[leelegend]

download client:

http://www.stanford.edu/~kasson/fold...P%20Client.EXE

and run the installer to a easier directory than the standard, i usually pick:

C:\fah

for the next steps you need to be logged in with admin rights to the machine and on Vista turn off UAC from the security centre.

if you have tried to install this before, open a cmd window and type:

smpd -remove
mpiexec -remove

check the task manager to make sure neither smpd or mpiexec or any stray cores are running (force them to quite if they are)

either double click on the Install.bat file that was extracted from the installer or from a cmd prompt type:

smpd -install
mpiexec -register

it will ask you for your username and password, make sure you enter the username/password of the user you are going to boot/use the machine with.

This will then come up with the message that it has encrypted the password in the registery.

check the task manager and make sure both services are running.

you can then run the client in the usual way:

fah -forceasm -verbosity 9

for the brave among you, you can get it running as a service, by setting that option in the config (1st run, or from fah -configonly).....but make sure you go into admistrative tools/services and change the "Log on" option from "system account" to the actual user/password of the user who is going to boot the machine. and then change the registry value to incorporate the flags that you require.

i have this running on 2 vista and 3 xp machines with no issues.

if you are suffering from loads of client communication errors, either the mpiexec or smpd services are not install properly (password/persmissions/etc), unistall and reinstall them as above.

i hope this helps.

lee

[LandShark]

and since putting it as service is not recommended yet.

here you can put it in the startup folder to have it auto start at start up and also hide the cmd box!

put cmdow in the same folding directory of the WinSMP along with the fah.bat file. create a shortcut of the fah.bat file and put it in the startup folder. and that's it. WinSMP client will start on start up and will also hide the cmd box from the task bar too!! this method works fine on both XP & Vista.

cmdow.exe:
http://www.commandline.co.uk/cmdow/

fah.bat file:

start cmdow.exe /RUN /HID fah.exe -verbosity 9 -forceasm -advmethods

and a bonus of using cmdow is you can hide or unhide the cmd box at will with these two bat files!! I like it better than service as I could have the cmd box pop back up to check on the folding status/log, especially when it's still in beta stage. and I could also hide the cmd box back too when done!!

to unhide:

cmdow.exe "C:\Program Files\Folding@Home Windows SMP Client V1.01\fah.exe" /VIS

to hide:

cmdow.exe "C:\Program Files\Folding@Home Windows SMP Client V1.01\fah.exe" /HID

[Joe Camel]

just set one up on lunch break

as these were FARM machines ("old" win 2k SP2), i needed to install .NET 2 (which needed windoz installer 3 to be installed first) but other than that, it was easy breezy.

any of you who have run the GPU client will recognize the SMP client.



note: if you have nothing as your password, you will need to change it to something, it wont work with no password entered



now to see if its still going when i get back home...

[Joe Camel]

so far so good....

6600 up and running

can someone convert this into PPD for me (please) TY


[18:19:22] Folding@Home Gromacs SMP Core
[18:19:22] Version 1.74 (March 10, 2007)
[18:19:22]
[18:19:22] Preparing to commence simulation
[18:19:22] - Ensuring status. Please wait.
[18:19:25] - Starting from initial work packet
[18:19:25]
[18:19:25] Project: 2610 (Run 1, Clone 404, Gen 0)
[18:19:25]
[18:19:26] Assembly optimizations on if available.
[18:19:26] Entering M.D.
[18:19:47] 8 percent)
[18:19:47] cket
[18:19:47]
[18:19:47] Project: 2610 (Run 1, Clone 404, Gen 0)
[18:19:47]
[18:19:48] 610 (Run 1, Clone 404, Gen 0)
[18:19:48]
[18:19:49] Entering M.D.
[18:19:55] Rejecting checkpoint
[18:19:56] Protein: Protein
[18:19:56] Writing local files
[18:19:56] Extra SSE boost OK.
[18:19:57] Writing local files
[18:19:57] Completed 0 out of 500000 steps (0 percent)
[18:31:12] Writing local files
[18:31:12] Completed 5000 out of 500000 steps (1 percent)
[18:42:26] Writing local files
[18:42:26] Completed 10000 out of 500000 steps (2 percent)
[18:53:39] Writing local files
[18:53:39] Completed 15000 out of 500000 steps (3 percent)
[19:04:52] Writing local files
[19:04:52] Completed 20000 out of 500000 steps (4 percent)
[19:16:05] Writing local files
[19:16:05] Completed 25000 out of 500000 steps (5 percent)
[19:27:18] Writing local files
[19:27:19] Completed 30000 out of 500000 steps (6 percent)
[19:38:31] Writing local files
[19:38:32] Completed 35000 out of 500000 steps (7 percent)
[19:49:44] Writing local files
[19:49:44] Completed 40000 out of 500000 steps (8 percent)
[20:00:57] Writing local files
[20:00:57] Completed 45000 out of 500000 steps (9 percent)
[20:12:09] Writing local files
[20:12:09] Completed 50000 out of 500000 steps (10 percent)
[20:23:22] Writing local files
[20:23:22] Completed 55000 out of 500000 steps (11 percent)

