Hardware preformance

[luke738]

I use FAH GPU Tracker v2 and have noticed that the PPD on my i7-2600k 4.5Ghz has dropped down to only 8000PPD. Is this a problem with the program or my CPU and is there a way to fix it? I also have my GTX 580 running at the same time, and I'm sure it isn't overheating.

[Jesse_V]

Hi there luke738, welcome to the F@h support forum!

Can you post your log in between Code tags? Like post your log in between

Code: Select all

. That way we can better see what's going on with your client.
What client are you using? What PPD is normal for you?
If you are using a passkey, perhaps you've fallen below the 80% successful return rate needed to get bonus points. Without more information, that is just speculation but it might explain your PPD drop.

[luke738]

Code: Select all

 

--- Opening Log file [January 18 02:58:36 UTC] 


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

                       Folding@Home Client Version 6.34

                          http://folding.stanford.edu

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

Launch directory: C:\Users\Luke\Desktop\F@H\SMP
Executable: C:\Users\Luke\Desktop\F@H\FAH.exe
Arguments: -oneunit -forceasm -smp 6 

[02:58:36] - Ask before connecting: No
[02:58:36] - User name: luke738 (Team 0)
[02:58:36] - User ID: 7A43691213427B42
[02:58:36] - Machine ID: 2
[02:58:36] 
[02:58:36] Work directory not found. Creating...
[02:58:36] Could not open work queue, generating new queue...
[02:58:36] - Preparing to get new work unit...
[02:58:36] Cleaning up work directory
[02:58:36] + Attempting to get work packet
[02:58:36] Passkey found
[02:58:36] - Connecting to assignment server
[02:58:36] - Successful: assigned to (171.64.65.75).
[02:58:36] + News From Folding@Home: Welcome to Folding@Home
[02:58:36] Loaded queue successfully.
[02:58:38] + Closed connections
[02:58:38] 
[02:58:38] + Processing work unit
[02:58:38] Core required: FahCore_a3.exe
[02:58:38] Core found.
[02:58:38] Working on queue slot 01 [January 18 02:58:38 UTC]
[02:58:38] + Working ...
[02:58:38] 
[02:58:38] *------------------------------*
[02:58:38] Folding@Home Gromacs SMP Core
[02:58:38] Version 2.27 (Dec. 15, 2010)
[02:58:38] 
[02:58:38] Preparing to commence simulation
[02:58:38] - Assembly optimizations manually forced on.
[02:58:38] - Not checking prior termination.
[02:58:38] - Expanded 863220 -> 1568576 (decompressed 181.7 percent)
[02:58:38] Called DecompressByteArray: compressed_data_size=863220 data_size=1568576, decompressed_data_size=1568576 diff=0
[02:58:38] - Digital signature verified
[02:58:38] 
[02:58:38] Project: 10133 (Run 92, Clone 0, Gen 35)
[02:58:38] 
[02:58:38] Assembly optimizations on if available.
[02:58:38] Entering M.D.
[02:58:44] Mapping NT from 6 to 6 
[02:58:44] Completed 0 out of 2000000 steps  (0%)

Folding@Home Client Shutdown at user request.

Folding@Home Client Shutdown.


--- Opening Log file [January 18 03:05:57 UTC] 


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

                       Folding@Home Client Version 6.34

                          http://folding.stanford.edu

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

Launch directory: C:\Users\Luke\Desktop\F@H\SMP
Executable: C:\Users\Luke\Desktop\F@H\FAH.exe
Arguments: -oneunit -forceasm -smp 6 

[03:05:57] - Ask before connecting: No
[03:05:57] - User name: luke738 (Team 0)
[03:05:57] - User ID: 7A43691213427B42
[03:05:57] - Machine ID: 2
[03:05:57] 
[03:05:58] Loaded queue successfully.
[03:05:58] 

I am using the SMP client, allowing it to use all availible cores as well. My usual ppd is around 17000, but I have seen others getting more with my CPU and clock speed. I don't think it could be the passkey thing as I've only finished one WU on my cpu (and about 3 on my GPU). Just noticed the, but FAH GPU tracker is now reporting 10000ppd, seems odd.

[Jesse_V]

Ah. I've never used FAH GPU Tracker, but I believe that it is assuming that you qualify for bonus points. I checked your stats and I see that you've only completed a few WUs. In order to get bonus points (which will significantly boost your PPD) you need to meet the following criteria:
1) Use a passkey with a specific username
2) Complete 10 WUs using the passkey and the username
3) Maintain an 80% successful return rate on that passkey/username combo

If you don't meet the criteria, you won't get bonus points. The Tracker assumes that you do, so it's overestimating your PPD. But your real stats show much less. Continue folding, and I believe that this problem will go away by itself. Unless there's a specific reason for doing this, I don't believe you necessarily need the -forceasm flag. It forcefully enables some optimizations, but those optimizations should be on by default anyway (unless there's some kind of rare glitch) so I'm not sure that there's a real need for that flag. But you can leave it on if you want, it's not hurting anything. Also, the 2600K has 4 real cores but with hyperthreading it shows as 8 cores, so -smp 8 is truly the highest you can go. Have you determined that -smp 6 will get you better PPD? If you remove it altogether, it will detect the number of cores (8) and use -smp 8 by default. That might get you better PPD, but I believe that changing it will restart your progress on the WU; a safety feature to ensure scientific integrity of that WU.

[luke738]

Thanks for the help, I wait to see if that's just the problem. Also, in FAH GPU Tracker I can't see an option to disable -forceasm. I actually never determined -smp 6 was better, I think I had it set there as I was doing something slightly more cpu intensive and set it that way. Also, FAH GPU Tracker does let you change that setting mid WU, though that might just mean it finishes that WU and applies the setting on the next. Anyway, Thanks.

[bruce]

In FAH GPU Tracker I can't see an option to disable -forceasm.

It's not important. The SMP cores do the exact same computations whether you have that flag or not.

[PantherX]

...I don't believe you necessarily need the -forceasm flag. It forcefully enables some optimizations, but those optimizations should be on by default anyway (unless there's some kind of rare glitch) so I'm not sure that there's a real need for that flag...

FYI, except FahCore_78, the other FAHCores (a3, a4, a5) have the optimizations hard-coded in them. It means that they can't be disabled at all so the FahCore will always runs with optimizations on regardless of how it was terminated.