Optimize GPU vs CPU

If you're new to FAH and need help getting started or you have very basic questions, start here.

Moderators: Site Moderators, FAHC Science Team

Pellicle
Posts: 5
Joined: Sat May 02, 2020 2:47 pm

Optimize GPU vs CPU

Post by Pellicle »

Question
I have a Ryzen &3800 system with a AMD RX 590 GPU.
The system is currently dedicated to folding@home
I noticed when both GPU and CPU are running loads sometines the GPU utilization drops significantly. If I pause the CPU the utilization of the GPU will increase significantly.
The CPU when running typically is showing all cores loaded to 100% and running at full frequency.
I tried an experiment to free up one core and this results in the GPU utilization being increased to near maximum and the CPU.
It appears to me that this is resulting in more units processed for a given time since the GPU output is maximized with a slight degradation to the CPU.
I am assuming the GPU is being throttled by not having enough CPU capacity devoted to feeding it.
Is this a reasonable strategy to maximize. I have not been using folding@home for long and am looking to maximize contribution of my system to the project.
Input would be appreciated.
Neil-B
Posts: 1996
Joined: Sun Mar 22, 2020 5:52 pm
Hardware configuration: 1: 2x Xeon E5-2697v3@2.60GHz, 512GB DDR4 LRDIMM, SSD Raid, Win10 Ent 20H2, Quadro K420 1GB, FAH 7.6.21
2: Xeon E3-1505Mv5@2.80GHz, 32GB DDR4, NVME, Win10 Pro 20H2, Quadro M1000M 2GB, FAH 7.6.21 (actually have two of these)
3: i7-960@3.20GHz, 12GB DDR3, SSD, Win10 Pro 20H2, GTX 750Ti 2GB, GTX 1080Ti 11GB, FAH 7.6.21
Location: UK

Re: Optimize GPU vs CPU

Post by Neil-B »

If you post you log inc the first 200 lines or so with the system configuration someone may be able to more accurately advise … viewtopic.php?f=61&t=26036.

You do need at least one CPU thread to service the GPU.
2x Xeon E5-2697v3, 512GB DDR4 LRDIMM, SSD Raid, W10-Ent, Quadro K420
Xeon E3-1505Mv5, 32GB DDR4, NVME, W10-Pro, Quadro M1000M
i7-960, 12GB DDR3, SSD, W10-Pro, GTX1080Ti
i9-10850K, 64GB DDR4, NVME, W11-Pro, RTX3070

(Green/Bold = Active)
Pellicle
Posts: 5
Joined: Sat May 02, 2020 2:47 pm

Re: Optimize GPU vs CPU

Post by Pellicle »

Below is an excerpt from the log for perusal.
Thank you for your feedback and assistance.

