata_01.edr
12:58:29:Unit 01:fcCheckPointResume: name 01/wudata_01.xtc Verified 01/wudata_01.xtc
12:58:29:Unit 01:fcCheckPointResume: state restored 2
12:58:29:Unit 01:Resumed from checkpoint
12:58:29:Unit 01:Completed   3150003 out of 15000000 steps (21%).
12:59:00:Slot 00 unpaused
12:59:00:Starting Unit 00
12:59:00:Running core: C:/Users/JuhanOma/FAHv7/Data/cores/www.stanford.edu/~pande/Win32/AMD64/NVIDIA/Fermi/Core_15.fah/FahCore_15.exe -dir 00 -suffix 01 -lifeline 1720 -version 701 -checkpoint 30 -gpu 0 -forceasm -verbose
12:59:00:Started core on PID 3972
12:59:00:FahCore 0x15 started
12:59:00:Started thread 9 on PID 1720
12:59:00:Unit 00:
12:59:00:Unit 00:*------------------------------*
12:59:00:Unit 00:Folding@Home GPU Core
12:59:00:Unit 00:Version                2.20 (Tue Aug 2 12:06:37 PDT 2011)
12:59:00:Unit 00:Build host             SimbiosNvdWin7
12:59:00:Unit 00:Board Type             NVIDIA/CUDA
12:59:00:Unit 00:Core                   15
12:59:00:Unit 00:
12:59:00:Unit 00:Window's signal control handler registered.
12:59:00:Unit 00:Preparing to commence simulation
12:59:00:Unit 00:- Ensuring status. Please wait.
12:59:09:Unit 00:- Assembly optimizations manually forced on.
12:59:09:Unit 00:- Not checking prior termination.
12:59:09:Unit 00:sizeof(CORE_PACKET_HDR) = 512 file=<>
12:59:09:Unit 00:- Expanded 124761 -> 501826 (decompressed 402.2 percent)
12:59:09:Unit 00:Called DecompressByteArray: compressed_data_size=124761 data_size=501826, decompressed_data_size=501826 diff=0
12:59:10:Unit 00:- Digital signature verified
12:59:10:Unit 00:
12:59:10:Unit 00:Project: 7620 (Run 549, Clone 0, Gen 46)
12:59:10:Unit 00:
12:59:10:Unit 00:Assembly optimizations on if available.
12:59:10:Unit 00:Entering M.D.
12:59:11:Unit 00:Tpr hash 00/wudata_01.tpr:  903831628 1764458592 3658980640 1703571288 3854552176
12:59:11:Unit 00:calling fah_main gpuDeviceId=0
12:59:12:Unit 00:Working on Protein
12:59:12:Unit 00:Client config unavailable.
12:59:12:Unit 00:Starting GUI Server
13:00:23:Unit 00:Setting checkpoint frequency: 400000
13:00:23:Unit 00:Completed         3 out of 40000000 steps (0%).
13:05:31:Unit 01:Completed   3300000 out of 15000000 steps (22%).
13:12:39:Unit 01:Completed   3450000 out of 15000000 steps (23%).
13:19:41:Unit 01:Completed   3600000 out of 15000000 steps (24%).
13:24:55:Unit 00:Completed    400000 out of 40000000 steps (1%).
13:26:39:Unit 01:Completed   3750000 out of 15000000 steps (25%).
13:33:36:Unit 01:Completed   3900000 out of 15000000 steps (26%).
13:40:47:Unit 01:Completed   4050000 out of 15000000 steps (27%).
13:48:25:Unit 01:Completed   4200000 out of 15000000 steps (28%).
13:49:30:Unit 00:Completed    800000 out of 40000000 steps (2%).
13:56:15:Unit 01:Completed   4350000 out of 15000000 steps (29%).
14:04:46:Unit 01:Completed   4500000 out of 15000000 steps (30%).
14:13:01:Unit 01:Completed   4650000 out of 15000000 steps (31%).
14:14:04:Unit 00:Completed   1200000 out of 40000000 steps (3%).
14:20:56:Unit 01:Completed   4800000 out of 15000000 steps (32%).
14:28:00:Unit 01:Completed   4950000 out of 15000000 steps (33%).
14:35:08:Unit 01:Completed   5100000 out of 15000000 steps (34%).
14:38:38:Unit 00:Completed   1600000 out of 40000000 steps (4%).
14:42:18:Unit 01:Completed   5250000 out of 15000000 steps (35%).
14:49:41:Unit 01:Completed   5400000 out of 15000000 steps (36%).
14:57:08:Unit 01:Completed   5550000 out of 15000000 steps (37%).
15:03:12:Unit 00:Completed   2000000 out of 40000000 steps (5%).
15:04:27:Unit 01:Completed   5700000 out of 15000000 steps (38%).