[dfonda]

Thats around 1949.44 ppd Joe. Thats @11.25 min perf rame on the 2610.
On the 2651(1760points) @11.25 you would get 2252 ppd.
The 2652 at 1148 points would get you 1469 ppd @11.25.

[Joe Camel]

Originally Posted by dfonda

Thats around 1950 ppd Joe.

OMF(olding)G thats just SICK!!!

this rig was getting about 520 PPD on patty-melts and Super-vill'n (2x clients)


sure hope its doing 3.75 times the science
(minus Beta bonus )


EDIT:
just set up the 6400 rig ... easy breezy (again)

will post screen shot and log as soon as i get 10+ frames in


EDIT:

lol ill edit the quote

[LandShark]

P2610 is a 1523 point WU.

so at 11:15min, you are getting around 1944ppd Joe.

for comparison, a e6400@just 2.9ghz is doing about 1100ppd from the P2610.

p.s. "how" you get 3400ppd out of it dfonda??

[dfonda]

Sorry about that first post I was in a rush to get the answer first.I was editing as fast as I could.

p.s. "how" you get 3400ppd out of it dfonda??

I inserted 2610(The protein) for points instead of 1523 my bad.

[Joe Camel]

6400 up and running...

(editing in screen shot and log...)

first thoughts: SMP LOVES cache


[19:38:32] Project: 2610 (Run 0, Clone 41, Gen 2)
[19:38:32]
[19:38:37] Entering M.D.
[19:38:44] iting local files
[19:38:44] riting local files
[19:38:45]
[19:38:45] xtra SSE boost OK.
[19:38:45]
[19:38:46] Completed 0 out of 500000 steps (0 percent)
[19:55:54] Writing local files
[19:55:55] Completed 5000 out of 500000 steps (1 percent)
[20:13:03] Writing local files
[20:13:04] Completed 10000 out of 500000 steps (2 percent)
[20:30:09] Writing local files
[20:30:10] Completed 15000 out of 500000 steps (3 percent)
[20:47:16] Writing local files
[20:47:17] Completed 20000 out of 500000 steps (4 percent)
[21:04:23] Writing local files
[21:04:24] Completed 25000 out of 500000 steps (5 percent)
[21:21:30] Writing local files
[21:21:30] Completed 30000 out of 500000 steps (6 percent)
[21:38:47] Writing local files
[21:38:48] Completed 35000 out of 500000 steps (7 percent)

[LandShark]

Originally Posted by Joe Camel

6400 up and running...

(editing in screen shot and log...)

first thoughts: SMP LOVES cache

yes, but not all! only a couple kind of WU are cache lover like the old P1495. WinSMP's 2610, LinuxSMP's 2609. all the rest (that we've encounter) is cache independent.

[inkfx]

Hey guys, I've got a slight problem. For the past 2 nights I've woken up to my computer powered off. I never had problems running dual instances of F@H consoles or single instance + GPU console. I've got Orthos running for 7 hours and 17 minutes now. This overclock hasn't given the slightest sign of instability. So I'm leaning towards the SMP client to be the culprit. Any ideas? I will let it run again tonight and see what happens.

[allen1090]

could anyone tell me how much memory these units are using? sis's rig running vista with only 1Gb ram isn't very pretty

[LandShark]

Originally Posted by inkfx

Hey guys, I've got a slight problem. For the past 2 nights I've woken up to my computer powered off. I never had problems running dual instances of F@H consoles or single instance + GPU console. I've got Orthos running for 7 hours and 17 minutes now. This overclock hasn't given the slightest sign of instability. So I'm leaning towards the SMP client to be the culprit. Any ideas? I will let it run again tonight and see what happens.

well, it's known that SMP (Win or Linux) folding put more stress than anything else. I could fold dual instances/orthos (big or small or mix)/pcmark/whatever on my 6600@3.33ghz all day long, but I'll have to back down to 3.25ghz to have it fold SMP w/out error or reboot!

