a little help

[d0a0b]

ok, after reading the front page folding post a few days ago, I got inspired and thought I'd jump into the folding mix, I have some spare cpu cycles I have access to and wanted to help out.

So I followed the SMP F@H Guide by Zero82z guide and set things up, but I'm wonding if it's running slow. Should it really take about 24 hours to finish a WU on this machine?, I just want to make sure I have everything set up properly.

Machine is Windows Server Enterprise SP1 64bit
Intel Xeon E5450 @3ghz 2 proc's
32Gb Ram

Here's an excerpt from the log (user id edited out, not sure if that was necessary or not, so i did it anyway).

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

                       Folding@Home Client Version 6.34

                          http://folding.stanford.edu

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

Launch directory: C:\Folding@home SMP
Executable: C:\Folding@home SMP\Folding@home-Win32-x86.exe
Arguments: -smp 

[19:08:51] - Ask before connecting: No
[19:08:51] - User name: d0a0b (Team 33)
[19:08:51] - User ID: ***********************
[19:08:51] - Machine ID: 1
[19:08:51] 
[19:08:51] Loaded queue successfully.
[19:08:51] 
[19:08:51] + Processing work unit
[19:08:51] Core required: FahCore_a3.exe
[19:08:51] Core found.
[19:08:51] Working on queue slot 01 [May 9 19:08:51 UTC]
[19:08:51] + Working ...
[19:08:52] 
[19:08:52] *------------------------------*
[19:08:52] Folding@Home Gromacs SMP Core
[19:08:52] Version 2.27 (Dec. 15, 2010)
[19:08:52] 
[19:08:52] Preparing to commence simulation
[19:08:52] - Ensuring status. Please wait.
[19:09:01] - Looking at optimizations...
[19:09:01] - Working with standard loops on this execution.
[19:09:01] - Previous termination of core was improper.
[19:09:01] - Files status OK
[19:09:02] - Expanded 1764792 -> 1948728 (decompressed 110.4 percent)
[19:09:02] Called DecompressByteArray: compressed_data_size=1764792 data_size=1948728, decompressed_data_size=1948728 diff=0
[19:09:02] - Digital signature verified
[19:09:02] 
[19:09:02] Project: 6943 (Run 0, Clone 109, Gen 6)
[19:09:02] 
[19:09:02] Entering M.D.
[19:09:08] Using Gromacs checkpoints
[19:09:08] Mapping NT from 8 to 8 
[19:09:08] Resuming from checkpoint
[19:09:08] Verified work/wudata_01.log
[19:09:08] Verified work/wudata_01.trr
[19:09:08] Verified work/wudata_01.edr
[19:09:09] Completed 17620 out of 500000 steps  (3%)
[19:15:58] Completed 20000 out of 500000 steps  (4%)
[19:32:42] Completed 25000 out of 500000 steps  (5%)
[19:46:21] Completed 30000 out of 500000 steps  (6%)
[19:59:48] Completed 35000 out of 500000 steps  (7%)
[20:14:32] Completed 40000 out of 500000 steps  (8%)
[20:28:49] Completed 45000 out of 500000 steps  (9%)
[20:42:09] Completed 50000 out of 500000 steps  (10%)

 

[Wyxe]

"Mapping NT from 8 to 8".......You have more than 8 cores don't you? Since you just entered -smp, I'm wondering if it defaults to 8 cores. Are all your cores running 100% while folding?

 

[d0a0b]

It appears so, Here's a cap of the task manager.

 

[d0a0b]

They're quad core proc's so 8 cores total.

 

[Wyxe]

Pardon my ignorance on Xeon processors, but no HT?

 

[Haitch]

Pardon my ignorance on Xeon processors, but no HT?

Yep, the E5450 is a quad core, quad thread Xeon, no HT. The size of that page file and the physical memory in use makes me wonder what else is running on that box though.

H.

 

[jebo_4jc]

Those are non hyperthreaded CPUs, so yes that looks correct. You've done the hard work, so good job!

Your username isn't important. I would loudly share it with anybody. All they could do is contribute points under your username

One thing, you didn't put a passkey in there did you? If not, you'll want to do that to get more points. It makes you eligible for bonus points for turning in work units quickly.

