Page 1 of 1

Returning to F@H

Posted: Fri Dec 19, 2014 3:44 am
by Kornflake
Hello all! I'm returning to F@H! I bought a new Alienware Area-51 from Dell primarily for F@H but I might do some gaming with it too. I didn't want to bother building it and now I have support if something goes wrong: even though it cost me extra!
  • Windows 8.1 Pro, 64Bit
    Intel Core i7-5930K Processor (6-cores, 15MB Cache, Overclocked up to 3.9 GHz w/ Turbo Boost) (F@H says I have 12 CPU's)
    Triple NVIDIA GeForce GTX 980 graphics with 12GB total (3x 4GB) GDDR5 -NVIDIA SLI Enabled (came disabled which is what I wanted!)
    32GB Quad Channel DDR4 at 2133MHz
    512GB SSD 6Gb/s Main + 4TB 6kRPM SATA 6Gb/s Storage
    Alienware 1500 Watt Multi-GPU Approved Power Supply
The purpose of this post is to try and confirm some information I've gathered since returning. I'm running the 344.80 driver because it contained specific fixes for my DisplayPort 4k Monitor setup.

Probably most notably it looks like I'm stuck in the situation documented here with the Maxwell cards and OpenCL: https://forums.geforce.com/default/topi ... ver-bug/1/
I posted in that forum and also here to try and make some progress on that front:
viewtopic.php?f=16&t=26980&p=272269#p272269

The Core 18s are only estimated at 80K PPD.

It looks like a new web control was introduced since I last used F@H. Personally I'd prefer to use the original version 7 FAHControl as I couldn't see anything that the web control allowed me to do beyond what I can already do in the FAHControl. So I think it was safe for me to remove that Client Option.

I think it might make sense for my system to participate in Beta or Advanced but I want to get some feedback. I'm pretty good at reviewing my log files for "Warnings & Errors".

It looks like the install automatically withheld 3 of my "CPUs" from my CPU slot for running the GPUs because it says CPU:9 but I want to confirm. I get a small PPD increase on my GPU slots when pausing the CPU slot so it might be throttling it a bit. (approx 270K to 290K PPD on my Core 17 9201 project) My GPU temp is being held at 80C so I think that might also be limiting their performance some.

Here is my log, please review for any improvement considerations:

