Not using 100% CPU usage with Deino?

[hackman2007]

Hello. Windows Task Manager used to report that I was using 100% CPU Usage on my computer. I got knocked off the Internet a few minutes ago and now it won't go back up to 100% CPU usage.

Folding starts up just fine and looks like it's running. This is a fresh install of the new version of Folding@Home.

Specs
Intel Core 2 Quad Q6600 stock clocked at 2.4GHz
2GB Corsair XMS2 DDR2 667 4-4-4-12
OCZ 700w GamerXstream PSU
Sapphire Radeon X1950GT PCI Express
Gigabyte P35-DS3R (Revision 2.0)
Windows Vista Home Premium 32 bit with SP1

Anyone know what is going on?

Edit: It looks like the Folding@Home percentage times are fine, as they are completing in time perfectly. Maybe a graphical bug? If you guys need more information, just ask



Here is my configuration file.

Code: Select all

[settings]
username=hackman2007
team=11108
passkey=
asknet=no
machineid=1
bigpackets=normal
local=1

[http]
active=no
host=localhost
port=8080
usereg=no

[core]
priority=96
cpuusage=100
disableassembly=no
checkpoint=15
ignoredeadlines=no
nocpulock=0
addr=

[power]
battery=no

[clienttype]
type=3

And here is my FAHlog.txt file.

Code: Select all

--- Opening Log file [August 1 20:48:15 UTC]


# Windows SMP Console Edition #################################################
###############################################################################

                       Folding@Home Client Version 6.22 SMP Beta2

                          http://folding.stanford.edu

###############################################################################
###############################################################################

Launch directory: C:\Program Files\Folding@Home Windows SMP Client
Executable: C:\Program Files\Folding@Home Windows SMP Client\Folding@home-Win32-x86.exe
Arguments: -deino -smp -forceasm -verbosity 9

Warning:
By using the -forceasm flag, you are overriding
safeguards in the program. If you did not intend to
do this, please restart the program without -forceasm.
If work units are not completing fully (and particularly
if your machine is overclocked), then please discontinue
use of the flag.

[20:48:15] - Ask before connecting: No
[20:48:15] - User name: hackman2007 (Team 11108)
[20:48:15] - User ID: 789AE08846C7D7B6
[20:48:15] - Machine ID: 1
[20:48:15]
[20:48:15] Loaded queue successfully.
[20:48:15]
[20:48:15] - Autosending finished units... [August 1 20:48:15 UTC]
[20:48:15] + Processing work unit
[20:48:15] Trying to send all finished work units
[20:48:15] Work type a1 not eligible for variable processors
[20:48:15] + No unsent completed units remaining.
[20:48:15] Core required: FahCore_a1.exe
[20:48:15] - Autosend completed
[20:48:15] Core found.
[20:48:15] Working on queue slot 02 [August 1 20:48:15 UTC]
[20:48:15] + Working ...
[20:48:15] - Calling 'mpiexec -np 4 -channel auto -env MPICH_USE_SMP_OPTIMIZATIONS 1 -host 127.0.0.1 FahCore_a1.exe -dir work/ -suffix 02 -priority 96 -checkpoint 15 -forceasm -verbose -lifeline 3620 -version 622'

[20:48:19]
[20:48:19] *------------------------------*
[20:48:19] Folding@Home Gromacs SMP Core
[20:48:19] Version 1.76 (February 23, 2008)
[20:48:19]
[20:48:19] Preparing to commence simulation
[20:48:19] - Ensuring status. Please wait.
[20:48:36] - Assembly optimizations manually forced on.
[20:48:36] - Not checking prior termination.
[20:48:52] - Expanded 4741105 -> 24426905 (decompressed 515.2 percent)
[20:48:53]
[20:48:53] Project: 2665 (Run 3, Clone 835, Gen 34)
[20:48:53]
[20:48:54] Assembly optimizations on if available.
[20:48:54] Entering M.D.
[20:49:06] Calling FAH init
[20:49:07] Read topology
[20:49:08] ocal files
[20:49:08] rom checkpoint)
[20:49:08] Read checkpoint
[20:49:08] Protein: HGG in water
[20:49:08] Writing local files
[20:49:09] Completed 107500 out of 250000 steps  (43 percent)
[20:49:17] Extra SSE boost OK.
[21:04:19] Timered checkpoint triggered.

