SMP Client (x4) !

feriso

feriso

feriso

Soldato
Joined
30 Dec 2004
Posts
2,869
Location
Stoke-on-Trent
OK got SMp client running and looking at the activity monitor I've got four (yes 4) instances of it running - now surely that ain't right?

If not can I kill some of them and how do I stop them from turning up again on rebooting.
 
The SMP client is designed for four CPUs [or a single quad-core CPU] so naturally there will be four processes. Since you will have a dual-core CPU it is running two processes on each core.

You can't kill the processes as it would kill the client and also the WU. It is completely normal.

I have 8 processes [2 SMP clients] running and it doesn't affect my PCs performance at all so I wouldn't worry.
 
That's OK then, except my only concern is that it's takign a hell of a long time - seems like 6hours for 0.5% (!)
 
feriso said:
That's OK then, except my only concern is that it's takign a hell of a long time - seems like 6hours for 0.5% (!)
Click to expand...
that just ain't right!
how long per frame?
 
SiriusB said:
I have 8 processes [2 SMP clients] running and it doesn't affect my PCs performance at all so I wouldn't worry.
Click to expand...
Are you meeting the preferred deadlines (not the final deadline, the earlier one). If not, then you'll be doing them no favours at all by doing that, as all the work will have to be sent out twice.

My Mac Mini is borderline - just about hits the deadlines most of the time, but any interruption and it's in trouble. :eek:
 
If it's the machine in your signature then you should be getting around either 10mins or 30mins a frame depending on if it's the smaller or larger WU that you've got.

Are you getting the SSE message when it starts work?

eg.
[15:47:10] *------------------------------*
[15:47:10] Folding@Home Gromacs SMP Core
[15:47:10] Version 1.73 (November 27, 2006)
[15:47:10]
[15:47:10] Preparing to commence simulation
[15:47:10] - Ensuring status. Please wait.
[15:47:10] y forced on.
[15:47:10] - Not checking prior termination.
[15:47:11] - Expanded 2960436 -> 15158509 (decompressed 512.0 percent)
[15:47:12]
[15:47:12] Project: 2605 (Run 0, Clone 414, Gen 7)
[15:47:12]
[15:47:12] Assembly optimizations on if available.
[15:47:12] Entering M.D.
[15:47:27] on if available.
[15:47:27] Entering M.D.
[15:47:35] Calling FAH init
[15:47:35] Read topology
[15:47:36] (Starting from checkpoint)
[15:47:36] Read checkpoint
[15:47:38] Protein: Protein in POPC
[15:47:38] Writing local files
[15:47:38] Completed 130000 out of 500000 steps (26 percent)
[15:47:39] Extra SSE boost OK.
[15:57:49] Timered checkpoint triggered.
[16:07:49] Timered checkpoint triggered.
[16:16:14] Writing local files
[16:16:14] Completed 135000 out of 500000 steps (27 percent)
[16:26:15] Timered checkpoint triggered.
[16:36:15] Timered checkpoint triggered.
[16:44:32] Writing local files
[16:44:32] Completed 140000 out of 500000 steps (28 percent)
Click to expand...

that's my [email protected] running SMP in Ubuntu via VMware
edit: you may not have as much detail reported as I do since I'm using the -verbosity 9 switch but you should definately be seeing the SSE boost message - if not then you need to restart the client
 
Last edited:
Yeah I get that . I also get a 3DNow boost as well (?!) not heard of that one before.

I've left it running at home so I'll check if after work.
 
feriso said:
Yeah I get that . I also get a 3DNow boost as well (?!) not heard of that one before.

I've left it running at home so I'll check if after work.
Click to expand...
hmmm seems odd - you using any other flags when starting it?

If it's using 3DNow instead of SSE that could explian it I guess.

Also worth checking that you're not low on RAM or have anything else stealing the CPU time
 
rich99million said:
hmmm seems odd - you using any other flags when starting it?

If it's using 3DNow instead of SSE that could explian it I guess.

Also worth checking that you're not low on RAM or have anything else stealing the CPU time
Click to expand...

Nothing much else is stealing CPU time as the four cores take about :- 55/45/45/43% each

First time I looked at that my head went funny cos it came to nearly 200% - then I figured it was on each core so it would be OK :o
 
feriso said:
Nothing much else is stealing CPU time as the four cores take about :- 55/45/45/43% each

First time I looked at that my head went funny cos it came to nearly 200% - then I figured it was on each core so it would be OK :o
Click to expand...
Is it a native Ubuntu install or are you using VMware in windows?

My CPU % figures add up to under 100% though I'm using VMware so it might have something to do with that.
 
Berserker said:
Are you meeting the preferred deadlines (not the final deadline, the earlier one). If not, then you'll be doing them no favours at all by doing that, as all the work will have to be sent out twice.

My Mac Mini is borderline - just about hits the deadlines most of the time, but any interruption and it's in trouble. :eek:
Click to expand...

Normal SMP WUs I can bang out 2 in 24 hours. Larger ones [1760 pointers] I can crunch two in about 2 days.

All within the preferred deadline :)
 
Running under VMWare for the moment - hopefully I'll go native over the weekend

Seems to be OK now - I think its was the fact that I didn't have the VM in focus on the screen, Got back from work and it's doing about 40min/%
 
Berserker said:
Are you meeting the preferred deadlines (not the final deadline, the earlier one). If not, then you'll be doing them no favours at all by doing that, as all the work will have to be sent out twice.

My Mac Mini is borderline - just about hits the deadlines most of the time, but any interruption and it's in trouble. :eek:
Click to expand...

Just curious, there's no fahmon for mac so how do you find out your deadlines?
 
feriso said:
Yeah I get that . I also get a 3DNow boost as well (?!) not heard of that one before.

I've left it running at home so I'll check if after work.
Click to expand...

The Gromacs33 linux core uses both SSE and 3DNow! on AMD machines. I'm not getting a 3dNow for any SMP unit.
 
You must log in or register to reply here.
Back
Top Bottom