Code: Select all

*********************** Log Started 2014-12-18T09:01:41Z ***********************
09:01:41:************************* Folding@home Client *************************
09:01:41:      Website: http://folding.stanford.edu/
09:01:41:    Copyright: (c) 2009-2014 Stanford University
09:01:41:       Author: Joseph Coffland <joseph@cauldrondevelopment.com>
09:01:41:         Args: 
09:01:41:       Config: C:/ProgramData/FAHClient/config.xml
09:01:41:******************************** Build ********************************
09:01:41:      Version: 7.4.4
09:01:41:         Date: Mar 4 2014
09:01:41:         Time: 20:26:54
09:01:41:      SVN Rev: 4130
09:01:41:       Branch: fah/trunk/client
09:01:41:     Compiler: Intel(R) C++ MSVC 1500 mode 1200
09:01:41:      Options: /TP /nologo /EHa /Qdiag-disable:4297,4103,1786,279 /Ox -arch:SSE
09:01:41:               /QaxSSE2,SSE3,SSSE3,SSE4.1,SSE4.2 /Qopenmp /Qrestrict /MT /Qmkl
09:01:41:     Platform: win32 XP
09:01:41:         Bits: 32
09:01:41:         Mode: Release
09:01:41:******************************* System ********************************
09:01:41:          CPU: Intel(R) Core(TM) i7-5930K CPU @ 3.50GHz
09:01:41:       CPU ID: GenuineIntel Family 6 Model 63 Stepping 2
09:01:41:         CPUs: 12
09:01:41:       Memory: 31.89GiB
09:01:41:  Free Memory: 30.40GiB
09:01:41:      Threads: WINDOWS_THREADS
09:01:41:   OS Version: 6.2
09:01:41:  Has Battery: false
09:01:41:   On Battery: false
09:01:41:   UTC Offset: -5
09:01:41:          PID: 772
09:01:41:          CWD: C:/ProgramData/FAHClient
09:01:41:           OS: Windows 8.1 Pro
09:01:41:      OS Arch: AMD64
09:01:41:         GPUs: 3
09:01:41:        GPU 0: NVIDIA:4 GM204 [GeForce GTX 980]
09:01:41:        GPU 1: NVIDIA:4 GM204 [GeForce GTX 980]
09:01:41:        GPU 2: NVIDIA:4 GM204 [GeForce GTX 980]
09:01:41:         CUDA: 5.2
09:01:41:  CUDA Driver: 6050
09:01:41:Win32 Service: false
09:01:41:***********************************************************************
09:01:41:<config>
09:01:41:  <!-- Network -->
09:01:41:  <proxy v=':8080'/>
09:01:41:
09:01:41:  <!-- Slot Control -->
09:01:41:  <power v='FULL'/>
09:01:41:
09:01:41:  <!-- User Information -->
09:01:41:  <passkey v='********************************'/>
09:01:41:  <team v='182919'/>
09:01:41:  <user v='Kornflake'/>
09:01:41:
09:01:41:  <!-- Folding Slots -->
09:01:41:  <slot id='0' type='CPU'/>
09:01:41:  <slot id='1' type='GPU'/>
09:01:41:  <slot id='2' type='GPU'/>
09:01:41:  <slot id='3' type='GPU'/>
09:01:41:</config>
09:01:41:Trying to access database...
09:01:41:Successfully acquired database lock
09:01:41:Enabled folding slot 00: READY cpu:9
09:01:41:Enabled folding slot 01: READY gpu:0:GM204 [GeForce GTX 980]
09:01:41:Enabled folding slot 02: READY gpu:1:GM204 [GeForce GTX 980]
09:01:41:Enabled folding slot 03: READY gpu:2:GM204 [GeForce GTX 980]
09:01:42:WU00:FS00:Connecting to 171.67.108.200:8080
09:01:42:WU01:FS01:Connecting to 171.67.108.200:80
09:01:42:WU02:FS02:Connecting to 171.67.108.200:80
09:01:42:WU03:FS03:Connecting to 171.67.108.200:80
09:01:42:WU00:FS00:Assigned to work server 171.64.65.99
09:01:42:WU00:FS00:Requesting new work unit for slot 00: READY cpu:9 from 171.64.65.99
09:01:42:WU00:FS00:Connecting to 171.64.65.99:8080
09:01:43:WU03:FS03:Assigned to work server 171.67.108.52
09:01:43:WU03:FS03:Requesting new work unit for slot 03: READY gpu:2:GM204 [GeForce GTX 980] from 171.67.108.52
09:01:43:WU01:FS01:Assigned to work server 140.163.4.235
09:01:43:WU03:FS03:Connecting to 171.67.108.52:8080
09:01:43:WU01:FS01:Requesting new work unit for slot 01: READY gpu:0:GM204 [GeForce GTX 980] from 140.163.4.235
09:01:43:WU02:FS02:Assigned to work server 171.67.108.52
09:01:43:WU01:FS01:Connecting to 140.163.4.235:8080
09:01:43:WU02:FS02:Requesting new work unit for slot 02: READY gpu:1:GM204 [GeForce GTX 980] from 171.67.108.52
09:01:43:WU02:FS02:Connecting to 171.67.108.52:8080