[7im]

Download this client, and drop in on top of the old one. http://www.stanford.edu/~kasson/folding/FAH6.22beta2-SMP-shm.exe Works better for some people.

[hackman2007]

Download this client, and drop in on top of the old one. http://www.stanford.edu/~kasson/folding/FAH6.22beta2-SMP-shm.exe Works better for some people.

Hey 7im, thanks for the reply, but I think I'm using that client as we speak.

Make sure you see my Edit in the first post (it's in bold).

[7im]

Actually, you aren't. I can tell from the fahlog.

[20:48:15] - Calling 'mpiexec -np 4 -channel auto -env MPICH_USE_SMP_OPTIMIZATIONS 1 -host 127.0.0.1 FahCore_a1.exe -dir work/ -suffix 02 -priority 96 -checkpoint 15 -forceasm -verbose -lifeline 3620 -version 622'

The client I linked will use the -shm switch instead of auto. Works better for some people. Save your old executable just in case, but do try this alternate version at your convenience.

[hackman2007]

Actually, you aren't. I can tell from the fahlog.

[20:48:15] - Calling 'mpiexec -np 4 -channel auto -env MPICH_USE_SMP_OPTIMIZATIONS 1 -host 127.0.0.1 FahCore_a1.exe -dir work/ -suffix 02 -priority 96 -checkpoint 15 -forceasm -verbose -lifeline 3620 -version 622'

The client I linked will use the -shm switch instead of auto. Works better for some people. Save your old executable just in case, but do try this alternate version at your convenience.

Actually, what I'll do is see if Folding is going to finish in the appropriate time and keep the checkpoints and percentages okay.

If they are, I'll just leave it. I don't want to mess something up and end up having to nuke the work unit or anything like that.

I'll keep you posted.

I'm still thinking this may be a bug with either Vista or Folding. I'm personally favoring a bug towards Vista.

[hackman2007]

Okay, this is definitely a visual bug. It does not affect the performance of Folding@Home at all. I'm getting the same PPD/WU that I was before this started happening.

Here is the FAHlog.txt logfile after I posted this topic.

Code: Select all

[20:48:19] Folding@Home Gromacs SMP Core
[20:48:19] Version 1.76 (February 23, 2008)
[20:48:19]
[20:48:19] Preparing to commence simulation
[20:48:19] - Ensuring status. Please wait.
[20:48:36] - Assembly optimizations manually forced on.
[20:48:36] - Not checking prior termination.
[20:48:52] - Expanded 4741105 -> 24426905 (decompressed 515.2 percent)
[20:48:53]
[20:48:53] Project: 2665 (Run 3, Clone 835, Gen 34)
[20:48:53]
[20:48:54] Assembly optimizations on if available.
[20:48:54] Entering M.D.
[20:49:06] Calling FAH init
[20:49:07] Read topology
[20:49:08] ocal files
[20:49:08] rom checkpoint)
[20:49:08] Read checkpoint
[20:49:08] Protein: HGG in water
[20:49:08] Writing local files
[20:49:09] Completed 107500 out of 250000 steps  (43 percent)
[20:49:17] Extra SSE boost OK.
[21:04:19] Timered checkpoint triggered.
[21:09:45] Writing local files
[21:09:46] Completed 110000 out of 250000 steps  (44 percent)
[21:24:46] Timered checkpoint triggered.
[21:30:04] Writing local files
[21:30:04] Completed 112500 out of 250000 steps  (45 percent)
[21:45:04] Timered checkpoint triggered.
[21:50:18] Writing local files
[21:50:19] Completed 115000 out of 250000 steps  (46 percent)
[22:05:20] Timered checkpoint triggered.
[22:10:33] Writing local files
[22:10:34] Completed 117500 out of 250000 steps  (47 percent)
[22:25:34] Timered checkpoint triggered.
[22:30:58] Writing local files
[22:30:58] Completed 120000 out of 250000 steps  (48 percent)
[22:45:59] Timered checkpoint triggered.
[22:51:16] Writing local files
[22:51:16] Completed 122500 out of 250000 steps  (49 percent)
[23:06:17] Timered checkpoint triggered.
[23:11:32] Writing local files
[23:11:32] Completed 125000 out of 250000 steps  (50 percent)
[23:26:33] Timered checkpoint triggered.
[23:32:14] Writing local files
[23:32:14] Completed 127500 out of 250000 steps  (51 percent)

