Two identical servers, different performance

[grizli]

Hello everyone,
So I have 2 boxes that have been folding for a while now. Since day one - machine A has been producing more PPD than second box.
They have been ordered together, identical everything.
Machine A PPD - 36k
Machine B PPD - 25k

The difference is significant, so I decided to see if someone can suggest how to resolve this and make machine B perform the same as A.

Specs on both machines are:
2xE5620, 12Gb RAM.

Thanks

[Macaholic]

Welcome to the forums. Have you checked your log files? It is highly unlikely that both machines have worked on the same project work units. Various units are worth varying points depending on several factors. That would explain the difference in points you are seeing.

[7im]

Hello grizli, welcome to the forum.

As another helpful diagnostic step, please post the first 50 lines of each of the fahlog.txt files, starting with the ############### Client Version Here ###############

[spitter3]

Switch hard drives, if the machines are the same its the quickest way to find out if its a programming issue or a hardware issue!!

[Nathan_P]

Switch hard drives, if the machines are the same its the quickest way to find out if its a programming issue or a hardware issue!!

Why would he need to do that. HDD performance is virtually meaningless for folding.

Client configuration - only viewable from the asked for log files, incorrect cpu/ram set up, heat throttling and different PPD from different projects are all far more likely problems.

[Grandpa_01]

Swapping hard drives could rule out client configuration. If you swap the drives and machine A drops to 25K and machine B goes up to 35K then it is definitely client configuration. But if all remains the same that only leaves one alternative. Hardware which is highly Likely since no to CPU's, sets of ram etc. that I have ever seen preformed exactly the same.

[7im]

Might be easier just to swap the FAH folders.

[k1wi]

Swapping folders might be a good idea, because it could be something as simple as a heatsink being incorrectly mounted, or dodgy ram. I note that it's a dual-cpu set up which I imagine could add complexity somewhat considerably.

[Mactin]

Hello Grizli,
Welcome to the forums.
For How long ? The gap is big but might not me statisticaly significant if the time frame is not long enogh.
H

[grizli]

Here are 2 latest starts of each machine from FAHlog.txt:
Machine A: (faster one)

Code: Select all

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

                       Folding@Home Client Version 6.34

                          http://folding.stanford.edu

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

Launch directory: C:\temp\fa
Executable: FAH6.34-win32-SMP.exe
Arguments: -smp 16 -bigadv 

[14:32:08] - Ask before connecting: No
[14:32:08] - User name: grizli 
[14:32:08] - User ID: xxxxxxxxxxxxxxxxxxx
[14:32:08] - Machine ID: 1
[14:32:08] 
[14:32:08] Loaded queue successfully.
[14:32:08] - Preparing to get new work unit...
[14:32:08] Project: 6900 (Run 13, Clone 10, Gen 34)
[14:32:08] Cleaning up work directory


[14:32:08] + Attempting to send results [September 5 14:32:08 UTC]
[14:32:09] + Attempting to get work packet
[14:32:09] Passkey found
[14:32:09] - Connecting to assignment server
[14:32:09] - Successful: assigned to (171.67.108.22).
[14:32:09] + News From Folding@Home: Welcome to Folding@Home
[14:32:10] Loaded queue successfully.
[14:32:58] + Closed connections
[14:32:58] 
[14:32:58] + Processing work unit
[14:32:58] Core required: FahCore_a5.exe
[14:32:58] Core found.
[14:32:58] Working on queue slot 06 [September 5 14:32:58 UTC]
[14:32:58] + Working ...
[14:32:58] 
[14:32:58] *------------------------------*
[14:32:58] Folding@Home Gromacs SMP Core
[14:32:58] Version 2.27 (Mar 12, 2010)
[14:32:58] 
[14:32:58] Preparing to commence simulation
[14:32:58] - Looking at optimizations...
[14:32:58] - Created dyn
[14:32:58] - Files status OK
[14:33:05] - Expanded 25468947 -> 31941441 (decompressed 125.4 percent)
[14:33:05] Called DecompressByteArray: compressed_data_size=25468947 data_size=31941441, decompressed_data_size=31941441 diff=0
[14:33:05] - Digital signature verified
[14:33:05] 
[14:33:05] Project: 2686 (Run 6, Clone 9, Gen 150)
[14:33:05] 
[14:33:05] Assembly optimizations on if available.
[14:33:05] Entering M.D.
[14:33:12] Mapping NT from 16 to 16 
[14:33:16] Completed 0 out of 250000 steps  (0%)
[15:00:33] Completed 2500 out of 250000 steps  (1%)