09:01:43:WU01:FS01:Downloading 4.18MiB
09:01:43:WU02:FS02:Downloading 1.53MiB
09:01:43:WU03:FS03:Downloading 1.52MiB
09:01:44:WU00:FS00:Downloading 1.98MiB
09:01:47:WU02:FS02:Download complete
09:01:47:WU02:FS02:Received Unit: id:02 state:DOWNLOAD error:NO_ERROR project:9201 run:911 clone:2 gen:65 core:0x17 unit:0x0000006b6652edc45399f9d90741f5c0
09:01:47:WU02:FS02:Starting
09:01:47:WU02:FS02:Running FahCore: "C:\Program Files (x86)\FAHClient/FAHCoreWrapper.exe" C:/ProgramData/FAHClient/cores/web.stanford.edu/~pande/Win32/AMD64/NVIDIA/Fermi/Core_17.fah/FahCore_17.exe -dir 02 -suffix 01 -version 704 -lifeline 772 -checkpoint 15 -gpu 1 -gpu-vendor nvidia
09:01:47:WU02:FS02:Started FahCore on PID 5392
09:01:48:WU02:FS02:Core PID:5416
09:01:48:WU02:FS02:FahCore 0x17 started
09:01:48:WU02:FS02:0x17:*********************** Log Started 2014-12-18T09:01:48Z ***********************
09:01:48:WU02:FS02:0x17:Project: 9201 (Run 911, Clone 2, Gen 65)
09:01:48:WU02:FS02:0x17:Unit: 0x0000006b6652edc45399f9d90741f5c0
09:01:48:WU02:FS02:0x17:CPU: 0x00000000000000000000000000000000
09:01:48:WU02:FS02:0x17:Machine: 2
09:01:48:WU02:FS02:0x17:Reading tar file state.xml
09:01:48:WU02:FS02:0x17:Reading tar file system.xml
09:01:48:WU02:FS02:0x17:Reading tar file integrator.xml
09:01:48:WU02:FS02:0x17:Reading tar file core.xml
09:01:48:WU02:FS02:0x17:Digital signatures verified
09:01:48:WU02:FS02:0x17:Folding@home GPU core17
09:01:48:WU02:FS02:0x17:Version 0.0.52
09:01:49:WU01:FS01:Download 68.81%
09:01:49:WU03:FS03:Download 82.05%
09:01:49:WU03:FS03:Download complete
09:01:50:WU00:FS00:Download 59.86%
09:01:50:WU03:FS03:Received Unit: id:03 state:DOWNLOAD error:NO_ERROR project:9201 run:932 clone:4 gen:107 core:0x17 unit:0x000000936652edc45399fab23e1bdd2c
09:01:50:WU03:FS03:Starting
09:01:50:WU03:FS03:Running FahCore: "C:\Program Files (x86)\FAHClient/FAHCoreWrapper.exe" C:/ProgramData/FAHClient/cores/web.stanford.edu/~pande/Win32/AMD64/NVIDIA/Fermi/Core_17.fah/FahCore_17.exe -dir 03 -suffix 01 -version 704 -lifeline 772 -checkpoint 15 -gpu 2 -gpu-vendor nvidia
09:01:50:WU03:FS03:Started FahCore on PID 5444
09:01:50:WU03:FS03:Core PID:5456
09:01:50:WU03:FS03:FahCore 0x17 started
09:01:50:WU03:FS03:0x17:*********************** Log Started 2014-12-18T09:01:50Z ***********************
09:01:50:WU03:FS03:0x17:Project: 9201 (Run 932, Clone 4, Gen 107)
09:01:50:WU03:FS03:0x17:Unit: 0x000000936652edc45399fab23e1bdd2c
09:01:50:WU03:FS03:0x17:CPU: 0x00000000000000000000000000000000
09:01:50:WU03:FS03:0x17:Machine: 3
09:01:50:WU03:FS03:0x17:Reading tar file state.xml
09:01:50:WU03:FS03:0x17:Reading tar file system.xml
09:01:51:WU03:FS03:0x17:Reading tar file integrator.xml
09:01:51:WU03:FS03:0x17:Reading tar file core.xml
09:01:51:WU03:FS03:0x17:Digital signatures verified
09:01:51:WU03:FS03:0x17:Folding@home GPU core17
09:01:51:WU03:FS03:0x17:Version 0.0.52
09:01:51:WU01:FS01:Download complete
09:01:52:WU01:FS01:Received Unit: id:01 state:DOWNLOAD error:NO_ERROR project:10471 run:0 clone:171 gen:60 core:0x18 unit:0x0000004f538b3dbb53beab7486b30c0e
09:01:52:WU01:FS01:Downloading core from http://web.stanford.edu/~pande/Win32/AMD64/NVIDIA/Fermi/Core_18.fah
09:01:52:WU01:FS01:Connecting to web.stanford.edu:80
09:01:52:WU01:FS01:FahCore 18: Downloading 2.54MiB
09:01:56:WU00:FS00:Download 88.21%
09:01:57:WU00:FS00:Download complete
09:01:57:WU00:FS00:Received Unit: id:00 state:DOWNLOAD error:NO_ERROR project:7809 run:2 clone:99 gen:36 core:0xa4 unit:0x000000380a3b1e874e31094434291df3
09:01:57:WU00:FS00:Starting
09:01:57:WU00:FS00:Running FahCore: "C:\Program Files (x86)\FAHClient/FAHCoreWrapper.exe" C:/ProgramData/FAHClient/cores/web.stanford.edu/~pande/Win32/AMD64/Core_a4.fah/FahCore_a4.exe -dir 00 -suffix 01 -version 704 -lifeline 772 -checkpoint 15 -np 9
09:01:57:WU00:FS00:Started FahCore on PID 5608
09:01:57:WU00:FS00:Core PID:5620
09:01:57:WU00:FS00:FahCore 0xa4 started
09:01:58:WU01:FS01:FahCore 18: 91.20%