Originally Posted by allen1090

could anyone tell me how much memory these units are using? sis's rig running vista with only 1Gb ram isn't very pretty

well, you know having just a gig of memory for vista isn't pretty to start with already! I'm not sure how "ok" she feel folding on it.... you know, some people feels like it's a bit slow, but still ok when using 512mb on XP and folding big WU like QMD/1495. but a SMP unit could use 200+MB of system memory.

[pscout]

@allen1090

2610 uses ~ 350 MB
2651 ~ 250 MB
2652 ~ 130 MB

I haven't used vista yet, but I don't think i would like to run a 2610 on a 1 GB box i was also trying to use heavily.

[pscout]

@inkfx

Orthos really doesn't cut it for ensuring folding stability as LS said.

It is good for initial testing and generating higher temps than folding usually will ... but i would not waste hours running it as a final test before folding.

A much better FINAL test is to use test wu's that you have already completed.

Just backup a wu after it is downloaded ... copy the work directory & queue.dat as a minimum ... i like to also copy the unitinfo.txt just so i can easily see what wu it is later ... and save them all in a directory eg 'SAVED WU".
You don't really need to stop the client to take the copy ... as long as it is not writing to the files in work when you do the copy.

here is a post i wrote quite a while ago on the technique http://www.ocforums.com/showpost.php...49&postcount=2

I think i wrote some more detailed ones too but couldn't find the bookmark(s)

Best current wu's are prolly 2610 for win smp, or multiple instances of 1495.

[pscout]

Originally Posted by Joe Camel

just set one up on lunch break

as these were FARM machines ("old" win 2k SP2), i needed to install .NET 2 (which needed windoz installer 3 to be installed first) but other than that, it was easy breezy.

any of you who have run the GPU client will recognize the SMP client.



note: if you have nothing as your password, you will need to change it to something, it wont work with no password entered



now to see if its still going when i get back home...

nice to see something on the SMP menu that you like

to the party

[inkfx]

Originally Posted by pscout

@inkfx

Orthos really doesn't cut it for ensuring folding stability as LS said.

It is good for initial testing and generating higher temps than folding usually will ... but i would not waste hours running it as a final test before folding.

A much better FINAL test is to use test wu's that you have already completed.

Just backup a wu after it is downloaded ... copy the work directory & queue.dat as a minimum ... i like to also copy the unitinfo.txt just so i can easily see what wu it is later ... and save them all in a directory eg 'SAVED WU".
You don't really need to stop the client to take the copy ... as long as it is not writing to the files in work when you do the copy.

here is a post i wrote quite a while ago on the technique http://www.ocforums.com/showpost.php...49&postcount=2

I think i wrote some more detailed ones too but couldn't find the bookmark(s)

Best current wu's are prolly 2610 for win smp, or multiple instances of 1495.

Thanks pscout, I will give that a try.

EDIT - I keep getting the 2651 WU. So I guess I will just stick with that and see what happens.

[pscout]

Originally Posted by inkfx

Thanks pscout, I will give that a try.

EDIT - I keep getting the 2651 WU. So I guess I will just stick with that and see what happens.

I am sure it will do. You don't really want the highest possible oc ... best to leave a small buffer to allow for buildup of dust, ambient temp changes etc.

Otherwize you have to keep re-oc'ing the rig.

One observation i have with the smp wu's is that an unstable rig due to oc tends to cause a reboot and usually does not cause an eue like folding uni clients does.
I guess the smp process/core sync stuff usaully takes the hit first instead of causing a computational error in one of the core's which is all that can get hit with a single core client.

[Spion]

i got it running as a service now thanks for the guide

[runsalone]

I installed the smp client on windows xp today and set it up. At around 27% done, I noticed it wasn't doing anything; none of the threads were consuming cpu cycles (cpu usage 0%). I left it for about an hour and came back to the same situation. I then closed the client and restarted my box, and started the client again. It then said something about it having to start over and started at 0% again. What the hell happened?

[Joe Camel]

well, the first rig (lunch setup) just finished its 1st WU

lots of down time ... 16 mins between last step (100) and first step (1) of new WU

[Mustanley]

Joe

The linux smp client also has delays between WUs like that. It's listed in the stanford smp faq somewhere. I think it's a bug they haven't worked out yet. Probably be fixed in the next client.