Code: Select all

14:36:19:Removing old file 'configs/config-20200426-150556.xml'
14:36:19:Saving configuration to config.xml
14:36:19:<config>
14:36:19:  <!-- Folding Core -->
14:36:19:  <core-priority v='low'/>
14:36:19:
14:36:19:  <!-- Network -->
14:36:19:  <proxy v=':8080'/>
14:36:19:
14:36:19:  <!-- Slot Control -->
14:36:19:  <pause-on-battery v='false'/>
14:36:19:  <power v='full'/>
14:36:19:
14:36:19:  <!-- User Information -->
14:36:19:  <passkey v='*****'/>
14:36:19:  <user v='Pellicle'/>
14:36:19:
14:36:19:  <!-- Folding Slots -->
14:36:19:  <slot id='0' type='CPU'>
14:36:19:    <cpus v='14'/>
14:36:19:    <paused v='False'/>
14:36:19:  </slot>
14:36:19:  <slot id='1' type='GPU'/>
14:36:19:</config>
14:36:19:FS00:Shutting core down
14:36:20:WU03:FS00:0xa7:WARNING:Console control signal 1 on PID 5096
14:36:20:WU03:FS00:0xa7:Exiting, please wait. . .
14:36:20:WU03:FS00:0xa7:Folding@home Core Shutdown: INTERRUPTED
14:36:21:WU03:FS00:FahCore returned: INTERRUPTED (102 = 0x66)
14:36:21:WU03:FS00:Starting
14:36:21:WARNING:WU03:FS00:Changed SMP threads from 15 to 14 this can cause some work units to fail
14:36:21:WU03:FS00:Running FahCore: "C:\Program Files (x86)\FAHClient/FAHCoreWrapper.exe" C:\Users\therm\AppData\Roaming\FAHClient\cores/cores.foldingathome.org/v7/win/64bit/avx/Core_a7.fah/FahCore_a7.exe -dir 03 -suffix 01 -version 706 -lifeline 8780 -checkpoint 15 -np 14
14:36:21:WU03:FS00:Started FahCore on PID 26296
14:36:21:WU03:FS00:Core PID:14820
14:36:21:WU03:FS00:FahCore 0xa7 started
14:36:21:WU03:FS00:0xa7:*********************** Log Started 2020-05-02T14:36:21Z ***********************
14:36:21:WU03:FS00:0xa7:************************** Gromacs Folding@home Core ***************************
14:36:21:WU03:FS00:0xa7:       Type: 0xa7
14:36:21:WU03:FS00:0xa7:       Core: Gromacs
14:36:21:WU03:FS00:0xa7:       Args: -dir 03 -suffix 01 -version 706 -lifeline 26296 -checkpoint 15 -np
14:36:21:WU03:FS00:0xa7:             14
14:36:21:WU03:FS00:0xa7:************************************ CBang *************************************
14:36:21:WU03:FS00:0xa7:       Date: Oct 26 2019
14:36:21:WU03:FS00:0xa7:       Time: 01:38:25
14:36:21:WU03:FS00:0xa7:   Revision: c46a1a011a24143739ac7218c5a435f66777f62f
14:36:21:WU03:FS00:0xa7:     Branch: master
14:36:21:WU03:FS00:0xa7:   Compiler: Visual C++ 2008
14:36:21:WU03:FS00:0xa7:    Options: /TP /nologo /EHa /wd4297 /wd4103 /Ox /MT
14:36:21:WU03:FS00:0xa7:   Platform: win32 10
14:36:21:WU03:FS00:0xa7:       Bits: 64
14:36:21:WU03:FS00:0xa7:       Mode: Release
14:36:21:WU03:FS00:0xa7:************************************ System ************************************
14:36:21:WU03:FS00:0xa7:        CPU: AMD Ryzen 7 3800X 8-Core Processor
14:36:21:WU03:FS00:0xa7:     CPU ID: AuthenticAMD Family 23 Model 113 Stepping 0
14:36:21:WU03:FS00:0xa7:       CPUs: 16
14:36:21:WU03:FS00:0xa7:     Memory: 31.93GiB
14:36:21:WU03:FS00:0xa7:Free Memory: 21.18GiB
14:36:21:WU03:FS00:0xa7:    Threads: WINDOWS_THREADS
14:36:21:WU03:FS00:0xa7: OS Version: 6.2
14:36:21:WU03:FS00:0xa7:Has Battery: false
14:36:21:WU03:FS00:0xa7: On Battery: false
14:36:21:WU03:FS00:0xa7: UTC Offset: -4
14:36:21:WU03:FS00:0xa7:        PID: 14820
14:36:21:WU03:FS00:0xa7:        CWD: C:\Users\therm\AppData\Roaming\FAHClient\work
14:36:21:WU03:FS00:0xa7:******************************** Build - libFAH ********************************
14:36:21:WU03:FS00:0xa7:    Version: 0.0.18
14:36:21:WU03:FS00:0xa7:     Author: Joseph Coffland <joseph@cauldrondevelopment.com>
14:36:21:WU03:FS00:0xa7:  Copyright: 2019 foldingathome.org
14:36:21:WU03:FS00:0xa7:   Homepage: https://foldingathome.org/
14:36:21:WU03:FS00:0xa7:       Date: Oct 26 2019
14:36:21:WU03:FS00:0xa7:       Time: 01:52:30
14:36:21:WU03:FS00:0xa7:   Revision: c1e3513b1bc0c16013668f2173ee969e5995b38e
14:36:21:WU03:FS00:0xa7:     Branch: master
14:36:21:WU03:FS00:0xa7:   Compiler: Visual C++ 2008
14:36:21:WU03:FS00:0xa7:    Options: /TP /nologo /EHa /wd4297 /wd4103 /Ox /MT
14:36:21:WU03:FS00:0xa7:   Platform: win32 10
14:36:21:WU03:FS00:0xa7:       Bits: 64
14:36:21:WU03:FS00:0xa7:       Mode: Release
14:36:21:WU03:FS00:0xa7:************************************ Build *************************************
14:36:21:WU03:FS00:0xa7:       SIMD: avx_256
14:36:21:WU03:FS00:0xa7:********************************************************************************
14:36:21:WU03:FS00:0xa7:Project: 16418 (Run 0, Clone 1541, Gen 134)
14:36:21:WU03:FS00:0xa7:Unit: 0x0000009396880e6e5e82558166bf3de8
14:36:21:WU03:FS00:0xa7:Digital signatures verified
14:36:21:WU03:FS00:0xa7:Reducing thread count from 14 to 13 to avoid domain decomposition with large prime factor 7
14:36:21:WU03:FS00:0xa7:Reducing thread count from 13 to 12 to avoid domain decomposition by a prime number > 3
14:36:21:WU03:FS00:0xa7:Calling: mdrun -s frame134.tpr -o frame134.trr -x frame134.xtc -cpi state.cpt -cpt 15 -nt 12
14:36:21:WU03:FS00:0xa7:Steps: first=33500000 total=250000
14:36:22:WU03:FS00:0xa7:Completed 21422 out of 250000 steps (8%)
14:36:36:WU03:FS00:0xa7:Completed 22500 out of 250000 steps (9%)
14:36:54:Removing old file 'configs/config-20200426-150657.xml'
14:36:54:Saving configuration to config.xml
14:36:54:<config>
14:36:54:  <!-- Folding Core -->
14:36:54:  <core-priority v='low'/>
14:36:54:
14:36:54:  <!-- Network -->
14:36:54:  <proxy v=':8080'/>
14:36:54:
14:36:54:  <!-- Slot Control -->
14:36:54:  <pause-on-battery v='false'/>
14:36:54:  <power v='full'/>
14:36:54:
14:36:54:  <!-- User Information -->
14:36:54:  <passkey v='*****'/>
14:36:54:  <user v='Pellicle'/>
14:36:54:
14:36:54:  <!-- Folding Slots -->
14:36:54:  <slot id='0' type='CPU'>
14:36:54:    <cpus v='14'/>
14:36:54:    <paused v='False'/>
14:36:54:  </slot>
14:36:54:  <slot id='1' type='GPU'/>
14:36:54:</config>
14:37:08:WU03:FS00:0xa7:Completed 25000 out of 250000 steps (10%)
14:37:40:WU03:FS00:0xa7:Completed 27500 out of 250000 steps (11%)
14:38:13:WU03:FS00:0xa7:Completed 30000 out of 250000 steps (12%)
14:38:46:WU03:FS00:0xa7:Completed 32500 out of 250000 steps (13%)
14:39:18:WU03:FS00:0xa7:Completed 35000 out of 250000 steps (14%)
14:39:51:WU03:FS00:0xa7:Completed 37500 out of 250000 steps (15%)
14:40:08:WU00:FS01:0x22:Completed 1650000 out of 5000000 steps (33%)
14:40:23:WU03:FS00:0xa7:Completed 40000 out of 250000 steps (16%)
14:40:56:WU03:FS00:0xa7:Completed 42500 out of 250000 steps (17%)
14:41:28:WU03:FS00:0xa7:Completed 45000 out of 250000 steps (18%)
14:42:00:WU03:FS00:0xa7:Completed 47500 out of 250000 steps (19%)
14:42:33:WU03:FS00:0xa7:Completed 50000 out of 250000 steps (20%)