09:01:58:WU01:FS01:FahCore 18: Download complete
09:01:58:WU00:FS00:0xa4:
09:01:58:WU00:FS00:0xa4:*------------------------------*
09:01:58:WU00:FS00:0xa4:Folding@Home Gromacs GB Core
09:01:58:WU00:FS00:0xa4:Version 2.27 (Dec. 15, 2010)
09:01:58:WU00:FS00:0xa4:
09:01:58:WU00:FS00:0xa4:Preparing to commence simulation
09:01:58:WU00:FS00:0xa4:- Looking at optimizations...
09:01:58:WU00:FS00:0xa4:- Created dyn
09:01:58:WU00:FS00:0xa4:- Files status OK
09:01:58:WU00:FS00:0xa4:- Expanded 2079648 -> 5386224 (decompressed 258.9 percent)
09:01:58:WU00:FS00:0xa4:Called DecompressByteArray: compressed_data_size=2079648 data_size=5386224, decompressed_data_size=5386224 diff=0
09:01:58:WU00:FS00:0xa4:- Digital signature verified
09:01:58:WU00:FS00:0xa4:
09:01:58:WU00:FS00:0xa4:Project: 7809 (Run 2, Clone 99, Gen 36)
09:01:58:WU00:FS00:0xa4:
09:01:58:WU00:FS00:0xa4:Assembly optimizations on if available.
09:01:58:WU00:FS00:0xa4:Entering M.D.
09:01:58:WU01:FS01:Valid core signature
09:01:58:WU01:FS01:Unpacked 8.47MiB to cores/web.stanford.edu/~pande/Win32/AMD64/NVIDIA/Fermi/Core_18.fah/FahCore_18.exe
09:01:59:WU01:FS01:Starting
09:01:59:WU01:FS01:Running FahCore: "C:\Program Files (x86)\FAHClient/FAHCoreWrapper.exe" C:/ProgramData/FAHClient/cores/web.stanford.edu/~pande/Win32/AMD64/NVIDIA/Fermi/Core_18.fah/FahCore_18.exe -dir 01 -suffix 01 -version 704 -lifeline 772 -checkpoint 15 -gpu 0 -gpu-vendor nvidia
09:01:59:WU01:FS01:Started FahCore on PID 5688
09:01:59:WU01:FS01:Core PID:5700
09:01:59:WU01:FS01:FahCore 0x18 started
09:02:00:WU01:FS01:0x18:*********************** Log Started 2014-12-18T09:02:00Z ***********************
09:02:00:WU01:FS01:0x18:Project: 10471 (Run 0, Clone 171, Gen 60)
09:02:00:WU01:FS01:0x18:Unit: 0x0000004f538b3dbb53beab7486b30c0e
09:02:00:WU01:FS01:0x18:CPU: 0x00000000000000000000000000000000
09:02:00:WU01:FS01:0x18:Machine: 1
09:02:00:WU01:FS01:0x18:Reading tar file state.xml
09:02:01:WU01:FS01:0x18:Reading tar file system.xml
09:02:01:WU01:FS01:0x18:Reading tar file integrator.xml
09:02:01:WU01:FS01:0x18:Reading tar file core.xml
09:02:01:WU01:FS01:0x18:Digital signatures verified
09:02:01:WU01:FS01:0x18:Folding@home GPU core18
09:02:01:WU01:FS01:0x18:Version 0.0.3
09:02:04:WU00:FS00:0xa4:Mapping NT from 9 to 9 
09:02:04:WU00:FS00:0xa4:Completed 0 out of 1500000 steps  (0%)
09:02:09:WU02:FS02:0x17:Completed 0 out of 5000000 steps (0%)
09:02:09:WU02:FS02:0x17:Temperature control disabled. Requirements: single Nvidia GPU, tmax must be < 110 and twait >= 900
09:02:12:WU03:FS03:0x17:Completed 0 out of 5000000 steps (0%)
09:02:12:WU03:FS03:0x17:Temperature control disabled. Requirements: single Nvidia GPU, tmax must be < 110 and twait >= 900
09:02:32:WU01:FS01:0x18:Completed 0 out of 5000000 steps (0%)
09:02:32:WU01:FS01:0x18:Temperature control disabled. Requirements: single Nvidia GPU, tmax must be < 110 and twait >= 900
09:03:55:WU02:FS02:0x17:Completed 50000 out of 5000000 steps (1%)
09:04:00:WU03:FS03:0x17:Completed 50000 out of 5000000 steps (1%)
09:05:42:WU02:FS02:0x17:Completed 100000 out of 5000000 steps (2%)
09:05:49:WU03:FS03:0x17:Completed 100000 out of 5000000 steps (2%)
09:07:30:WU02:FS02:0x17:Completed 150000 out of 5000000 steps (3%)
09:07:39:WU03:FS03:0x17:Completed 150000 out of 5000000 steps (3%)
09:09:02:WU00:FS00:0xa4:Completed 15000 out of 1500000 steps  (1%)
09:09:19:WU02:FS02:0x17:Completed 200000 out of 5000000 steps (4%)
09:09:29:WU03:FS03:0x17:Completed 200000 out of 5000000 steps (4%)
09:11:08:WU02:FS02:0x17:Completed 250000 out of 5000000 steps (5%)
09:11:19:WU03:FS03:0x17:Completed 250000 out of 5000000 steps (5%)
09:12:16:WU01:FS01:0x18:Completed 50000 out of 5000000 steps (1%)
09:12:58:WU02:FS02:0x17:Completed 300000 out of 5000000 steps (6%)
09:13:09:WU03:FS03:0x17:Completed 300000 out of 5000000 steps (6%)
09:14:48:WU02:FS02:0x17:Completed 350000 out of 5000000 steps (7%)
Anything else I should be aware of?