15:11:50:Unit 01:Completed   5850000 out of 15000000 steps (39%).
15:19:16:Unit 01:Completed   6000000 out of 15000000 steps (40%).
15:26:40:Unit 01:Completed   6150000 out of 15000000 steps (41%).
15:27:48:Unit 00:Completed   2400000 out of 40000000 steps (6%).
15:34:02:Unit 01:Completed   6300000 out of 15000000 steps (42%).
15:41:24:Unit 01:Completed   6450000 out of 15000000 steps (43%).
15:48:45:Unit 01:Completed   6600000 out of 15000000 steps (44%).
15:52:22:Unit 00:Completed   2800000 out of 40000000 steps (7%).
15:54:53:Slot 06 finishing
15:55:12:Slot 05 unpaused
15:55:13:Connecting to assign3.stanford.edu:8080
15:55:13:News: Welcome to Folding@Home
15:55:13:Assigned to work server 128.143.199.97
15:55:13:Requesting new work unit for slot 05: READY smp:4 from 128.143.199.97
15:55:13:Connecting to 128.143.199.97:8080
15:55:16:Slot 05: Downloading 2.02MiB
15:55:22:Slot 05: 68.79%
15:55:24:Slot 05: Download complete
15:55:24:Received Unit: id:02 state:DOWNLOAD error:OK project:7508 run:0 clone:78 gen:50 core:0xa3 unit:0x0000003bfbcb017d4e4c1572a803b8ad
15:55:24:Starting Unit 02
15:55:24:Running core: C:/Users/JuhanOma/FAHv7/Data/cores/www.stanford.edu/~pande/Win32/AMD64/beta/Core_a3.fah/FahCore_a3.exe -dir 02 -suffix 01 -lifeline 1720 -version 701 -checkpoint 30 -np 4 -forceasm -verbose
15:55:24:Started core on PID 10040
15:55:24:FahCore 0xa3 started
15:55:24:Started thread 10 on PID 1720
15:55:25:Unit 02:
15:55:25:Unit 02:*------------------------------*
15:55:25:Unit 02:Folding@Home Gromacs SMP Core
15:55:25:Unit 02:Version 2.27 (Dec. 15, 2010)
15:55:25:Unit 02:
15:55:25:Unit 02:Preparing to commence simulation
15:55:25:Unit 02:- Assembly optimizations manually forced on.
15:55:25:Unit 02:- Not checking prior termination.
15:55:25:Unit 02:- Expanded 2113428 -> 3093280 (decompressed 146.3 percent)
15:55:25:Unit 02:Called DecompressByteArray: compressed_data_size=2113428 data_size=3093280, decompressed_data_size=3093280 diff=0
15:55:25:Unit 02:- Digital signature verified
15:55:26:Unit 02:
15:55:26:Unit 02:Project: 7508 (Run 0, Clone 78, Gen 50)
15:55:26:Unit 02:
15:55:26:Unit 02:Assembly optimizations on if available.
15:55:26:Unit 02:Entering M.D.
15:55:31:Unit 02:Mapping NT from 4 to 4
15:55:33:Unit 02:Completed 0 out of 500000 steps  (0%)
15:55:53:Unit 01:Completed   6750000 out of 15000000 steps (45%).
16:03:06:Unit 01:Completed   6900000 out of 15000000 steps (46%).
16:10:27:Unit 01:Completed   7050000 out of 15000000 steps (47%).
16:16:55:Unit 00:Completed   3200000 out of 40000000 steps (8%).
16:17:54:Unit 01:Completed   7200000 out of 15000000 steps (48%).
16:25:16:Unit 01:Completed   7350000 out of 15000000 steps (49%).
16:32:35:Unit 01:Completed   7500000 out of 15000000 steps (50%).
16:39:49:Unit 01:Completed   7650000 out of 15000000 steps (51%).
16:39:54:Server connection id=2 on 0.0.0.0:36330 from 127.0.0.1
16:39:55:Started thread 11 on PID 1720
16:39:55:SocketDevice::write(): Send error: 10053: An established connection was aborted by the software in your host machine.
16:39:55:SocketDevice::write(): Send error: 10053: An established connection was aborted by the software in your host machine.
16:39:55:Server connection id=1 ended
16:41:29:Unit 00:Completed   3600000 out of 40000000 steps (9%).
16:44:41:Downloading project 7508 description
16:44:41:Connecting to fah-web.stanford.edu:80
16:44:42:SocketDevice::read(): End of stream
16:44:42:Downloading project 7620 description
16:44:42:Connecting to fah-web.stanford.edu:80
16:44:43:SocketDevice::read(): End of stream
16:46:07:Unit 02:Completed 5000 out of 500000 steps  (1%)