14:43:05:WU03:FS00:0xa7:Completed 52500 out of 250000 steps (21%)
14:43:37:WU03:FS00:0xa7:Completed 55000 out of 250000 steps (22%)
14:44:01:WU00:FS01:0x22:Completed 1700000 out of 5000000 steps (34%)
14:44:10:WU03:FS00:0xa7:Completed 57500 out of 250000 steps (23%)
14:44:42:WU03:FS00:0xa7:Completed 60000 out of 250000 steps (24%)
14:45:15:WU03:FS00:0xa7:Completed 62500 out of 250000 steps (25%)
14:45:48:WU03:FS00:0xa7:Completed 65000 out of 250000 steps (26%)
14:46:20:WU03:FS00:0xa7:Completed 67500 out of 250000 steps (27%)
14:46:53:WU03:FS00:0xa7:Completed 70000 out of 250000 steps (28%)
14:47:25:WU03:FS00:0xa7:Completed 72500 out of 250000 steps (29%)
14:47:54:WU00:FS01:0x22:Completed 1750000 out of 5000000 steps (35%)
14:47:58:WU03:FS00:0xa7:Completed 75000 out of 250000 steps (30%)
14:48:30:WU03:FS00:0xa7:Completed 77500 out of 250000 steps (31%)
14:49:02:WU03:FS00:0xa7:Completed 80000 out of 250000 steps (32%)
14:49:35:WU03:FS00:0xa7:Completed 82500 out of 250000 steps (33%)
14:50:07:WU03:FS00:0xa7:Completed 85000 out of 250000 steps (34%)
14:50:40:WU03:FS00:0xa7:Completed 87500 out of 250000 steps (35%)
14:51:12:WU03:FS00:0xa7:Completed 90000 out of 250000 steps (36%)
14:51:45:WU03:FS00:0xa7:Completed 92500 out of 250000 steps (37%)
14:51:47:WU00:FS01:0x22:Completed 1800000 out of 5000000 steps (36%)
14:52:18:WU03:FS00:0xa7:Completed 95000 out of 250000 steps (38%)
14:52:50:WU03:FS00:0xa7:Completed 97500 out of 250000 steps (39%)
14:53:22:WU03:FS00:0xa7:Completed 100000 out of 250000 steps (40%)
14:53:55:WU03:FS00:0xa7:Completed 102500 out of 250000 steps (41%)
14:54:27:WU03:FS00:0xa7:Completed 105000 out of 250000 steps (42%)
14:55:00:WU03:FS00:0xa7:Completed 107500 out of 250000 steps (43%)
14:55:32:WU03:FS00:0xa7:Completed 110000 out of 250000 steps (44%)
14:55:39:WU00:FS01:0x22:Completed 1850000 out of 5000000 steps (37%)
14:56:05:WU03:FS00:0xa7:Completed 112500 out of 250000 steps (45%)
14:56:38:WU03:FS00:0xa7:Completed 115000 out of 250000 steps (46%)
14:57:10:WU03:FS00:0xa7:Completed 117500 out of 250000 steps (47%)
14:57:42:WU03:FS00:0xa7:Completed 120000 out of 250000 steps (48%)
14:58:15:WU03:FS00:0xa7:Completed 122500 out of 250000 steps (49%)
14:58:47:WU03:FS00:0xa7:Completed 125000 out of 250000 steps (50%)
14:59:20:WU03:FS00:0xa7:Completed 127500 out of 250000 steps (51%)
14:59:33:WU00:FS01:0x22:Completed 1900000 out of 5000000 steps (38%)
14:59:52:WU03:FS00:0xa7:Completed 130000 out of 250000 steps (52%)
15:00:25:WU03:FS00:0xa7:Completed 132500 out of 250000 steps (53%)
15:00:57:WU03:FS00:0xa7:Completed 135000 out of 250000 steps (54%)
15:01:29:WU03:FS00:0xa7:Completed 137500 out of 250000 steps (55%)
15:02:02:WU03:FS00:0xa7:Completed 140000 out of 250000 steps (56%)
15:02:34:WU03:FS00:0xa7:Completed 142500 out of 250000 steps (57%)
15:03:07:WU03:FS00:0xa7:Completed 145000 out of 250000 steps (58%)
15:03:23:WU00:FS01:0x22:Completed 1950000 out of 5000000 steps (39%)
15:03:39:WU03:FS00:0xa7:Completed 147500 out of 250000 steps (59%)
15:04:12:WU03:FS00:0xa7:Completed 150000 out of 250000 steps (60%)
15:04:44:WU03:FS00:0xa7:Completed 152500 out of 250000 steps (61%)
15:05:16:WU03:FS00:0xa7:Completed 155000 out of 250000 steps (62%)
15:05:49:WU03:FS00:0xa7:Completed 157500 out of 250000 steps (63%)
15:06:21:WU03:FS00:0xa7:Completed 160000 out of 250000 steps (64%)
15:06:54:WU03:FS00:0xa7:Completed 162500 out of 250000 steps (65%)
15:07:11:WU00:FS01:0x22:Completed 2000000 out of 5000000 steps (40%)
15:07:26:WU03:FS00:0xa7:Completed 165000 out of 250000 steps (66%)
15:07:59:WU03:FS00:0xa7:Completed 167500 out of 250000 steps (67%)
15:08:31:WU03:FS00:0xa7:Completed 170000 out of 250000 steps (68%)
15:09:04:WU03:FS00:0xa7:Completed 172500 out of 250000 steps (69%)
15:09:36:WU03:FS00:0xa7:Completed 175000 out of 250000 steps (70%)
15:10:08:WU03:FS00:0xa7:Completed 177500 out of 250000 steps (71%)
15:10:40:WU03:FS00:0xa7:Completed 180000 out of 250000 steps (72%)
15:10:59:WU00:FS01:0x22:Completed 2050000 out of 5000000 steps (41%)
15:11:13:WU03:FS00:0xa7:Completed 182500 out of 250000 steps (73%)
15:11:45:WU03:FS00:0xa7:Completed 185000 out of 250000 steps (74%)
15:12:18:WU03:FS00:0xa7:Completed 187500 out of 250000 steps (75%)
15:12:50:WU03:FS00:0xa7:Completed 190000 out of 250000 steps (76%)
15:13:23:WU03:FS00:0xa7:Completed 192500 out of 250000 steps (77%)
15:13:55:WU03:FS00:0xa7:Completed 195000 out of 250000 steps (78%)
15:14:27:WU03:FS00:0xa7:Completed 197500 out of 250000 steps (79%)
15:14:48:WU00:FS01:0x22:Completed 2100000 out of 5000000 steps (42%)
15:15:00:WU03:FS00:0xa7:Completed 200000 out of 250000 steps (80%)
15:15:32:WU03:FS00:0xa7:Completed 202500 out of 250000 steps (81%)
15:16:04:WU03:FS00:0xa7:Completed 205000 out of 250000 steps (82%)
15:16:37:WU03:FS00:0xa7:Completed 207500 out of 250000 steps (83%)
15:17:09:WU03:FS00:0xa7:Completed 210000 out of 250000 steps (84%)
15:17:41:WU03:FS00:0xa7:Completed 212500 out of 250000 steps (85%)
15:18:14:WU03:FS00:0xa7:Completed 215000 out of 250000 steps (86%)
15:18:36:WU00:FS01:0x22:Completed 2150000 out of 5000000 steps (43%)
15:18:46:WU03:FS00:0xa7:Completed 217500 out of 250000 steps (87%)
15:19:18:WU03:FS00:0xa7:Completed 220000 out of 250000 steps (88%)
15:19:51:WU03:FS00:0xa7:Completed 222500 out of 250000 steps (89%)
15:20:23:WU03:FS00:0xa7:Completed 225000 out of 250000 steps (90%)
15:20:55:WU03:FS00:0xa7:Completed 227500 out of 250000 steps (91%)
15:21:28:WU03:FS00:0xa7:Completed 230000 out of 250000 steps (92%)
15:22:00:WU03:FS00:0xa7:Completed 232500 out of 250000 steps (93%)
15:22:24:WU00:FS01:0x22:Completed 2200000 out of 5000000 steps (44%)
15:22:33:WU03:FS00:0xa7:Completed 235000 out of 250000 steps (94%)
15:23:05:WU03:FS00:0xa7:Completed 237500 out of 250000 steps (95%)
15:23:37:WU03:FS00:0xa7:Completed 240000 out of 250000 steps (96%)
15:24:10:WU03:FS00:0xa7:Completed 242500 out of 250000 steps (97%)
15:24:42:WU03:FS00:0xa7:Completed 245000 out of 250000 steps (98%)
15:25:14:WU03:FS00:0xa7:Completed 247500 out of 250000 steps (99%)
15:25:15:WU02:FS00:Connecting to 65.254.110.245:80
15:25:15:WU02:FS00:Assigned to work server 128.252.203.2
15:25:15:WU02:FS00:Requesting new work unit for slot 00: RUNNING cpu:14 from 128.252.203.2
15:25:15:WU02:FS00:Connecting to 128.252.203.2:8080
15:25:16:WU02:FS00:Downloading 25.12MiB
15:25:21:WU02:FS00:Download complete
15:25:21:WU02:FS00:Received Unit: id:02 state:DOWNLOAD error:NO_ERROR project:14591 run:0 clone:305 gen:34 core:0xa7 unit:0x0000002780fccb025e8cf8c8e4889a7c
