Anyone folding using new windows multi processor client?

[SemperFi]

Just curious. I installed the other day and just turned in my first unit an hour or so ago. I don't know how many points it was worth yet. Thing that I find weird is when I was running the Linux client through vm all 4 cores would peg into the high 90's as far as percent usage. Since I switched to the windows client now 3 cores run in the range of 70-90 percent usage and the fourth one looks like about 25-30 percent usage. In task manager the total usage of all cores runs around 55% and will spike up to 70-80s percent momentarily.

I was wondering if anyone else has experienced this or should I be looking for problems in my setup?

 

[GLeeM]

I haven't used the new client yet - Assimilator1 has, maybe he can help?
IF this is normal, it would sure help when running one or more GPU clients!
From past experience with other clients I would have thought the new client would use closer to 100% CPU usage?

Just in case you don't know yet: you should start getting the bonus after you finish ten WUs. And you must use a "passkey" - IIRC you do.

 

[SemperFi]

Thanks Gleem, I do have a passkey so that is taken care of. I got busy last evening and forgot to check how many points I got but from my total I figure somewhere between 4000 and 4500. Not bad for the cores not running near 100%.

I just saw that my processor wasn't working as hard as when I was running the Linux version. Actually now that I think about it running the linux version I would finish a unit in about a day and this first unit I did took nearly 2. I need to keep an eye on this for a while.

Edit/ just checked the log and it is just 5% along and has had this wu for 17 hours so I am definately not putting out nearly as much so the lack of cpu usage has to have some effect on this.

 

[Assimilator1]

OK I take it we're talking about F@H .

Well I've just been watching the Task man. for about a minute & FahCore_a3 has been pegging the CPU (Q6600) soldily at 99% (& ~172MB) throughout that time.

IIRC I've got 'large WUs' & 'advmethod' on (I think I've got that last name right??).
Have you?

Last completed WU took just over 8 1/2hrs (Project 6015).
That's on a Q6600 @3.34 GHz, FSB 417 MHz, RAM @1042 MHz. Intel P45 chipset.
Partial log for that WU:-

[02:46:22] Completed 0 out of 500000 steps (0%)
[02:51:34] Completed 5000 out of 500000 steps (1%)
[02:56:45] Completed 10000 out of 500000 steps (2%)
[03:01:57] Completed 15000 out of 500000 steps (3%)
[03:07:09] Completed 20000 out of 500000 steps (4%)
[03:12:20] Completed 25000 out of 500000 steps (5%)
[03:17:32] Completed 30000 out of 500000 steps (6%)
[03:22:43] Completed 35000 out of 500000 steps (7%)
[03:27:55] Completed 40000 out of 500000 steps (8%)
[03:33:07] Completed 45000 out of 500000 steps (9%)
[03:38:18] Completed 50000 out of 500000 steps (10%)
[03:43:29] Completed 55000 out of 500000 steps (11%)
[03:48:40] Completed 60000 out of 500000 steps (12%)
[03:53:52] Completed 65000 out of 500000 steps (13%)
[03:59:04] Completed 70000 out of 500000 steps (14%)
[04:04:15] Completed 75000 out of 500000 steps (15%)
[04:09:27] Completed 80000 out of 500000 steps (16%)

[11:14:32] Completed 490000 out of 500000 steps (98%)
[11:19:42] Completed 495000 out of 500000 steps (99%)
[11:24:53] Completed 500000 out of 500000 steps (100%)

WU before that (Project 6023) also took ~8.5 hrs.
A 6024 WU again took ~8.5 hrs, 2653 WU ~14hrs (!)

What spec's your rig?

 

[SemperFi]

I am running core 78. I don't have large working units, just the normal. I don't remember on the advanced methods I believe it is on. I think you actually have to turn that one off, which I didn't do.

Also when I first installed it tried 7 times to download core 78 which I already had on another computer. So I transfered it off of my server as I was late for work. I figured the core was the same and it started folding after I transfered it.

