Weak PPDs on CPU with SMP6

AlexDeGruven

Just started getting back into the F@H game. I got the v7 FAHControl up and running, and my GPU (Radeon 6870) seems to be going along as expected (~7200PPD), but the CPU PPD seems REALLY weak (~1000PPD).

Am I doing something wrong in the setup? All 6 cores (FX-6100) are running 100%, so I know it's using them.

Tushon

FX had notoriously bad PPD performance compared to previous generation, doubly so considering the increased power draw. I didn't find anything to back up that statement online, but I remember it going around. Post the log from SMP slot (might have to turn it on expert mode) and filter to SMP only (I've never done this, just heard, might not actually be possible)

AlexDeGruven

Strange. I can understand things can work differently between processor generations, but only generating 10% of a comparable previous-gen CPU seems really off to me.

Here's the log for slot 01 (CPU) on my current FAHControl:

*********************** Log Started 2012-03-09T19:15:16Z ***********************
19:15:16:WU01:FS01:Cleaning up
19:15:17:WU01:FS01:Connecting to assign3.stanford.edu:8080
19:15:17:WU01:FS01:News: Welcome to Folding@Home
19:15:17:WU01:FS01:Assigned to work server 171.64.65.101
19:15:17:WU01:FS01:Requesting new work unit for slot 01: READY smp:6 from 171.64.65.101
19:15:17:WU01:FS01:Connecting to 171.64.65.101:8080
19:15:18:WU01:FS01:Downloading 295.51KiB
19:15:19:WU01:FS01:Download complete
19:15:19:WU01:FS01:Received Unit: id:01 state:DOWNLOAD error:OK project:7600 run:3 clone:5 gen:38 core:0xa4 unit:0x00000040664f2dcd4dee89747ccf4627
19:15:19:WU01:FS01:Starting
19:15:19:WU01:FS01:Running FahCore: "C:\Program Files (x86)\FAHClient/FAHCoreWrapper.exe" C:/Users/Nathan/AppData/Roaming/FAHClient/cores/www.stanford.edu/~pande/Win32/AMD64/Core_a4.fah/FahCore_a4.exe -dir 01 -suffix 01 -version 701 -lifeline 5192 -checkpoint 15 -np 6
19:15:19:WU01:FS01:Started FahCore on PID 4812
19:15:19:WU01:FS01:Core PID:5696
19:15:19:WU01:FS01:FahCore 0xa4 started
19:15:19:WU01:FS01:0xa4:
19:15:19:WU01:FS01:0xa4:*------------------------------*
19:15:19:WU01:FS01:0xa4:Folding@Home Gromacs GB Core
19:15:19:WU01:FS01:0xa4:Version 2.27 (Dec. 15, 2010)
19:15:19:WU01:FS01:0xa4:
19:15:19:WU01:FS01:0xa4:Preparing to commence simulation
19:15:19:WU01:FS01:0xa4:- Looking at optimizations...
19:15:19:WU01:FS01:0xa4:- Created dyn
19:15:19:WU01:FS01:0xa4:- Files status OK
19:15:19:WU01:FS01:0xa4:- Expanded 302089 -> 505352 (decompressed 167.2 percent)
19:15:19:WU01:FS01:0xa4:Called DecompressByteArray: compressed_data_size=302089 data_size=505352, decompressed_data_size=505352 diff=0
19:15:19:WU01:FS01:0xa4:- Digital signature verified
19:15:19:WU01:FS01:0xa4:
19:15:19:WU01:FS01:0xa4:Project: 7600 (Run 3, Clone 5, Gen 38)
19:15:19:WU01:FS01:0xa4:
19:15:19:WU01:FS01:0xa4:Assembly optimizations on if available.
19:15:19:WU01:FS01:0xa4:Entering M.D.
19:15:25:WU01:FS01:0xa4:Mapping NT from 6 to 6
19:15:26:WU01:FS01:0xa4:Completed 0 out of 2000000 steps (0%)
19:49:14:WU01:FS01:0xa4:Completed 20000 out of 2000000 steps (1%)
20:18:05:WU01:FS01:0xa4:Completed 40000 out of 2000000 steps (2%)

_k

14-19k PPD on 8150 @ 4.5 GHz for reference

You are at 1367.61ppd

AlexDeGruven

Yeah, but I have 6 cores. Even dropping one core to handle the GPU3 stuff shouldn't have me down at 1k, should it?

_k

You caught me before I could edit, I originally read OP as 10k.

Run either the CLI client out of v7 or f@htracker to verify.

AlexDeGruven

Hmph... f@htracker shows similar results. Can't quite figure out what I should be using for a command-line.