15:25:46:WU03:FS00:0xa7:Completed 250000 out of 250000 steps (100%)
15:25:47:WU03:FS00:0xa7:Saving result file ..\logfile_01.txt
15:25:47:WU03:FS00:0xa7:Saving result file frame134.trr
15:25:47:WU03:FS00:0xa7:Saving result file frame134.xtc
15:25:47:WU03:FS00:0xa7:Saving result file md.log
15:25:47:WU03:FS00:0xa7:Saving result file science.log
15:25:47:WU03:FS00:0xa7:Folding@home Core Shutdown: FINISHED_UNIT
15:25:48:WU03:FS00:FahCore returned: FINISHED_UNIT (100 = 0x64)
15:25:48:WU03:FS00:Sending unit results: id:03 state:SEND error:NO_ERROR project:16418 run:0 clone:1541 gen:134 core:0xa7 unit:0x0000009396880e6e5e82558166bf3de8
15:25:48:WU03:FS00:Uploading 2.74MiB to 150.136.14.110
15:25:48:WU02:FS00:Starting
15:25:48:WU03:FS00:Connecting to 150.136.14.110:8080
15:25:48:WU02:FS00:Running FahCore: "C:\Program Files (x86)\FAHClient/FAHCoreWrapper.exe" C:\Users\therm\AppData\Roaming\FAHClient\cores/cores.foldingathome.org/v7/win/64bit/avx/Core_a7.fah/FahCore_a7.exe -dir 02 -suffix 01 -version 706 -lifeline 8780 -checkpoint 15 -np 14
15:25:48:WU02:FS00:Started FahCore on PID 4164
15:25:48:WU02:FS00:Core PID:16008
15:25:48:WU02:FS00:FahCore 0xa7 started
15:25:48:WU02:FS00:0xa7:*********************** Log Started 2020-05-02T15:25:48Z ***********************
15:25:48:WU02:FS00:0xa7:************************** Gromacs Folding@home Core ***************************
15:25:48:WU02:FS00:0xa7:       Type: 0xa7
15:25:48:WU02:FS00:0xa7:       Core: Gromacs
15:25:48:WU02:FS00:0xa7:       Args: -dir 02 -suffix 01 -version 706 -lifeline 4164 -checkpoint 15 -np
15:25:48:WU02:FS00:0xa7:             14
15:25:48:WU02:FS00:0xa7:************************************ CBang *************************************
15:25:48:WU02:FS00:0xa7:       Date: Oct 26 2019
15:25:48:WU02:FS00:0xa7:       Time: 01:38:25
15:25:48:WU02:FS00:0xa7:   Revision: c46a1a011a24143739ac7218c5a435f66777f62f
15:25:48:WU02:FS00:0xa7:     Branch: master
15:25:48:WU02:FS00:0xa7:   Compiler: Visual C++ 2008
15:25:48:WU02:FS00:0xa7:    Options: /TP /nologo /EHa /wd4297 /wd4103 /Ox /MT
15:25:48:WU02:FS00:0xa7:   Platform: win32 10
15:25:48:WU02:FS00:0xa7:       Bits: 64
15:25:48:WU02:FS00:0xa7:       Mode: Release
15:25:48:WU02:FS00:0xa7:************************************ System ************************************
15:25:48:WU02:FS00:0xa7:        CPU: AMD Ryzen 7 3800X 8-Core Processor
15:25:48:WU02:FS00:0xa7:     CPU ID: AuthenticAMD Family 23 Model 113 Stepping 0
15:25:48:WU02:FS00:0xa7:       CPUs: 16
15:25:48:WU02:FS00:0xa7:     Memory: 31.93GiB
15:25:48:WU02:FS00:0xa7:Free Memory: 21.14GiB
15:25:48:WU02:FS00:0xa7:    Threads: WINDOWS_THREADS
15:25:48:WU02:FS00:0xa7: OS Version: 6.2
15:25:48:WU02:FS00:0xa7:Has Battery: false
15:25:48:WU02:FS00:0xa7: On Battery: false
15:25:48:WU02:FS00:0xa7: UTC Offset: -4
15:25:48:WU02:FS00:0xa7:        PID: 16008
15:25:48:WU02:FS00:0xa7:        CWD: C:\Users\therm\AppData\Roaming\FAHClient\work
15:25:48:WU02:FS00:0xa7:******************************** Build - libFAH ********************************
15:25:48:WU02:FS00:0xa7:    Version: 0.0.18
15:25:48:WU02:FS00:0xa7:     Author: Joseph Coffland <joseph@cauldrondevelopment.com>
15:25:48:WU02:FS00:0xa7:  Copyright: 2019 foldingathome.org
15:25:48:WU02:FS00:0xa7:   Homepage: https://foldingathome.org/
15:25:48:WU02:FS00:0xa7:       Date: Oct 26 2019
15:25:48:WU02:FS00:0xa7:       Time: 01:52:30
15:25:49:WU02:FS00:0xa7:   Revision: c1e3513b1bc0c16013668f2173ee969e5995b38e
15:25:49:WU02:FS00:0xa7:     Branch: master
15:25:49:WU02:FS00:0xa7:   Compiler: Visual C++ 2008
15:25:49:WU02:FS00:0xa7:    Options: /TP /nologo /EHa /wd4297 /wd4103 /Ox /MT
15:25:49:WU02:FS00:0xa7:   Platform: win32 10
15:25:49:WU02:FS00:0xa7:       Bits: 64
15:25:49:WU02:FS00:0xa7:       Mode: Release
15:25:49:WU02:FS00:0xa7:************************************ Build *************************************
15:25:49:WU02:FS00:0xa7:       SIMD: avx_256
15:25:49:WU02:FS00:0xa7:********************************************************************************
15:25:49:WU02:FS00:0xa7:Project: 14591 (Run 0, Clone 305, Gen 34)
15:25:49:WU02:FS00:0xa7:Unit: 0x0000002780fccb025e8cf8c8e4889a7c
15:25:49:WU02:FS00:0xa7:Reading tar file core.xml
15:25:49:WU02:FS00:0xa7:Reading tar file frame34.tpr
15:25:49:WU02:FS00:0xa7:Digital signatures verified
15:25:49:WU02:FS00:0xa7:Reducing thread count from 14 to 13 to avoid domain decomposition with large prime factor 7
15:25:49:WU02:FS00:0xa7:Reducing thread count from 13 to 12 to avoid domain decomposition by a prime number > 3
15:25:49:WU02:FS00:0xa7:Calling: mdrun -s frame34.tpr -o frame34.trr -x frame34.xtc -cpt 15 -nt 12
15:25:49:WU02:FS00:0xa7:Steps: first=2125000 total=62500
15:25:50:WU03:FS00:Upload complete
15:25:50:WU03:FS00:Server responded WORK_ACK (400)
15:25:50:WU03:FS00:Final credit estimate, 6023.00 points
15:25:50:WU03:FS00:Cleaning up
15:25:53:WU02:FS00:0xa7:Completed 1 out of 62500 steps (0%)
15:26:12:WU00:FS01:0x22:Completed 2250000 out of 5000000 steps (45%)
15:26:51:WU02:FS00:0xa7:Completed 625 out of 62500 steps (1%)
15:27:50:WU02:FS00:0xa7:Completed 1250 out of 62500 steps (2%)
15:28:49:WU02:FS00:0xa7:Completed 1875 out of 62500 steps (3%)
15:29:47:WU02:FS00:0xa7:Completed 2500 out of 62500 steps (4%)
15:30:12:WU00:FS01:0x22:Completed 2300000 out of 5000000 steps (46%)
15:30:46:WU02:FS00:0xa7:Completed 3125 out of 62500 steps (5%)
15:31:45:WU02:FS00:0xa7:Completed 3750 out of 62500 steps (6%)
15:32:44:WU02:FS00:0xa7:Completed 4375 out of 62500 steps (7%)
15:33:42:WU02:FS00:0xa7:Completed 5000 out of 62500 steps (8%)
15:34:12:WU00:FS01:0x22:Completed 2350000 out of 5000000 steps (47%)
15:34:41:WU02:FS00:0xa7:Completed 5625 out of 62500 steps (9%)
15:35:40:WU02:FS00:0xa7:Completed 6250 out of 62500 steps (10%)
15:36:29:WU01:FS01:Sending unit results: id:01 state:SEND error:NO_ERROR project:14416 run:0 clone:1053 gen:18 core:0x22 unit:0x0000001f0d5262775e84c8a79aeea156
15:36:29:WU01:FS01:Uploading 231.61MiB to 13.82.98.119
15:36:29:WU01:FS01:Connecting to 13.82.98.119:8080
15:36:39:WU02:FS00:0xa7:Completed 6875 out of 62500 steps (11%)
15:37:02:WU01:FS01:Upload 0.08%
15:37:02:WARNING:WU01:FS01:Exception: Failed to send results to work server: Transfer failed
15:37:02:WU01:FS01:Trying to send results to collection server
15:37:02:WU01:FS01:Uploading 231.61MiB to 52.224.109.74
15:37:02:WU01:FS01:Connecting to 52.224.109.74:8080
15:37:02:ERROR:WU01:FS01:Exception: Transfer failed
15:37:37:WU02:FS00:0xa7:Completed 7500 out of 62500 steps (12%)
JimboPalmer
Posts: 2522
Joined: Mon Feb 16, 2009 4:12 am
Location: Greenwood MS USA

