2 SMP and 2 GPU clients getting 7k ppd...

MaddBomber83

MaddBomber83

Weaksauce
Joined
Jun 20, 2006
Messages
109
I have two computers that I'm running a SMP and GPU client on each and am getting 7k ppd total. I was told in another thread that this was low and to make a new thread to figure it out. Any help would be appreciated!

My main computer is running an i7 920 @ 3.6 ghz.
It gets 249 ppd with the SMP client.

I have the SMP client set up as a service with a -smp flag.

My card is an ATI 4870 and gets 2000 ppd.


My Wife's rig has a Cor Duo @ 3.0 ghz and gets 150 ppd
Her 8800 GTX is pushing about 4k ppd.

Not sure what other info is needed to help find any problems.

Thank you in advance!

 
Thank you for the links. I have done what is shown there and will have to wait for new cores and such to be loaded.

I think the GPU clients are doing well. Both cards show 100% utilization. The changes you had me made increased the ATI card to ~ 2500 ppd.

Now the SMP clients... I'm waiting for them to finish what they have for the new cores. I notice that my CPU is not running at full speed. I have a display that shows each core and they are all ziging around 20-50% with the average being about 40%. Is there a reason they are not using more cycles?

Should CPU affinity be yes or no?

Thank you in advance!
 
MaddBomber83 said:
Thank you for the links. I have done what is shown there and will have to wait for new cores and such to be loaded.

I think the GPU clients are doing well. Both cards show 100% utilization. The changes you had me made increased the ATI card to ~ 2500 ppd.

Now the SMP clients... I'm waiting for them to finish what they have for the new cores. I notice that my CPU is not running at full speed. I have a display that shows each core and they are all ziging around 20-50% with the average being about 40%. Is there a reason they are not using more cycles?

Should CPU affinity be yes or no?

Thank you in advance!
Click to expand...


make sure your using the -smp flag on the shortcut..

cpu affinity should be no i believe..

should look like this in your config.. make sure you use your passkey as well..



smpconfig.png


to get into the config again change the -smp flag in the shortcut to -configonly
 
Last edited:
MaddBomber83 said:
Now the SMP clients... I'm waiting for them to finish what they have for the new cores. I notice that my CPU is not running at full speed. I have a display that shows each core and they are all ziging around 20-50% with the average being about 40%. Is there a reason they are not using more cycles?
Click to expand...
By default there should be use of all available CPU cycles in your system. How many SMP clients are you running? If it's only one machine there should be only one SMP client. Make sure you are using the -smp flag as sirmonkey advised and keep only one client when processing SMP WUs. If you are running a different setup than I described, wait until the current WUs complete and then change it.
 
APOLLO said:
By default there should be use of all available CPU cycles in your system. How many SMP clients are you running?
Click to expand...

To be clear I have 2 computers. Each computer has 1 SMP Client and 1 GPU client running on it. I'm using HFM.Net to monitor the PPD.

Below is my SMP Client
"C:\Folding@home SMP\Folding@home-Win32-x86.exe" -smp -bigadv -config

[settings]
username=maddbomber83
team=33
passkey="mypasskey"
asknet=no
machineid=1
bigpackets=big
local=31

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

[core]
cpuusage=100
nocpulock=1
checkpoint=5
priority=0
addr=

[clienttype]
type=3

After making the adjustments both computers are doing ~350 ppd on the SMP client. The ATI card is about 3k and the NVidea card is about 4k.

Both SMP Clients are not using all the CPU cycles.
 
MaddBomber83 said:
To be clear I have 2 computers. Each computer has 1 SMP Client and 1 GPU client running on it. I'm using HFM.Net to monitor the PPD.
Click to expand...
My mistake for misreading your posts. :eek:

After making the adjustments both computers are doing ~350 ppd on the SMP client. The ATI card is about 3k and the NVidea card is about 4k.
Click to expand...
Both GPU clients appear to be OK with the productions you posted. The SMP clients are way below normal especially for recent architectures...

Both SMP Clients are not using all the CPU cycles.
Click to expand...
Undoubtedly, that's the reason your production is so low. I'm at a loss to explain this result if you've indeed used the -smp flag in your shortcut. The only suggestion I can make is to check if the specific flags for the CPU clients have been set at the proper place, i.e. in the client shortcut's target box. Perhaps posting a portion of the log of one of your SMP clients could shed some light?
 
Thank you for your help!

The flags are in the Target: box on the shortcut tab.


Code: 
--- Opening Log file [September 26 15:00:10 UTC] 


# Windows CPU Console Edition #################################################
###############################################################################

                       Folding@Home Client Version 6.30

                          http://folding.stanford.edu

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

Launch directory: C:\Folding@home SMP
Service: C:\Folding@home SMP\Folding@home-Win32-x86.exe
Arguments: -svcstart -d C:\Folding@home SMP 

Launched as a service.
Entered C:\Folding@home SMP to do work.