[orion456]

I have the same slow down on more than one rig since upgrading to 6.2. System is showing 15-23% usage along with System Idle showing 5-15% usage. I was accidently folding on one 5.91b6 client but in the SMP folder with the new files and it was showing the same slow down as the new 6.2 client.

Is the an SMP problem, or a client problem? How do I go back to using the older system until this gets fixed?

[hackman2007]

I have the same slow down on more than one rig since upgrading to 6.2. System is showing 15-23% usage along with System Idle showing 5-15% usage. I was accidently folding on one 5.91 client but in the SMP folder with the new files and it was showing the same slow down as the new 6.2 client.

Is the an SMP problem, or a client problem? How do I go back to using the older system until this gets fixed?

Do the frame times and the percentages look around the same figure before you switched?

[7im]

Orion, same slowdown on which version(s) of the SMP client? v5.91-MPICH, v5.92-Deino, v6.22-MPICH, v6.22-Deino Auto from download page, v6.22-Deino -SHM from Bebergs post?

Berberg posted links to the v5.91 and v5.92 clients with extended deadlines. If you want to go backwards, then please go find his post and go back to the old version.

However, no one is yet to point out a clear path to what caused the slowdown. hackman2007 won't try the fix I suggested, and orion is reloading the old client. We won't find this problem unless people are willing to test or try things out. Sorry.

On another note, the FAHCORES do all the folding, and the fahcores didn't change. If you are seeing a slowdown, it's likely because you installed a different version of the client than you were running before, i.e. switched from MPICH to DEINO, or are running the DEINO client without the -smp AND -deino switches. Or you were running the Deino -shm in v5.92 and then tried to run the v6.22 Deino -auto client from the download page.

Either way, more feedback is needed before Pande Group will be able to figure this out. Any volunteers?

[AndreworNIL]

I volunteer! But i have to ask you to carefully instruct me what to do. I tried shm executable, but it has another issue. GPU executable using one core, three SMP executables using a core each and one SMP executable does not using a core (no cores left i think ). I'm going to see how this impact the PPD and report of results.

[7im]

I'm not sure how to instruct you to break your SMP install to slow it down as they posted above. And even then, I'm not sure of the fix either.

What I had seen cause the problem before was...
installed a different version of the client than you were running before, i.e. switched from MPICH to DEINO, or are running the DEINO client without the -smp AND -deino switches.
Or you were running the Deino -shm in v5.92 and then tried to run the v6.22 Deino -auto client from the download page.

The fixes would be to use the same type of client as your previous install. If it was v5.91, then upgrade to v6.22-MPICH. If it was v5.92, then download the custom -SHM client I linked to above.

[P5-133XL]

OK, I am willing to experiment a bit.

DL 6.20 deino SMP using -smp parameter and noticed that the task manager was consistantly leaving 50% idle with 4 core processes running.
Added the -deino parameter and that didn't change anything
DL 6.22beta2 and I now get the following error CoreStatus 63 (99) error - Error starting core process.

Windows XP Media Center (SP3) Q6600@3.2GHz; 4GB Ram;

Code: Select all

[08:22:29] + Processing work unit
[08:22:29] Work type a1 not eligible for variable processors
[08:22:29] Core required: FahCore_a1.exe
[08:22:29] Core found.
[08:22:29] Working on queue slot 02 [August 2 08:22:29 UTC]
[08:22:29] + Working ...
[08:22:29] - Calling 'mpiexec -np 4 -channel shm -env MPICH_USE_SMP_OPTIMIZATIONS 1 -host 127.0.0.1 FahCore_a1.exe -dir work/ -suffix 02 -checkpoint 15 -verbose -lifeline 3112 -version 622'