Re: Optimize GPU vs CPU

Post by JimboPalmer »

Welcome to Folding@Home!

You did not include the first 200 lines so some advice is fragmentary.

Set your number of threads to 12 (F@H calls them CPUs)

You have 16 threads but F@H reserves one for feeding the GPU.
That leaves you 15, but F@H hates large Primes and their multiples, it is safest to only use numbers with 2 and 3 as factors.

15 is 3 * 5, so F@H dislikes it
14 is 2* 7 so F@H dislikes it
13 is Prime, so F@H dislikes it
12 is 2*2*3 so it is ideal

In the taskbar to the lower right of the screen, you should see a F@H molecule icon, click it (you may need to click an Up Arrow to see it ^)

The second item in this menu is Advanced Control, click it

On this screen to the left is a Configure button, click it

Now you get a screen with a Slots tab, click it

On this white field should be a cpu item, click it and then click edit

By default F@H set the number of CPUs to -1 meaning let the software decide.

Type the number you want, and click save.

Then capture a new log and show us that.

viewtopic.php?f=24&t=26036

I hope this helps!
Tsar of all the Rushers
I tried to remain childlike, all I achieved was childish.
A friend to those who want no friends
HugoNotte
Posts: 66
Joined: Tue Apr 07, 2020 7:09 pm

Re: Optimize GPU vs CPU

Post by HugoNotte »

FAH does want 1 thread (logical core) reserved for the GPU. If I'm not mistaking, your processor got 8 physical cores and 16 logical cores / threads. Therefore the standard setting in the CPU slot settings would be -1.
Otherwise you can define 15 "cores" in the cpu slot settings, to ensure the 16th is available for the GPU. If I read your log correctly, you have defined 14 CPU cores for FAH. You might run into trouble with that, since FAH doesn't like multiples of larger primes, such as 7 and onwards. Best is to stick to core counts which are multiples of 2 and / or 3. 5 is apparently still an acceptable factor, too. You might be better off with 15 or 12 cores or make 2 CPU slots with 9 and 6 cores respectively.
Pellicle
Posts: 5
Joined: Sat May 02, 2020 2:47 pm

Re: Optimize GPU vs CPU

Post by Pellicle »

Thanks to both JimboPalmer and HugoNotte,
BY trial and error I saw that when I set to 14 threads the CPU utilization decreased. Cores 5 through 8 showed decreased utilization running significantly lower in frequency.
But the GPU was running at a higher utilization than the default settings.
I increased threads allocated to 15 and the CPU utilization went back to full utilization with all cores running at max with corresponding power and heat increases. However I am seeing decreased GPU utilization again. So I have set CPU core allocation to 13. GPU now is running at higher utilization >90% as opposed to around 70%.
Assuming my system readings are a reflection of output rate from F@H I am thinking this might be a good compromise. If my assessment is incorrect please comment.
Attached is full log after change