[15:00:10] - Ask before connecting: No
[15:00:10] - User name: maddbomber83 (Team 33)
[15:00:10] - User ID: 49B133BE05536607
[15:00:10] - Machine ID: 1
[15:00:10] 
[15:00:10] Loaded queue successfully.
[15:00:10] 
[15:00:10] + Processing work unit
[15:00:10] Core required: FahCore_b4.exe
[15:00:10] Core found.
[15:00:10] Working on queue slot 07 [September 26 15:00:10 UTC]
[15:00:10] + Working ...
[15:00:14] *********************** Log Started 26/Sep/2010 15:00:14 ***********************
[15:00:14] ************************** ProtoMol Folding@Home Core **************************
[15:00:14]   Version: 25
[15:00:14]      Type: 180
[15:00:14]      Core: ProtoMol
[15:00:14]   Website: http://folding.stanford.edu/
[15:00:14] Copyright: (c) 2009 Stanford University
[15:00:14]    Author: Joseph Coffland <joseph@cauldrondevelopment.com>
[15:00:14]      Args: -dir work/ -suffix 07 -nice 19 -checkpoint 5 -service -lifeline 1648
[15:00:14]            -version 630
[15:00:14] ************************************ Build *************************************
[15:00:14]      Date: May 18 2010
[15:00:14]      Time: 23:43:52
[15:00:14]  Revision: 1819
[15:00:14]  Compiler: Intel(R) C++ MSVC 1500 mode 1110
[15:00:14]   Options: /TP /nologo /EHsc /wd4297 /wd4103 /wd1786 /arch:IA32 /Ox
[15:00:14]            /QaxSSE2,SSE3,SSSE3,SSE4.1,SSE4.2 /Qrestrict /MT
[15:00:14]   Defines: _CRT_SECURE_NO_WARNINGS NDEBUG HAVE_GEEKINFO BOOST_ALL_NO_LIB
[15:00:14]            XML_STATIC HAVE_EXPAT HAVE_OPENSSL HAVE_LIBFAH HAVE_SIMTK_LAPACK
[15:00:14]  Platform: Windows XP
[15:00:14]      Bits: 32
[15:00:14]      Mode: Release
[15:00:14] ************************************ System ************************************
[15:00:14]        OS: Microsoft Windows 7 Home Premium
[15:00:14]       CPU: Intel(R) Core(TM) i7 CPU 920 @ 2.67GHz
[15:00:14]    CPU ID: GenuineIntel Family 6 Model 26 Stepping 4
[15:00:14]      CPUs: 16 Logical, 1 Physical
[15:00:14]    Memory: 5.99 GB
[15:00:14]   Threads: Windows
[15:00:14] ********************************************************************************
[15:00:15] Project: 10055 (Run 911, Clone 0, Gen 18)
[15:00:15] Unit: 0x0000001a0001329c4c61915e0000365a
[15:00:15] User: 0x00000000000000000000000000000000
[15:00:15] Machine: 1
[15:00:15] Digital signatures verified
[15:00:16] Completed 201000 out of 499375 steps (40%)
[15:03:50] Completed 204800 out of 499375 steps (41%)
[15:08:33] Completed 209800 out of 499375 steps (42%)
[15:13:20] Completed 214700 out of 499375 steps (42%)
[15:18:35] Completed 219700 out of 499375 steps (43%)
[15:23:57] Completed 224700 out of 499375 steps (44%)
[15:29:19] Completed 229700 out of 499375 steps (45%)
[15:34:10] Completed 234700 out of 499375 steps (46%)
[15:38:50] Completed 239700 out of 499375 steps (48%)
[15:43:36] Completed 244700 out of 499375 steps (49%)
[15:48:16] Completed 249700 out of 499375 steps (50%)
[15:52:48] Completed 254700 out of 499375 steps (51%)
[15:57:22] Completed 259700 out of 499375 steps (52%)
[16:02:14] Completed 264700 out of 499375 steps (53%)
[16:07:09] Completed 269700 out of 499375 steps (54%)
 
OK, it's a good thing I inquired about a log. You are working on a standard uniprocessor WU not SMP. It explains both the low PPD and why you are not seeing full utilization. In the arguments section of the log near the top should list the -smp flag but it isn't there.
 
After posting that log file I was looking at it too.

When I use my shortcut it has the -smp and -bigadv flags there, but when the service starts it does not.

So... rather than use the -smp flag in the shortcut I put it at the very end of the advanced options when it asks you for additional parameters. I then deleted my work folder so it would download a new file and now I have 100% cpu use on all cores. I'll have to give it a while and see what my PPD is now.

Thank you for your help! Hopefully this fixed it.
 
MaddBomber83 said:
When I use my shortcut it has the -smp and -bigadv flags there, but when the service starts it does not.

So... rather than use the -smp flag in the shortcut I put it at the very end of the advanced options when it asks you for additional parameters. I then deleted my work folder so it would download a new file and now I have 100% cpu use on all cores. I'll have to give it a while and see what my PPD is now.

Thank you for your help! Hopefully this fixed it.
Click to expand...
No problem. If you are seeing all cores engaged at the very least you have a SMP WU. To receive -bigadv you will also need to use that flag in conjunction with the -smp flag. Which WU did your client download, BTW?
 
Service: C:\Folding@home SMP\Folding@home-Win32-x86.exe
Arguments: -svcstart -d C:\Folding@home SMP
Click to expand...

3 days late to the thread and it looks like you're already all set but ......

Your service was starting with SMP instead of -SMP, missing the - That's why it was running the classic client.

Welcome aboard! :D
 
CraftyOtter said:
3 days late to the thread and it looks like you're already all set but ......

Your service was starting with SMP instead of -SMP, missing the - That's why it was running the classic client.

Welcome aboard! :D
Click to expand...


Kinda "C:\Folding@home SMP" was the directory. There was no -smp flag on the service even though the shortcut had a -smp flag.

Code: 
Arguments: -svcstart -d C:\Folding@Home SMP -smp -bigadv
is how it looks now.

Still waiting for her computer to get through its first % to see what PPD it is, but my SMP client is doing ~8k now.
 
MaddBomber83 said:
Still waiting for her computer to get through its first % to see what PPD it is, but my SMP client is doing ~8k now.
Click to expand...
The production indicates it's a standard SMP WU. It can take a long while to receive a -bigadv WU because Stanford isn't always sending these units. With the flag you've set you should see one eventually.
 
I would not set the -bigadv flag until you have finished 10 SMP WU with your passkey, otherwise that is a bunch of work for very little points.
 
You must log in or register to reply here.
Back
Top