[08:22:29]
[08:22:29] *------------------------------*
[08:22:29] Folding@Home Gromacs SMP Core
[08:22:29] Version 1.74 (March 10, 2007)
[08:22:29]
[08:22:29] Preparing to commence simulation
[08:22:29] - Ensuring status. Please wait.
[08:22:46] - Looking at optimizations...
[08:22:46] - Working with standard loops on this execution.
[08:22:46] Examination of work files indicates 8 consecutive improper terminations of core.
[08:22:57] - Expanded 4676442 -> 24111057 (decompressed 515.5 percent)
[08:22:57]
[08:22:57] Project: 2665 (Run 1, Clone 901, Gen 39)
[08:22:57]
[08:23:01] Entering M.D.
[08:23:11] CoreStatus = 63 (99)
[08:23:11] + Error starting Folding@Home core.
[08:23:16]
[08:23:16] + Processing work unit
[08:23:16] Work type a1 not eligible for variable processors
[08:23:16] Core required: FahCore_a1.exe
[08:23:16] Core found.
[08:23:16] Working on queue slot 02 [August 2 08:23:16 UTC]
[08:23:16] + Working ...
[08:23:16] - Calling 'mpiexec -np 4 -channel shm -env MPICH_USE_SMP_OPTIMIZATIONS 1 -host 127.0.0.1 FahCore_a1.exe -dir work/ -suffix 02 -checkpoint 15 -verbose -lifeline 3112 -version 622'

[08:23:16]
[08:23:16] *------------------------------*
[08:23:16] Folding@Home Gromacs SMP Core
[08:23:16] Version 1.74 (March 10, 2007)
[08:23:16]
[08:23:16] Preparing to commence simulation
[08:23:16] - Ensuring status. Please wait.
[08:23:33] - Looking at optimizations...
[08:23:33] - Working with standard loops on this execution.
[08:23:33] Examination of work files indicates 8 consecutive improper terminations of core.
[08:23:44] - Expanded 4676442 -> 24111057 (decompressed 515.5 percent)
[08:23:44]
[08:23:44] Project: 2665 (Run 1, Clone 901, Gen 39)
[08:23:44]
[08:23:48] Entering M.D.
[08:23:58] CoreStatus = 63 (99)
[08:23:58] + Error starting Folding@Home core.
[08:23:58] - Attempting to download new core...
[08:23:58] + Downloading new core: FahCore_a1.exe
[08:23:58] Downloading core (/~pande/Win32/x86/beta/Core_a1.fah from www.stanford.edu)
[08:23:58] Initial: AFDE; + 10240 bytes downloaded
[08:23:58] Initial: AD21; + 20480 bytes downloaded

...

[08:23:59] Initial: 34A0; + 778240 bytes downloaded
[08:23:59] Initial: DD6C; + 788480 bytes downloaded
[08:23:59] Initial: D2E9; + 789667 bytes downloaded
[08:23:59] Verifying core Core_a1.fah...
[08:23:59] Signature is VALID
[08:23:59]
[08:23:59] Trying to unzip core FahCore_a1.exe
[08:23:59] Decompressed FahCore_a1.exe (2035712 bytes) successfully
[08:24:04] + Core successfully engaged
[08:24:09]
[08:24:09] + Processing work unit
[08:24:09] Work type a1 not eligible for variable processors
[08:24:09] Core required: FahCore_a1.exe
[08:24:09] Core found.
[08:24:09] Working on queue slot 02 [August 2 08:24:09 UTC]
[08:24:09] + Working ...
[08:24:09] - Calling 'mpiexec -np 4 -channel shm -env MPICH_USE_SMP_OPTIMIZATIONS 1 -host 127.0.0.1 FahCore_a1.exe -dir work/ -suffix 02 -checkpoint 15 -verbose -lifeline 3112 -version 622'