Thanks!

Re: Returning to F@H

Posted: Fri Dec 19, 2014 7:18 am
by bruce
Nice system.

Your observations pretty much covers it.

Depending on when the drivers are fixed and/or when specific changes are made to the FahCores. your GPUs will do most of your folding. While folding with 6 or with 9 or even more threads will change the amount of work produced by your i7, but it's going to be a very small addition so it's not really that important, particularly since you have hardware for six SSE/FP processors which are shared with pairs of logical CPUs. I'd leave it at 9.

Re: Returning to F@H

Posted: Fri Dec 19, 2014 8:58 am
by Napoleon
Kornflake wrote:I get a small PPD increase on my GPU slots when pausing the CPU slot so it might be throttling it a bit. (approx 270K to 290K PPD on my Core 17 9201 project) My GPU temp is being held at 80C so I think that might also be limiting their performance some.
FahCore_17 seems to rely on the CPU to a greater extent than previous GPU cores. The initial preparation of a fresh WU is entirely CPU dependent (single-threaded performance), same goes for checkpointing and periodical validity checks during folding. It's not a huge amount of (CPU) time, but since such a large portion of PPD consists of QRB in your case, rather tiny linear changes can cause significant fluctuations in PPD.

AFAIK, FahCore_18 works around the NV OpenCL bug via emulation, so that should certainly be very much CPU dependent. The triple 980's being your biggest guns by a wide margin, you should optimize for them, which is to say that you may want to reserve a full CPU core for each.

That requires installing and configuring a 3rd party affinity/priority management utility like Process Lasso (I prefer WinAFC). Set up FahCore_1?.exe to run on logical CPUs 1,3 and 5. Use lCPUs 6-11 for CPU:6 (FahCore_a?.exe). That way each GPU feeder process should have exclusive access to a full CPU core, while CPU:6 is a little bit of extra - "because you can". That would leave lCPUs 0,2 and 4 available for monitoring utilities (FahControl and HFM for example) and CPU folding would have hardly any load imbalance, although CPU:9 will most likely be slightly faster than CPU:6. Be that as it may, the extra PPD from GPUs should more than make up for the slack. To make sure, use CoreInfo to verify the specifics of your CPU topology.

You could test the theory without installing any 3rd party utilities by tweaking affinities manually using Task Manager. FYI, Task Manager settings are lost when a new WU is started (or current ones are paused/resumed).