Tushon

Try changing the slot settings for SMP to

client-type
advanced

max-packet-size
big

And delete the current WU (not sure how to do that, I need to play with v7 more)

AlexDeGruven

Did that, and it spiked up for a moment, but then started creeping back down toward 1k.

Fake Edit: Down to ~200 now. Weird.
Fake Edit2: After that dive, it's now cranking up. At about 12k now. Seems much better.
Fake Edit3: Now up to 15.5k, so I think we're in the right space now.
Last Edit: Turned GPU back on, and we're up over 20k now.

*********************** Log Started 2012-03-09T21:20:44Z ***********************
21:20:44:WU01:FS01:Cleaning up
21:20:45:WU01:FS01:Connecting to assign3.stanford.edu:8080
21:20:45:WU01:FS01:News: Welcome to Folding@Home
21:20:45:WU01:FS01:Assigned to work server 129.74.85.15
21:20:45:WU01:FS01:Requesting new work unit for slot 01: READY smp:6 from 129.74.85.15
21:20:45:WU01:FS01:Connecting to 129.74.85.15:8080
21:20:46:WU01:FS01:Downloading 37.56KiB
21:20:46:WU01:FS01:Download complete
21:20:46:WU01:FS01:Received Unit: id:01 state:DOWNLOAD error:OK project:10090 run:76 clone:2 gen:83 core:0xa4 unit:0x0000014b0001329c4dfb541a08a34df6
21:20:46:WU01:FS01:Starting
21:20:46:WU01:FS01:Running FahCore: "C:\Program Files (x86)\FAHClient/FAHCoreWrapper.exe" C:/Users/Nathan/AppData/Roaming/FAHClient/cores/www.stanford.edu/~pande/Win32/AMD64/Core_a4.fah/FahCore_a4.exe -dir 01 -suffix 01 -version 701 -lifeline 5980 -checkpoint 15 -np 6
21:20:46:WU01:FS01:Started FahCore on PID 3728
21:20:46:WU01:FS01:Core PID:2456
21:20:46:WU01:FS01:FahCore 0xa4 started
21:20:46:WU01:FS01:Downloading project 10090 description
21:20:46:WU01:FS01:Connecting to fah-web.stanford.edu:80
21:20:47:WU01:FS01:0xa4:
21:20:47:WU01:FS01:0xa4:*------------------------------*
21:20:47:WU01:FS01:0xa4:Folding@Home Gromacs GB Core
21:20:47:WU01:FS01:0xa4:Version 2.27 (Dec. 15, 2010)
21:20:47:WU01:FS01:0xa4:
21:20:47:WU01:FS01:0xa4:Preparing to commence simulation
21:20:47:WU01:FS01:0xa4:- Looking at optimizations...
21:20:47:WU01:FS01:0xa4:- Created dyn
21:20:47:WU01:FS01:0xa4:- Files status OK
21:20:47:WU01:FS01:0xa4:- Expanded 37946 -> 194344 (decompressed 512.1 percent)
21:20:47:WU01:FS01:Project 10090 description downloaded successfully
21:20:47:WU01:FS01:0xa4:Called DecompressByteArray: compressed_data_size=37946 data_size=194344, decompressed_data_size=194344 diff=0
21:20:47:WU01:FS01:0xa4:- Digital signature verified
21:20:47:WU01:FS01:0xa4:
21:20:47:WU01:FS01:0xa4:Project: 10090 (Run 76, Clone 2, Gen 83)
21:20:47:WU01:FS01:0xa4:
21:20:47:WU01:FS01:0xa4:Assembly optimizations on if available.
21:20:47:WU01:FS01:0xa4:Entering M.D.
21:20:52:WU01:FS01:0xa4:Mapping NT from 6 to 6
21:21:05:WU01:FS01:0xa4:Completed 0 out of 10000000 steps (0%)

Tushon

I'm guessing the FX client needs the advanced WUs to take advantage, but who knows.

AlexDeGruven

Something is still not right. COuld still be FX related. PPD is down to 2600 for the CPU. I saw it over 22k as I was leaving work today.

Edit: Once it finishes this WU, I'm going to knock it down to 4 and see if that helps, since it appears GPU tends to use quite a bit of CPU.

Tushon

Yeah, unfortunately the AMD cores are still extremely CPU intensive (as in, they locked up an entire core on my computer for the brief time I had an AMD GPU in my box). When you drop it to 4, you may want to try and specify that the GPU uses core 6 or something like that.

csimon

What if he/you dropped it to 3 cores and enabled turbo core?
I'm in about the same range ...about 3k ppd on my 1100T @ 3.8ghz (6 cores).