[08:24:10]
[08:24:10] *------------------------------*
[08:24:10] Folding@Home Gromacs SMP Core
[08:24:10] Version 1.74 (March 10, 2007)
[08:24:10]
[08:24:10] Preparing to commence simulation
[08:24:10] - Ensuring status. Please wait.
[08:24:27] - Looking at optimizations...
[08:24:27] - Working with standard loops on this execution.
[08:24:27] Examination of work files indicates 8 consecutive improper terminations of core.
[08:24:37] - Expanded 4676442 -> 24111057 (decompressed 515.5 percent)
[08:24:38]
[08:24:38] Project: 2665 (Run 1, Clone 901, Gen 39)
[08:24:38]
[08:24:41] Entering M.D.
[08:24:52] CoreStatus = 63 (99)
[08:24:52] + Error starting Folding@Home core.
[08:24:57]
[08:24:57] + Processing work unit
[08:24:57] Work type a1 not eligible for variable processors
[08:24:57] Core required: FahCore_a1.exe
[08:24:57] Core found.
[08:24:57] Working on queue slot 02 [August 2 08:24:57 UTC]
[08:24:57] + Working ...
[08:24:57] - Calling 'mpiexec -np 4 -channel shm -env MPICH_USE_SMP_OPTIMIZATIONS 1 -host 127.0.0.1 FahCore_a1.exe -dir work/ -suffix 02 -checkpoint 15 -verbose -lifeline 3112 -version 622'

[08:24:57]
[08:24:57] *------------------------------*
[08:24:57] Folding@Home Gromacs SMP Core
[08:24:57] Version 1.74 (March 10, 2007)
[08:24:57]
[08:24:57] Preparing to commence simulation
[08:24:57] - Ensuring status. Please wait.
[08:25:14] - Looking at optimizations...
[08:25:14] - Working with standard loops on this execution.
[08:25:14] Examination of work files indicates 8 consecutive improper terminations of core.
[08:25:25] - Expanded 4676442 -> 24111057 (decompressed 515.5 percent)
[08:25:25]
[08:25:25] Project: 2665 (Run 1, Clone 901, Gen 39)
[08:25:25]
[08:25:29] Entering M.D.
[08:25:41] CoreStatus = 63 (99)
[08:25:41] + Error starting Folding@Home core.
[08:25:46]
[08:25:46] + Processing work unit
[08:25:46] Work type a1 not eligible for variable processors
[08:25:46] Core required: FahCore_a1.exe
[08:25:46] Core found.
[08:25:46] Working on queue slot 02 [August 2 08:25:46 UTC]
[08:25:46] + Working ...
[08:25:46] - Calling 'mpiexec -np 4 -channel shm -env MPICH_USE_SMP_OPTIMIZATIONS 1 -host 127.0.0.1 FahCore_a1.exe -dir work/ -suffix 02 -checkpoint 15 -verbose -lifeline 3112 -version 622'

[08:25:46]
[08:25:46] *------------------------------*
[08:25:46] Folding@Home Gromacs SMP Core
[08:25:46] Version 1.74 (March 10, 2007)
[08:25:46]
[08:25:46] Preparing to commence simulation
[08:25:46] - Ensuring status. Please wait.
[08:26:03] - Looking at optimizations...
[08:26:03] - Working with standard loops on this execution.
[08:26:03] Examination of work files indicates 8 consecutive improper terminations of core.
[08:26:14] - Expanded 4676442 -> 24111057 (decompressed 515.5 percent)
[08:26:14]
[08:26:14] Project: 2665 (Run 1, Clone 901, Gen 39)
[08:26:14]
[08:26:17] Entering M.D.
[08:26:28] CoreStatus = 63 (99)
[08:26:28] + Error starting Folding@Home core.
[08:26:28] - Attempting to download new core...
[08:26:28] + Downloading new core: FahCore_a1.exe
[08:26:28] Downloading core (/~pande/Win32/x86/beta/Core_a1.fah from www.stanford.edu)
[08:26:31] Initial: AFDE; + 10240 bytes downloaded
[08:26:31] Initial: AD21; + 20480 bytes downloaded
[08:26:31] Initial: CC38; + 30720 bytes downloaded
[08:26:31] Initial: 8501; + 40960 bytes downloaded

...