[Adam A. Wanderer]

I do have on hold over problem from the original beta build. Many other programs that use the GPU will intermittently "freeze" if F@H GPU slot is running. I can solve the problem by "pausing" the GPU slot until I'm finished with the other program, which is OK as long as I remember to unpause the GPU slot when I'm done. This was well reported with the original beta. The cause seems to be that the GPU card doesn't work on "Priorities" like the SMP CPU does, so all programs get equal time, or the time is divided between programs. There seems to have been no solution. I have the latest GPU driver. Probably, a differently designed card will be needed to solve this problem, or a far, far more powerful card.

[7im]

They added a GPU % usage slider in this last version, like the CPU % usage slider for CPU clients.

However, the fahcores do not support that feature very well just yet. It cycles between full power and no power instead of reducing the load by a certain percent.

They are taking steps in the right direction...

[Jesse_V]

They added a GPU % usage slider in this last version, like the CPU % usage slider for CPU clients.

However, the fahcores do not support that feature very well just yet. It cycles between full power and no power instead of reducing the load by a certain percent.

They are taking steps in the right direction...

That is very encouraging. I realize there are several tickets open about this subject, and I hope that they get it figured out when v7 becomes the main client. As SMP+GPU are now the default if on supported hardware, it would be a shame if newcomers were heavily discouraged by any lag and decided to not participate. I realize that this scenario would likely only affect a small portion of any newcomers, but it was almost non-existent with the currently recommended uniprocessor client. I'm sure there have been extensive discussions about this issue and hopefully there will be more scientific gain from the default SMP+GPU than there will be harm. Working GPU/CPU sliders will very much help.

[bruce]

The V7 client can manage specific GPUs, but that's probably something for future policy decisions.

Let's suppose that you have a GPU which causes significant lags when used by FAH. Further suppose that they decide to do that by de-whitelisting that GPU. Would that be a good idea? What about someone else who wants to dedicate that GPU to FAH but is willing to accept the lag? Either way, there's no easy solution.

Let's suppose that the FahCores can be updated to provide a smoother modulation that relatively long periods of on/off. Even with that enhancement, it's still not an ideal solution. I'm not sure we should expect a newcomer to recognize the problem, come here and ask for help, and be told to reduce the slider setting. On the other hand, setting a default %GPU to a smaller number is a bad idea because there's no general setting that applies to a majority of GPUs other than the maximum performance setting.

Trying to make FAHClient smart enough to know whether you're going to have a lag problem and to deal with it is very, very complex. Even if it's possible, adding that much code that will only be used once isn't going to happen, either.

[Adam A. Wanderer]

They added a GPU % usage slider in this last version, like the CPU % usage slider for CPU clients.

However, the fahcores do not support that feature very well just yet. It cycles between full power and no power instead of reducing the load by a certain percent.

They are taking steps in the right direction...

Thanks! It does bother me that this would reduce my contribution of GPU units. Most of the time I'm not running a program that would experience "freeze" while the GPU slot is running at full power. Perhaps a feature that would reduce the slider automatically when you call up a GPU intensive program to say about, 70% or some other optional, preset amount, then return it to 100% when you were done with the program. Then, the rest of the time the GPU slot would have the full resources of the GPU card. That way, one could maintain the output of GPU completed work units.

[7im]

So like the CPU client has two thread priority settings (Idle, Low), the GPU should have these also ...

[bruce]

Perhaps a feature that would reduce the slider automatically when you call up a GPU intensive program to say about, 70% or some other optional, preset amount, then return it to 100% when you were done with the program. Then, the rest of the time the GPU slot would have the full resources of the GPU card. That way, one could maintain the output of GPU completed work units.

You'll notice that Ticket #83 has been open a long time and although it's probably simpler to implement than your suggestion, Development has been concentrating on more important issues. Hopefully some kind of enhancement will be added to V7 after the bugs are all addressed. FAH has never attempted to monitor other programs or to dynamically re-tune your system. (...and doing that sort of thing in various operating systems would make it prohibitively expensive to do.) The CPU clients can easily be managed by using the priority setting that's built in to any modern OS. There's no priority setting for GPU workloads; GPU processing is strictly a first-in-first-out ordering and the only software that might recognize a conflict would be the drivers (and GPU drivers are a poor substitute for a real OS that could manage the work requests for stream computing resources).

[JimF]

Hopefully some kind of enhancement will be added to V7 after the bugs are all addressed. FAH has never attempted to monitor other programs or to dynamically re-tune your system. (...and doing that sort of thing in various operating systems would make it prohibitively expensive to do.) The CPU clients can easily be managed by using the priority setting that's built in to any modern OS. There's no priority setting for GPU workloads; GPU processing is strictly a first-in-first-out ordering and the only software that might recognize a conflict would be the drivers (and GPU drivers are a poor substitute for a real OS that could manage the work requests for stream computing resources).

A simple solution is to set a timer to turn on the GPU at a set time each day (e.g., 10 pm), to make sure it is on if is not already. Then, you could turn it off and not have to worry that you would forget to turn it back on later. Of course, you could turn it back on manually, but this would be a backup. That would encourage people to fold on display cards who might otherwise not do so.