First, get a passkey from here http://www.stanford.edu/group/pandegroup/folding/FAQ-passkey.html

Then, make another shortcut just like the one you have, but instead of "-smp" put "-configonly".

Go through the configuration the same way, but add the passkey when prompted.

Then, you can start up folding again.

Thanks for joining!

(Edit: I see those are harpertown CPUs. They actually might not be able to finish bigadv units in time. You're probably better off sticking with smp just the way you have it)

 

[d0a0b]

apparently not on the e5450

http://ark.intel.com/MySearch.aspx?s=t&HyperThreading=true&FamilyText=Intel%C2%AE%20Xeon%C2%AE%20Processor

 

[d0a0b]

i did add the passkey in upon setup, I first tried -bigadv flag but it didn't seem to do anything, so i took that off. Once I finish this next WU I'm going to try changing the setting from normal to big, see if that makes any difference.

 

[d0a0b]

primarily sql server is using up that ram.

 

[jebo_4jc]

i did add the passkey in upon setup, I first tried -bigadv flag but it didn't seem to do anything, so i took that off. Once I finish this next WU I'm going to try changing the setting from normal to big, see if that makes any difference.

oh yeah. forgot about that. you do need that "big" setting

 

[jfb9301]

I don't think you can get the bigadv WUs until you finish your initial 10 under your passkey.

Its a way stanford makes sure that the big WUs go to those who are best suited to finish them within deadlines

so, fold 10 SMP, with 80% of them complete without error, and make the deadlines. Then you can get your big units.

 

[d0a0b]

Does stopping a WU and restarting it effect the WU at all? if so, Is there a way to stop a work unit and start it over from the beginning? Is there a way to cancel the current WU and start on a new one?

 

[d0a0b]

for the "big" setting i wasn't referring to the -bigadv flag but rather the one that looks like it corresponds to memory in the config.

 

[mplee]

for the "big" setting i wasn't referring to the -bigadv flag but rather the one that looks like it corresponds to memory in the config.

Is that the packet size flag? If so, it makes no difference when you're running -smp. It will prevent you from running bigadv though, if it's not set to big.

 

[d0a0b]

yes, that's the one

 

[Zero82z]

I don't think you can get the bigadv WUs until you finish your initial 10 under your passkey.

Yes, you can.

Does stopping a WU and restarting it effect the WU at all? if so, Is there a way to stop a work unit and start it over from the beginning? Is there a way to cancel the current WU and start on a new one?

Yes, no, yes but you shouldn't do it unless there's a problem with the unit.

 

[d0a0b]

My first work unit I had stopped twice just by closing out of window, changed some configs and then restarted it, however when it completed it gave errors, so I guess I got no credit for it, my second WU, I did not stop and it went through fine.

If I wanted to try out different settings, that means I'd be stopping and starting a few times, and I didn't want to wait a full 24 hours to find out there was errors from my stopping and starting.

If I could start fresh after making changes, I was thinking it would reduce the possibility of errors.

 

[jebo_4jc]

Generally, you should be able to stop and start with no issues. however, at times, it does cause a work unit to fail or result in a lack of credit, as you have seen.

However, it is generally frowned upon in the community to quit the current work unit and start with a new one volitionally. This results in the original work unit hanging in limbo for a period of time until the servers realize the unit hasn't been completed. At that point, the servers re-assign the work unit to another client, and the process starts again. It has been stated numerous times that their work is very time sensitive and it suffers a major setback when work units get delayed in this way.

Generally, if a work unit is going to have an issue, it will have an issue immediatly upon being restarted. It is somewhat unusual for it to have a problem when it is supposed to be completed. Therefore, in my testing, I generally stop and start the client as needed.

 

[d0a0b]

thanks, i'll do the same.

 

[SazanEyes]

You can adjust the checkpoint frequency in the F@H configuration options. This controls how often it saves work to disk, which is the point it will restart if you stop the client. You don't want it too short or it will be constantly writing to disk, but something like 5 minutes prevents you from losing too much work.