[08:26:32] Initial: DD6C; + 788480 bytes downloaded
[08:26:32] Initial: D2E9; + 789667 bytes downloaded
[08:26:32] Verifying core Core_a1.fah...
[08:26:32] Signature is VALID
[08:26:32]
[08:26:32] Trying to unzip core FahCore_a1.exe
[08:26:32] Decompressed FahCore_a1.exe (2035712 bytes) successfully
[08:26:37] + Core successfully engaged
[08:26:48]
[08:26:48] + Processing work unit
[08:26:48] Work type a1 not eligible for variable processors
[08:26:48] Core required: FahCore_a1.exe
[08:26:48] Core found.
[08:26:48] Working on queue slot 02 [August 2 08:26:48 UTC]
[08:26:48] + Working ...
[08:26:48] - Calling 'mpiexec -np 4 -channel shm -env MPICH_USE_SMP_OPTIMIZATIONS 1 -host 127.0.0.1 FahCore_a1.exe -dir work/ -suffix 02 -checkpoint 15 -verbose -lifeline 3112 -version 622'

[08:26:49]
[08:26:49] *------------------------------*
[08:26:49] Folding@Home Gromacs SMP Core
[08:26:49] Version 1.74 (March 10, 2007)
[08:26:49]
[08:26:49] Preparing to commence simulation
[08:26:49] - Ensuring status. Please wait.
[08:27:06] - Looking at optimizations...
[08:27:06] - Working with standard loops on this execution.
[08:27:06] Examination of work files indicates 8 consecutive improper terminations of core.
[08:27:14] Killing all core threads
[08:27:14] Killing 2 cores
[08:27:14] Killing core 0
[08:27:14] Killing core 1

Folding@Home Client Shutdown at user request.
[08:27:14] ***** Got a SIGTERM signal (2)
[08:27:14] Killing all core threads
[08:27:14] Killing 2 cores
[08:27:14] Killing core 0
[08:27:14] Killing core 1

Folding@Home Client Shutdown.


[orion456]

Orion, same slowdown on which version(s) of the SMP client? v5.91-MPICH, v5.92-Deino, v6.22-MPICH, v6.22-Deino Auto from download page, v6.22-Deino -SHM from Bebergs post?

I upgraded from v5.91-MPICH to v6.22Deino (WinXP Pro). However because I didn't notice that the client name had changed I continued to use my old shortcut and was folding with v.5.91 but now on Deino; it was using only about 80% core, with about 20% split between the System and System Idle. Changing to v6.22 made no difference, the slow down continuted. I dropped in the v.6.22 -SHM version - same slow down. I guess that means its in Deino. I also notice that DeinoPM.exe has 5 instances in Windows Task Manager, not sure if that is normal?

I want to run the Deino version because I'm having problems with my net link shutting down. Do you think a complete delete and reinstall might help?

Confirmed that deleting Deino and installing v6.22-MPICH there is no slow down.

EDIT: Aug 2 - I deleted all the files in my FahSMP folder but left the working folder. I uninstalled "smpd.exe -uninstall" and removed smpd.exe from the folder as well. I then installed v6.22-Deino with flags "-smp -deino -forceasm -verbosity 9" and upon running it showed the same slow down. Dropping in FAH6.22beta2-smp-shm.exe I now have full speed working. I'm not sure why that didn't work the last time; perhaps removing smpd.exe helped.

[hackman2007]

Hey 7im, I'm sorry if I sounded like I wasn't willing to test out what was going on with this client. I didn't mean to sound like that at all. I didn't want to screw up the current work unit it was working on and end up not returning it to Stanford in time.

Now that I have a work unit at 0% I'm willing to try messing since if it starts over I will definitely finish it in time.

Anyways, I downloaded the client you linked in one of your earlier posts. I deleted the other file that I downloaded. I made a shortcut to the new one inside of the Folding folder and added the following flags: -smp -forceasm -verbosity 9 -deino

Windows Task Manager says my computer is now running at 100% CPU Usage. I will post the appropriate logfiles and such when I get some percentages.

[7im]

And I'm sorry if I sounded like I was picking on you. I wasn't. I'd actually prefer that you not lose a work unit while testing. And thank you for sticking with it. Glad to have you and orion both jumping in to help everyone else who runs across this problem.

My initial guess is that SMPD might be stealing cycles, but I'm sure we'll figure it out soon enough.