Edit/ just checked and it is at 40% in the past 5 hours so there had to be one sent up early this morning.
I have an Athlon II with the other cores unlocked and overclocked slightly to 2.87Ghz and 4 gig of ram.
Window has SSE boost Ok which I believe is the advmethods flag so that is on.
I just went to the folder to look at the log file and there isn't one, nor is there a work folder like in the other client. I just wanted to check out the log file and see how many wu's have been sent up the dos window only goes back so far.

 

[Assimilator1]

Err, I'm still rather new to F@H, what's core 78? :$

Btw you mentioned 'other client', is that a GPU client or on another rig?

 

[SemperFi]

no it is the normal windows client on another rig. Well it is Fahcore_78 I just pulled it out of the air as I remembered it.

 

[Assimilator1]

See here for more info on the a3 core http://forums.anandtech.com/showthread.php?t=2045182
Their's a different flag for getting that 1.

 

[Slider76]

SemperFi, I was running a Notfreds Linux VM on my Quad core @3.45ghz, had switched over to the new windows smp client. Make sure that your windows client is estup to use all for cores of you processor, as you can change it when doing the initial config.

I found an improvement going from linux notfred (~3,500ppd) to windows smp3 (~4,500ppd).
The first 10 WU's you don't get any bonus points on, which really hurts production.

 

[SemperFi]

Ok it looks like I really screwed this thing up.

I didn't have smp enabled in the correct place therefor it wasn't enabled at all. I just went through the -config again and enabled the smp in the correct place this time. When I click the client the window pops up says read the license agreement and that it must be accepted. Then a bunch of errors then window closes. This is what is displayed.

4 cores detected
Unable to connect to 'Jeffs-Toy:8676',
sock error: generic socket failure, error stack:
MPIDU_Sock_post_connect(1228): unable to connect to Jeffs-Toy on port 8676, exhausted all endpoints (errno -1)
MPIDU_Sock_post_connect(1275): unable to connect to Jeffs-Toy on port 8676, No connection could be made because the target machine actively refused it. (errno 10061)
[02:32:09] Error running mpiexec. Please check your configuration and re-run install.bat if necessary

I reran install.bat as administrator and get the following readout.

C:\Windows\system32>"C:\Program Files (x86)\Folding@Home Windows SMP Client V1.0
1\install.bat"
'smpd' is not recognized as an internal or external command,
operable program or batch file.
'mpiexec' is not recognized as an internal or external command,
operable program or batch file.
'mpiexec' is not recognized as an internal or external command,
operable program or batch file.
Press any key to continue . . .

So basically I have no idea what the hell the problem is now.

Those smilies in the above path are supposed to be a colon and back slash.

 

[SemperFi]

Ok I uninstalled the folding and downloaded again the smp client. This time I put it in the root directory instead of the program files directory. Reran the install.bat and configured the new client with big working units and advmethods flag. Started the client it downloaded an a1 core and is currently folding so I guess that is where I need to be. I will keep an eye out for an a3 core.

Edit/ also put the smp switch in the shortcut properties correctly this time and task manager is showing 100% usage.

Edit 2/ Only problem I see now is how slow it is coming along. Almost 10 hours into this unit and only at 6%. My folding monitor says it is due in 1 day 18 hours and says it will finish in 6 days. I had installed the smp on my dual core machine but quit using it for the same reason. I'll ride it out to see what happens. Maybe the a3 core will make a difference.

 

[GLeeM]

My folding monitor says it is due in 1 day 18 hours and says it will finish in 6 days

What is the number of this project?

 

[SemperFi]

P3065 (R3, C460, G19)

The monitor now says it will finish in 20 hours and is due in 21. So that is better.

Edit/ now it is 3 days ETA on finishing. I don't know if I read it wrong or if it is jumping around like that. I will see I guess.

 

[Assimilator1]

To get the a3 core you need to put in a different flag to the default one, see link above.

 

[SemperFi]

Yea I looked at it when you posted it. In the setup there was a question in the advanced setup asking if I want to enable the advmethod flag. Is that what you are talking about? Or do I need to put that flag in the properties of the exe. file. I used the former in the advanced setup. Which one are you doing Assimilator? The flag in the setup or in your command line when you start your client?