Machine B:(slower one)

Code: Select all

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

                       Folding@Home Client Version 6.34

                          http://folding.stanford.edu

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

Launch directory: C:\temp\fa
Executable: FAH6.34-win32-SMP.exe
Arguments: -smp 16 -bigadv 

[12:25:24] - Ask before connecting: No
[12:25:24] - User name: grizli
[12:25:24] - User ID not found locally
[12:25:24] + Requesting User ID from server
[12:25:25] - Machine ID: 1
[12:25:25] 
[12:25:25] Loaded queue successfully.
[12:25:25] 
[12:25:25] + Processing work unit
[12:25:25] Core required: FahCore_a5.exe
[12:25:25] Core found.
[12:25:25] Working on queue slot 01 [September 20 12:25:25 UTC]
[12:25:25] + Working ...
[12:25:25] 
[12:25:25] *------------------------------*
[12:25:25] Folding@Home Gromacs SMP Core
[12:25:25] Version 2.27 (Mar 12, 2010)
[12:25:25] 
[12:25:25] Preparing to commence simulation
[12:25:25] - Ensuring status. Please wait.
[12:25:35] - Looking at optimizations...
[12:25:35] - Working with standard loops on this execution.
[12:25:35] - Previous termination of core was improper.
[12:25:35] - Files status OK
[12:25:41] - Expanded 24869644 -> 30796292 (decompressed 123.8 percent)
[12:25:41] Called DecompressByteArray: compressed_data_size=24869644 data_size=30796292, decompressed_data_size=30796292 diff=0
[12:25:42] - Digital signature verified
[12:25:42] 
[12:25:42] Project: 6900 (Run 3, Clone 24, Gen 57)
[12:25:42] 
[12:25:42] Entering M.D.
[12:25:48] Using Gromacs checkpoints
[12:25:49] Mapping NT from 16 to 16 
[12:26:00] Resuming from checkpoint
[12:26:01] Verified work/wudata_01.log
[12:26:03] Verified work/wudata_01.trr
[12:26:03] Verified work/wudata_01.xtc
[12:26:03] Verified work/wudata_01.edr
[12:26:04] Completed 192860 out of 250000 steps  (77%)
[13:11:06] Completed 195000 out of 250000 steps  (78%)

One thing I noticed - Machine B could not find UserID??? Why is that? Both have Passkey in client.cfg files.

Machines have been folding for over a month now.

Another thing to consider - I have another 2 boxes that are the same:
2xE5335 with 8Gb of RAM. The same story - 11K PPD vs 8K PPD.

Mod Edit: Added Code Tags - PantherX

[7im]

[12:25:24] + Requesting User ID from server

The above message should only happen once, when the client is first installed.

Since the client has been running for a month, that indicates a problem. The client needs to be run with a user account that has accesss to write to the Windows Registry.

[Magic Michael]

Have you checked if all cores are used ? Sometimes my Xeons have one core stuck in C1/Halt mode until I restart the folding client. CentOS though, not Windoze.

[grizli]

[12:25:24] + Requesting User ID from server

The above message should only happen once, when the client is first installed.

Since the client has been running for a month, that indicates a problem. The client needs to be run with a user account that has accesss to write to the Windows Registry.

I just restarted the client - went through fine this time. It was running under administrator account. Not sure what that was all about, but PPD still hasn't changed

[grizli]

Have you checked if all cores are used ? Sometimes my Xeons have one core stuck in C1/Halt mode until I restart the folding client. CentOS though, not Windoze.

I checked via Task manager - all 100%

[i7GTX550Ti8Gb]

seems 2 different projects to me, and what i noticed is that voth your machines are configured as machine number 1?

Not sure wether it helped, not a server expert

me