Code: Select all

17:56:01:Removing old file 'configs/config-20200426-182714.xml'
17:56:01:Saving configuration to config.xml
17:56:01:<config>
17:56:01:  <!-- Folding Core -->
17:56:01:  <core-priority v='low'/>
17:56:01:
17:56:01:  <!-- Network -->
17:56:01:  <proxy v=':8080'/>
17:56:01:
17:56:01:  <!-- Slot Control -->
17:56:01:  <pause-on-battery v='false'/>
17:56:01:  <power v='full'/>
17:56:01:
17:56:01:  <!-- User Information -->
17:56:01:  <passkey v='*****'/>
17:56:01:  <user v='Pellicle'/>
17:56:01:
17:56:01:  <!-- Folding Slots -->
17:56:01:  <slot id='0' type='CPU'>
17:56:01:    <cpus v='13'/>
17:56:01:    <paused v='False'/>
17:56:01:  </slot>
17:56:01:  <slot id='1' type='GPU'/>
17:56:01:</config>
17:56:01:FS00:Shutting core down
17:56:02:WU03:FS00:0xa7:WARNING:Console control signal 1 on PID 16932
17:56:02:WU03:FS00:0xa7:Exiting, please wait. . .
17:56:02:WU03:FS00:0xa7:Folding@home Core Shutdown: INTERRUPTED
17:56:03:WU03:FS00:FahCore returned: INTERRUPTED (102 = 0x66)
17:56:03:WU03:FS00:Starting
17:56:03:WARNING:WU03:FS00:Changed SMP threads from 15 to 13 this can cause some work units to fail
17:56:03:WU03:FS00:Running FahCore: "C:\Program Files (x86)\FAHClient/FAHCoreWrapper.exe" C:\Users\therm\AppData\Roaming\FAHClient\cores/cores.foldingathome.org/v7/win/64bit/avx/Core_a7.fah/FahCore_a7.exe -dir 03 -suffix 01 -version 706 -lifeline 8780 -checkpoint 15 -np 13
17:56:03:WU03:FS00:Started FahCore on PID 24740
17:56:03:WU03:FS00:Core PID:21080
17:56:03:WU03:FS00:FahCore 0xa7 started
17:56:03:WU03:FS00:0xa7:*********************** Log Started 2020-05-02T17:56:03Z ***********************
17:56:03:WU03:FS00:0xa7:************************** Gromacs Folding@home Core ***************************
17:56:03:WU03:FS00:0xa7:       Type: 0xa7
17:56:03:WU03:FS00:0xa7:       Core: Gromacs
17:56:03:WU03:FS00:0xa7:       Args: -dir 03 -suffix 01 -version 706 -lifeline 24740 -checkpoint 15 -np
17:56:03:WU03:FS00:0xa7:             13
17:56:03:WU03:FS00:0xa7:************************************ CBang *************************************
17:56:03:WU03:FS00:0xa7:       Date: Oct 26 2019
17:56:03:WU03:FS00:0xa7:       Time: 01:38:25
17:56:03:WU03:FS00:0xa7:   Revision: c46a1a011a24143739ac7218c5a435f66777f62f
17:56:03:WU03:FS00:0xa7:     Branch: master
17:56:03:WU03:FS00:0xa7:   Compiler: Visual C++ 2008
17:56:03:WU03:FS00:0xa7:    Options: /TP /nologo /EHa /wd4297 /wd4103 /Ox /MT
17:56:03:WU03:FS00:0xa7:   Platform: win32 10
17:56:03:WU03:FS00:0xa7:       Bits: 64
17:56:03:WU03:FS00:0xa7:       Mode: Release
17:56:03:WU03:FS00:0xa7:************************************ System ************************************
17:56:03:WU03:FS00:0xa7:        CPU: AMD Ryzen 7 3800X 8-Core Processor
17:56:03:WU03:FS00:0xa7:     CPU ID: AuthenticAMD Family 23 Model 113 Stepping 0
17:56:03:WU03:FS00:0xa7:       CPUs: 16
17:56:03:WU03:FS00:0xa7:     Memory: 31.93GiB
17:56:03:WU03:FS00:0xa7:Free Memory: 21.14GiB
17:56:03:WU03:FS00:0xa7:    Threads: WINDOWS_THREADS
17:56:03:WU03:FS00:0xa7: OS Version: 6.2
17:56:03:WU03:FS00:0xa7:Has Battery: false
17:56:03:WU03:FS00:0xa7: On Battery: false
17:56:03:WU03:FS00:0xa7: UTC Offset: -4
17:56:03:WU03:FS00:0xa7:        PID: 21080
17:56:03:WU03:FS00:0xa7:        CWD: C:\Users\therm\AppData\Roaming\FAHClient\work
17:56:03:WU03:FS00:0xa7:******************************** Build - libFAH ********************************
17:56:03:WU03:FS00:0xa7:    Version: 0.0.18
17:56:03:WU03:FS00:0xa7:     Author: Joseph Coffland <joseph@cauldrondevelopment.com>
17:56:03:WU03:FS00:0xa7:  Copyright: 2019 foldingathome.org
17:56:03:WU03:FS00:0xa7:   Homepage: https://foldingathome.org/
17:56:03:WU03:FS00:0xa7:       Date: Oct 26 2019
17:56:03:WU03:FS00:0xa7:       Time: 01:52:30
17:56:03:WU03:FS00:0xa7:   Revision: c1e3513b1bc0c16013668f2173ee969e5995b38e
17:56:03:WU03:FS00:0xa7:     Branch: master
17:56:03:WU03:FS00:0xa7:   Compiler: Visual C++ 2008
17:56:03:WU03:FS00:0xa7:    Options: /TP /nologo /EHa /wd4297 /wd4103 /Ox /MT
17:56:03:WU03:FS00:0xa7:   Platform: win32 10
17:56:03:WU03:FS00:0xa7:       Bits: 64
17:56:03:WU03:FS00:0xa7:       Mode: Release
17:56:03:WU03:FS00:0xa7:************************************ Build *************************************
17:56:03:WU03:FS00:0xa7:       SIMD: avx_256
17:56:03:WU03:FS00:0xa7:********************************************************************************
17:56:03:WU03:FS00:0xa7:Project: 14621 (Run 166, Clone 0, Gen 56)
17:56:03:WU03:FS00:0xa7:Unit: 0x000000419bf7a4d55e89f1ba97b4800f
17:56:03:WU03:FS00:0xa7:Digital signatures verified
17:56:03:WU03:FS00:0xa7:Reducing thread count from 13 to 12 to avoid domain decomposition by a prime number > 3
17:56:03:WU03:FS00:0xa7:Calling: mdrun -s frame56.tpr -o frame56.trr -cpi state.cpt -cpt 15 -nt 12
17:56:03:WU03:FS00:0xa7:Steps: first=0 total=250000
17:56:04:WU03:FS00:0xa7:Completed 210877 out of 250000 steps (84%)
17:56:30:WU03:FS00:0xa7:Completed 212500 out of 250000 steps (85%)
17:56:34:Removing old file 'configs/config-20200426-182815.xml'
17:56:34:Saving configuration to config.xml
17:56:34:<config>
17:56:34:  <!-- Folding Core -->
17:56:34:  <core-priority v='low'/>
17:56:34:
17:56:34:  <!-- Network -->
17:56:34:  <proxy v=':8080'/>
17:56:34:
17:56:34:  <!-- Slot Control -->
17:56:34:  <pause-on-battery v='false'/>
17:56:34:  <power v='full'/>
17:56:34:
17:56:34:  <!-- User Information -->
17:56:34:  <passkey v='*****'/>
17:56:34:  <user v='Pellicle'/>
17:56:34:
17:56:34:  <!-- Folding Slots -->
17:56:34:  <slot id='0' type='CPU'>
17:56:34:    <cpus v='13'/>
17:56:34:    <paused v='False'/>
17:56:34:  </slot>
17:56:34:  <slot id='1' type='GPU'/>
17:56:34:</config>
17:57:09:WU03:FS00:0xa7:Completed 215000 out of 250000 steps (86%)
17:57:49:WU03:FS00:0xa7:Completed 217500 out of 250000 steps (87%)
17:58:29:WU03:FS00:0xa7:Completed 220000 out of 250000 steps (88%)
17:58:58:WU00:FS01:0x22:Completed 4100000 out of 5000000 steps (82%)
17:59:08:WU03:FS00:0xa7:Completed 222500 out of 250000 steps (89%)
17:59:48:WU03:FS00:0xa7:Completed 225000 out of 250000 steps (90%)
18:00:27:WU03:FS00:0xa7:Completed 227500 out of 250000 steps (91%)
18:01:07:WU03:FS00:0xa7:Completed 230000 out of 250000 steps (92%)
18:01:47:WU03:FS00:0xa7:Completed 232500 out of 250000 steps (93%)
18:02:26:WU03:FS00:0xa7:Completed 235000 out of 250000 steps (94%)
18:02:49:WU00:FS01:0x22:Completed 4150000 out of 5000000 steps (83%)