[Joe Camel]

no real big deal, just thought id post it here so others (who dont read the FaQ too) know its "normal"

thanks

[||Console||]

How do you figure out the ppd from this ->
[17:33:31] Folding@Home Gromacs SMP Core
[17:33:31] Version 1.74 (March 10, 2007)
[17:33:31]
[17:33:31] Preparing to commence simulation
[17:33:31] - Ensuring status. Please wait- Created dyn
[17:33:31] - Files status OK
[17:33:45] - Expanded 3615419 -> 21620008 (decompressed 597.9 percent)
[17:33:46] - Starting from initial work packet
[17:33:46]
[17:33:46] Project: 2610 (Run 1, Clone 402, Gen 0)
[17:33:46]
[17:33:46] Assembly optimizations on if available.
[17:33:46] Entering M.D.
[17:34:02] ial work packet
[17:34:02]
[17:34:02] Project: 2610 (Run 1, Clone 402, Gen 0)
[17:34:02]
[17:34:03] Entering M.D.
[17:34:10] Rejecting checkpoint
[17:34:13] Protein: Protein
[17:34:13] Writing local files
[17:34:14] Extra SSE boost OK.
[17:34:15] Writing local files
[17:34:16] Completed 0 out of 500000 steps (0 percent)
[18:00:58] Writing local files
[18:00:59] Completed 5000 out of 500000 steps (1 percent)
[18:27:39] Writing local files
[18:27:40] Completed 10000 out of 500000 steps (2 percent)
[18:54:22] Writing local files
[18:54:23] Completed 15000 out of 500000 steps (3 percent)

[nikhsub1]

OK, I chilled out and retried this SMP on my 2 vista rigs. My main rig, 64 bit forget it, it just won't run, it will (i've done everything in the guide) but it just sits there doing NOTHING much like this client, the 32 bit vista that did complete one unit, then proceeded to just do NOTHING for 11 hours. Again I will say this client is dung.

[04:18:50] Writing local files
[04:18:50] Completed 1000000 out of 1000000 steps  (100 percent)
[04:18:50] Writing final coordinates.
[04:18:51] Past main M.D. loop
[04:18:51] Will end MPI now
[04:19:51] 
[04:19:51] Finished Work Unit:
[04:19:51] - Reading up to 804816 from "work/wudata_01.arc": Read 804816
[04:19:51] - Reading up to 392976 from "work/wudata_01.xtc": Read 392976
[04:19:51] goefile size: 0
[04:19:51] logfile size: 645395
[04:19:51] Leaving Run
[04:19:55] - Writing 3095587 bytes of core data to disk...
[04:19:55]   ... Done.
[04:19:55] - Failed to delete work/wudata_01.sas
[04:19:55] - Failed to delete work/wudata_01.goe
[04:19:55] Warning:  check for stray files
[04:19:55] - Shutting down core
[04:21:55] 
[04:21:55] Folding@home Core Shutdown: FINISHED_UNIT
[04:21:55] 
[04:21:55] Folding@home Core Shutdown: FINISHED_UNIT
[04:22:00] CoreStatus = 64 (100)
[04:22:00] Unit 1 finished with 50 percent of time to deadline remaining.
[04:22:00] Updated performance fraction: 0.502218
[04:22:00] Sending work to server