I am at 4 days till this WU is done says HFM monitor. Says 11 hours till it is due. I was going to wait it out but I am missing out on a bunch of points. I got close to 3000 every day to day and a half on the VM and alfreds. I have been running this a1 unit since thurs or friday and am still at 34% I normally don't give a rip about points but I just made it into the 100 on our team with someone on my heels. I need about 20,000 more to insure I stay there for the immediate future. I don't know. Will the client flush this WU when time expires? Maybe it will get an a3 Wu next time. Open to suggestions but getting very aggravated.

 

[xbassman]

Hey SemperFi,

It sounds like you definitely have something messed up.
I would check that you aren't limiting the % your core get.

Here is a guide that will get you up and running right. (assuming you don't have other issues.)
Since you have already installed you can stop the client and start the guide under client configuration.

http://folding.stanford.edu/English/WinSMPGuideMPICH

I have installed using this guide on 2 dual cores and both run each core @ 100%.
On my main box (Core2 2.5ghz) I am finishing each A3 WU in around 23 hrs or so.

I just set -advmethods in the config. Then it is not necessary to add it to your shortcut.

 

[SemperFi]

Thanks xbassman I will go through the guide here in a bit I have a meeting coming up in a few. I have all 4 cores cranking at 100% now I figured that one out a few days ago. I didn't have -smp in the correct place in the shortcut properties.

 

[SemperFi]

I went through that guide before I installed the client. I did forget to put the -verbosity 9 in the shortcut so I did that.

The wu is now over due and the message, Timered checkpoint triggered, now appears after each percent that finishes. I was hoping it would flush the wu and get a new one. Hopefully an a3 core. But it isn't going to apparently. I tried shutting it down and restarting hoping that would do it but no go.

I couldn't stand it anymore. I just came back to my computer and only processed a percent or 3 since I got home from work. then stopped working and just gave the error listed above. I shut it down and deleted the work folder and just got the a3 core hopefully I will start to finally put up some points again. The past week has been killing me in points.

 

[Assimilator1]

I only have -smp in the shortcut, rest was done in setup.

And doh!, I missed that you had already posted in that thread before me lol! :$

 

[xbassman]

I only have -smp in the shortcut, rest was done in setup.

And doh!, I missed that you had already posted in that thread before me lol! :$

With mine, I did it all in the setup.
My shortcut is: "C:\FAH\fah6.exe"

I added the switches under "additional client parameters".

My log starts:
Launch directory: C:\FAH
Executable: C:\FAH\fah6.exe
Arguments: -smp -verbosity 9

Is the A3 doing any better?
Anyway I have a cheap core2quad coming tomorrow.
I am gonna install the new client and am now wondering if I will have issues as well.

 

[SemperFi]

I just got to my computer this morning. Looks like I got an update from microshaft and my computer rebooted so of course nothing was running. I just fired the client back up and am at 15% which took me 2 days to get to with the a1 core so I would say that it doing much better. Obviously I will have a better idea by this evening. I might try putting this back on my dual core processor if it works out well. I never did get an a3 for it either.

 

[Assimilator1]

Hmm, odd, mine automatically updated with the switch in place.....

 

[SemperFi]

Hmm, odd, mine automatically updated with the switch in place.....

Not sure what you mean by that.

I just got home and I am at 92%. Most definitely a MAJOR improvement. I can get used to this.

 

[xbassman]

Got my quad core today.

It is sitting @ 99% cpu usage.

Here's the limited benchmark from HFM.net:

Name: Quad
Path: \\Quad\fah\
Number of Frames Observed: 4

Min. Time / Frame : 00:06:37 - 5,515.9 PPD
Avg. Time / Frame : 00:06:44 - 5,373.2 PPD
Cur. Time / Frame : 00:06:56 - 5,186.5 PPD
R3F. Time / Frame : 00:06:45 - 5,327.3 PPD
All Time / Frame : 00:06:44 - 5,340.5 PPD
Eff. Time / Frame : 00:08:23 - 4,289.4 PPD

 

[SemperFi]

xbassman I didn't notice that benchmark on there I will have to do some checking with that. I just fired up the smp client on my old opteron dual core again and it doesn't look good. I thought I saw someone else using the smp client on dual core. I am going to give it a day before I pull the plug.

My WHS has a more recent athlon dual core, I may give it a try on that over the weekend.