18:03:06:WU03:FS00:0xa7:Completed 237500 out of 250000 steps (95%)
18:03:45:WU03:FS00:0xa7:Completed 240000 out of 250000 steps (96%)
18:04:25:WU03:FS00:0xa7:Completed 242500 out of 250000 steps (97%)
18:05:05:WU03:FS00:0xa7:Completed 245000 out of 250000 steps (98%)
18:05:45:WU03:FS00:0xa7:Completed 247500 out of 250000 steps (99%)
18:05:45:WU02:FS00:Connecting to 65.254.110.245:80
18:05:45:WU02:FS00:Assigned to work server 168.245.198.125
18:05:45:WU02:FS00:Requesting new work unit for slot 00: RUNNING cpu:13 from 168.245.198.125
18:05:45:WU02:FS00:Connecting to 168.245.198.125:8080
18:05:46:WU02:FS00:Downloading 13.97MiB
18:05:51:WU02:FS00:Download complete
18:05:51:WU02:FS00:Received Unit: id:02 state:DOWNLOAD error:NO_ERROR project:16408 run:0 clone:1373 gen:57 core:0xa7 unit:0x00000041a8f5c67d5e801d790b21ab48
18:06:25:WU03:FS00:0xa7:Completed 250000 out of 250000 steps (100%)
18:06:26:WU03:FS00:0xa7:Saving result file ..\logfile_01.txt
18:06:26:WU03:FS00:0xa7:Saving result file dhdl.xvg
18:06:26:WU03:FS00:0xa7:Saving result file frame56.trr
18:06:26:WU03:FS00:0xa7:Saving result file md.log
18:06:26:WU03:FS00:0xa7:Saving result file pullf.xvg
18:06:26:WU03:FS00:0xa7:Saving result file pullx.xvg
18:06:26:WU03:FS00:0xa7:Saving result file science.log
18:06:26:WU03:FS00:0xa7:Saving result file traj_comp.xtc
18:06:26:WU03:FS00:0xa7:Folding@home Core Shutdown: FINISHED_UNIT
18:06:26:WU03:FS00:FahCore returned: FINISHED_UNIT (100 = 0x64)
18:06:26:WU03:FS00:Sending unit results: id:03 state:SEND error:NO_ERROR project:14621 run:166 clone:0 gen:56 core:0xa7 unit:0x000000419bf7a4d55e89f1ba97b4800f
18:06:26:WU03:FS00:Uploading 6.47MiB to 155.247.164.213
18:06:26:WU03:FS00:Connecting to 155.247.164.213:8080
18:06:26:WU02:FS00:Starting
18:06:26:WU02:FS00:Running FahCore: "C:\Program Files (x86)\FAHClient/FAHCoreWrapper.exe" C:\Users\therm\AppData\Roaming\FAHClient\cores/cores.foldingathome.org/v7/win/64bit/avx/Core_a7.fah/FahCore_a7.exe -dir 02 -suffix 01 -version 706 -lifeline 8780 -checkpoint 15 -np 13
18:06:26:WU02:FS00:Started FahCore on PID 21616
18:06:26:WU02:FS00:Core PID:21064
18:06:26:WU02:FS00:FahCore 0xa7 started
18:06:26:WU02:FS00:0xa7:*********************** Log Started 2020-05-02T18:06:26Z ***********************
18:06:26:WU02:FS00:0xa7:************************** Gromacs Folding@home Core ***************************
18:06:26:WU02:FS00:0xa7:       Type: 0xa7
18:06:26:WU02:FS00:0xa7:       Core: Gromacs
18:06:26:WU02:FS00:0xa7:       Args: -dir 02 -suffix 01 -version 706 -lifeline 21616 -checkpoint 15 -np
18:06:26:WU02:FS00:0xa7:             13
18:06:26:WU02:FS00:0xa7:************************************ CBang *************************************
18:06:26:WU02:FS00:0xa7:       Date: Oct 26 2019
18:06:26:WU02:FS00:0xa7:       Time: 01:38:25
18:06:26:WU02:FS00:0xa7:   Revision: c46a1a011a24143739ac7218c5a435f66777f62f
18:06:26:WU02:FS00:0xa7:     Branch: master
18:06:26:WU02:FS00:0xa7:   Compiler: Visual C++ 2008
18:06:26:WU02:FS00:0xa7:    Options: /TP /nologo /EHa /wd4297 /wd4103 /Ox /MT
18:06:26:WU02:FS00:0xa7:   Platform: win32 10
18:06:26:WU02:FS00:0xa7:       Bits: 64
18:06:26:WU02:FS00:0xa7:       Mode: Release
18:06:26:WU02:FS00:0xa7:************************************ System ************************************
18:06:26:WU02:FS00:0xa7:        CPU: AMD Ryzen 7 3800X 8-Core Processor
18:06:26:WU02:FS00:0xa7:     CPU ID: AuthenticAMD Family 23 Model 113 Stepping 0
18:06:26:WU02:FS00:0xa7:       CPUs: 16
18:06:26:WU02:FS00:0xa7:     Memory: 31.93GiB
18:06:26:WU02:FS00:0xa7:Free Memory: 21.13GiB
18:06:26:WU02:FS00:0xa7:    Threads: WINDOWS_THREADS
18:06:26:WU02:FS00:0xa7: OS Version: 6.2
18:06:26:WU02:FS00:0xa7:Has Battery: false
18:06:26:WU02:FS00:0xa7: On Battery: false
18:06:26:WU02:FS00:0xa7: UTC Offset: -4
18:06:26:WU02:FS00:0xa7:        PID: 21064
18:06:26:WU02:FS00:0xa7:        CWD: C:\Users\therm\AppData\Roaming\FAHClient\work
18:06:26:WU02:FS00:0xa7:******************************** Build - libFAH ********************************
18:06:26:WU02:FS00:0xa7:    Version: 0.0.18
18:06:26:WU02:FS00:0xa7:     Author: Joseph Coffland <joseph@cauldrondevelopment.com>
18:06:27:WU02:FS00:0xa7:  Copyright: 2019 foldingathome.org
18:06:27:WU02:FS00:0xa7:   Homepage: https://foldingathome.org/
18:06:27:WU02:FS00:0xa7:       Date: Oct 26 2019
18:06:27:WU02:FS00:0xa7:       Time: 01:52:30
18:06:27:WU02:FS00:0xa7:   Revision: c1e3513b1bc0c16013668f2173ee969e5995b38e
18:06:27:WU02:FS00:0xa7:     Branch: master
18:06:27:WU02:FS00:0xa7:   Compiler: Visual C++ 2008
18:06:27:WU02:FS00:0xa7:    Options: /TP /nologo /EHa /wd4297 /wd4103 /Ox /MT
18:06:27:WU02:FS00:0xa7:   Platform: win32 10
18:06:27:WU02:FS00:0xa7:       Bits: 64
18:06:27:WU02:FS00:0xa7:       Mode: Release
18:06:27:WU02:FS00:0xa7:************************************ Build *************************************
18:06:27:WU02:FS00:0xa7:       SIMD: avx_256
18:06:27:WU02:FS00:0xa7:********************************************************************************
18:06:27:WU02:FS00:0xa7:Project: 16408 (Run 0, Clone 1373, Gen 57)
18:06:27:WU02:FS00:0xa7:Unit: 0x00000041a8f5c67d5e801d790b21ab48
18:06:27:WU02:FS00:0xa7:Reading tar file core.xml
18:06:27:WU02:FS00:0xa7:Reading tar file frame57.tpr
18:06:27:WU02:FS00:0xa7:Digital signatures verified
18:06:27:WU02:FS00:0xa7:Reducing thread count from 13 to 12 to avoid domain decomposition by a prime number > 3
18:06:27:WU02:FS00:0xa7:Calling: mdrun -s frame57.tpr -o frame57.trr -x frame57.xtc -cpt 15 -nt 12
18:06:27:WU02:FS00:0xa7:Steps: first=14250000 total=250000
18:06:28:WU02:FS00:0xa7:Completed 1 out of 250000 steps (0%)
18:06:31:WU03:FS00:Upload complete
18:06:31:WU03:FS00:Server responded WORK_ACK (400)
18:06:31:WU03:FS00:Final credit estimate, 6419.00 points
18:06:31:WU03:FS00:Cleaning up
18:06:38:WU00:FS01:0x22:Completed 4200000 out of 5000000 steps (84%)
18:08:23:WU02:FS00:0xa7:Completed 2500 out of 250000 steps (1%)
18:10:17:WU02:FS00:0xa7:Completed 5000 out of 250000 steps (2%)
18:10:33:WU00:FS01:0x22:Completed 4250000 out of 5000000 steps (85%)
18:12:12:WU02:FS00:0xa7:Completed 7500 out of 250000 steps (3%)
18:14:06:WU02:FS00:0xa7:Completed 10000 out of 250000 steps (4%)
18:14:28:WU00:FS01:0x22:Completed 4300000 out of 5000000 steps (86%)
18:16:01:WU02:FS00:0xa7:Completed 12500 out of 250000 steps (5%)
HaloJones
Posts: 906
Joined: Thu Jul 24, 2008 10:16 am