[04:22:00] + Attempting to send results
[04:22:00] - Reading file work/wuresults_01.dat from core
[04:22:00]   (Read 3095587 bytes from disk)
[04:22:00] Connecting to http://171.64.65.64:8080/
[04:22:50] Posted data.
[04:22:51] Initial: 0000; - Uploaded at ~59 kB/s
[04:22:51] - Averaged speed for that direction ~59 kB/s
[04:22:51] + Results successfully sent
[04:22:51] Thank you for your contribution to Folding@Home.
[04:22:51] + Starting local stats count at 1
[04:25:12] - Warning: Could not delete all work unit files (1): Core returned invalid code
[04:25:12] Trying to send all finished work units
[04:25:12] + No unsent completed units remaining.
[04:25:12] - Preparing to get new work unit...
[04:25:12] + Attempting to get work packet
[04:25:12] - Will indicate memory of 1022 MB
[04:25:12] - Connecting to assignment server
[04:25:12] Connecting to http://assign.stanford.edu:8080/
[04:25:12] Posted data.
[04:25:12] Initial: 40AB; - Successful: assigned to (171.64.65.64).
[04:25:12] + News From Folding@Home: Welcome to Folding@Home
[04:25:12] Loaded queue successfully.
[04:25:12] Connecting to http://171.64.65.64:8080/
[04:25:17] Posted data.
[04:25:17] Initial: 0000; - Receiving payload (expected size: 3962158)
[04:25:26] - Downloaded at ~429 kB/s
[04:25:26] - Averaged speed for that direction ~380 kB/s
[04:25:26] + Received work.
[04:25:26] Trying to send all finished work units
[04:25:26] + No unsent completed units remaining.
[04:25:26] + Closed connections
[04:25:26] 
[04:25:26] + Processing work unit
[04:25:26] Core required: FahCore_a1.exe
[04:25:26] Core found.
[04:25:26] Working on Unit 02 [March 22 04:25:26]
[04:25:26] + Working ...
[04:25:26] - Calling 'mpiexec -channel auto -np 4 FahCore_a1.exe -dir work/ -suffix 02 -checkpoint 30 -forceasm -verbose -lifeline 3188 -version 591'

[04:25:26] 
[04:25:26] *------------------------------*
[04:25:26] Folding@Home Gromacs SMP Core
[04:25:26] Version 1.74 (March 10, 2007)
[04:25:26] 
[04:25:26] Preparing to commence simulation
[04:25:26] - Ensuring status. Please wait.
[04:25:43] - Assembly optimizations manually forced on.
[04:25:43] - Not checking prior termination.
[04:26:06] - Expanded 3961646 -> 21619596 (decompressed 545.7 percent)
[04:26:07] - Starting from initial work packet
[04:26:07] 
[04:26:07] Project: 2610 (Run 0, Clone 58, Gen 1)
[04:26:07] 
[04:26:09] Assembly optimizations on if available.
[04:26:09] Entering M.D.
[08:05:06] - Autosending finished units...
[08:05:06] Trying to send all finished work units
[08:05:06] + No unsent completed units remaining.
[08:05:06] - Autosend completed
[14:05:06] - Autosending finished units...
[14:05:06] Trying to send all finished work units
[14:05:06] + No unsent completed units remaining.
[14:05:06] - Autosend completed
[15:16:27] Killing all core threads
[15:16:27] Killing SMP core threads
[15:16:27] Killing 4 cores
[15:16:27] Killing core 0
[15:16:27] Killing core 1
[15:16:27] Killing core 2
[15:16:27] Killing core 3

Folding@Home Client Shutdown at user request.
[15:16:27] ***** Got a SIGTERM signal (2)
[15:16:27] Killing all core threads
[15:16:27] Killing SMP core threads
[15:16:27] Killing 4 cores
[15:16:27] Killing core 0
[15:16:27] Killing core 1
[15:16:27] Killing core 2
[15:16:27] Killing core 3

Folding@Home Client Shutdown.

[inkfx]

2651 ~14 minute frames. So far so good. I hope it stays like this. One thing though, before when my CPU shut down I was getting like 20 minute or more frame times on 2651. Now it's dropped by ~6 minutes. Perhaps something was buggy with the client or WU I received previously?

[runsalone]

OK, I chilled out and retried this SMP on my 2 vista rigs. My main rig, 64 bit forget it, it just won't run, it will (i've done everything in the guide) but it just sits there doing NOTHING much like this client, the 32 bit vista that did complete one unit, then proceeded to just do NOTHING for 11 hours. Again I will say this client is dung.

I'm going to have to agree with this assessment of the windows smp beta client. As I posted above, the client 'paused' on me the first time I tried it. I reinstalled it before I went to bed and let it run. I wake up to find it sitting there doing nothing at 3% done in a new work unit.

I will likely wait until there's a smp client that isn't flakey before trying again.

[dfonda]

How do you figure out the ppd from this

Minutes between frames x 100 divided by 60(to get hours) points of protein divided by hours. Multiple that by 24 to get ppd.

The points for a 2610 are 1523 so that's about 820ppd for you.

[Junebug]

I like Landshark's suggestion on how to autostart the client, and to "hide" it. I have been having problems with family members getting overly aggressive at closing windows. This fix worked very nicely...

Thanks Landshark!

[inkfx]

Unfortunately the SMP client caused my computer to shut down, again. I backed down my OC to 430 x 7. I hope the loss of 200mhz is the gain of stability.