Re: Optimize GPU vs CPU

Post by HaloJones »

when you set it to 15, you may be dedicating what FAH thinks is a core to the GPU but you're really only dedicating half an actual hardware core.

try it with 12 as suggested. I think it will give the best mix.
single 1070

Image
Neil-B
Posts: 1996
Joined: Sun Mar 22, 2020 5:52 pm
Hardware configuration: 1: 2x Xeon E5-2697v3@2.60GHz, 512GB DDR4 LRDIMM, SSD Raid, Win10 Ent 20H2, Quadro K420 1GB, FAH 7.6.21
2: Xeon E3-1505Mv5@2.80GHz, 32GB DDR4, NVME, Win10 Pro 20H2, Quadro M1000M 2GB, FAH 7.6.21 (actually have two of these)
3: i7-960@3.20GHz, 12GB DDR3, SSD, Win10 Pro 20H2, GTX 750Ti 2GB, GTX 1080Ti 11GB, FAH 7.6.21
Location: UK

Re: Optimize GPU vs CPU

Post by Neil-B »

You may well have issues with a 13 slot as it is a prime number … better might be a 12 slot (highest count you can get that isn't likely to have issues and gets science done fastest and a 2nd slot with 2 or 3 depending on what suits your kit best and keeps the GPU happy.
2x Xeon E5-2697v3, 512GB DDR4 LRDIMM, SSD Raid, W10-Ent, Quadro K420
Xeon E3-1505Mv5, 32GB DDR4, NVME, W10-Pro, Quadro M1000M
i7-960, 12GB DDR3, SSD, W10-Pro, GTX1080Ti
i9-10850K, 64GB DDR4, NVME, W11-Pro, RTX3070

(Green/Bold = Active)
Pellicle
Posts: 5
Joined: Sat May 02, 2020 2:47 pm

Re: Optimize GPU vs CPU

Post by Pellicle »

Have set to 12 and 2 CPU threads using 2 slots.
Seems to be best again assuming utilization as shown by the system metrics is a refection of output.
Attached is full log.
Thanks for your input and help.

Code: Select all

17:56:01:Removing old file 'configs/config-20200426-182714.xml'
17:56:01:Saving configuration to config.xml
17:56:01:<config>
17:56:01:  <!-- Folding Core -->
17:56:01:  <core-priority v='low'/>
17:56:01:
17:56:01:  <!-- Network -->
17:56:01:  <proxy v=':8080'/>
17:56:01:
17:56:01:  <!-- Slot Control -->
17:56:01:  <pause-on-battery v='false'/>
17:56:01:  <power v='full'/>
17:56:01:
17:56:01:  <!-- User Information -->
17:56:01:  <passkey v='*****'/>
17:56:01:  <user v='Pellicle'/>
17:56:01:
17:56:01:  <!-- Folding Slots -->
17:56:01:  <slot id='0' type='CPU'>
17:56:01:    <cpus v='13'/>
17:56:01:    <paused v='False'/>
17:56:01:  </slot>
17:56:01:  <slot id='1' type='GPU'/>
17:56:01:</config>
17:56:01:FS00:Shutting core down
17:56:02:WU03:FS00:0xa7:WARNING:Console control signal 1 on PID 16932
17:56:02:WU03:FS00:0xa7:Exiting, please wait. . .
17:56:02:WU03:FS00:0xa7:Folding@home Core Shutdown: INTERRUPTED
17:56:03:WU03:FS00:FahCore returned: INTERRUPTED (102 = 0x66)
17:56:03:WU03:FS00:Starting
17:56:03:WARNING:WU03:FS00:Changed SMP threads from 15 to 13 this can cause some work units to fail
17:56:03:WU03:FS00:Running FahCore: "C:\Program Files (x86)\FAHClient/FAHCoreWrapper.exe" C:\Users\therm\AppData\Roaming\FAHClient\cores/cores.foldingathome.org/v7/win/64bit/avx/Core_a7.fah/FahCore_a7.exe -dir 03 -suffix 01 -version 706 -lifeline 8780 -checkpoint 15 -np 13
17:56:03:WU03:FS00:Started FahCore on PID 24740
17:56:03:WU03:FS00:Core PID:21080
17:56:03:WU03:FS00:FahCore 0xa7 started
17:56:03:WU03:FS00:0xa7:*********************** Log Started 2020-05-02T17:56:03Z ***********************
17:56:03:WU03:FS00:0xa7:************************** Gromacs Folding@home Core ***************************
17:56:03:WU03:FS00:0xa7:       Type: 0xa7
17:56:03:WU03:FS00:0xa7:       Core: Gromacs
17:56:03:WU03:FS00:0xa7:       Args: -dir 03 -suffix 01 -version 706 -lifeline 24740 -checkpoint 15 -np
17:56:03:WU03:FS00:0xa7:             13
17:56:03:WU03:FS00:0xa7:************************************ CBang *************************************
17:56:03:WU03:FS00:0xa7:       Date: Oct 26 2019
17:56:03:WU03:FS00:0xa7:       Time: 01:38:25
17:56:03:WU03:FS00:0xa7:   Revision: c46a1a011a24143739ac7218c5a435f66777f62f
17:56:03:WU03:FS00:0xa7:     Branch: master
17:56:03:WU03:FS00:0xa7:   Compiler: Visual C++ 2008
17:56:03:WU03:FS00:0xa7:    Options: /TP /nologo /EHa /wd4297 /wd4103 /Ox /MT
17:56:03:WU03:FS00:0xa7:   Platform: win32 10
17:56:03:WU03:FS00:0xa7:       Bits: 64
17:56:03:WU03:FS00:0xa7:       Mode: Release
17:56:03:WU03:FS00:0xa7:************************************ System ************************************
17:56:03:WU03:FS00:0xa7:        CPU: AMD Ryzen 7 3800X 8-Core Processor
17:56:03:WU03:FS00:0xa7:     CPU ID: AuthenticAMD Family 23 Model 113 Stepping 0
17:56:03:WU03:FS00:0xa7:       CPUs: 16
17:56:03:WU03:FS00:0xa7:     Memory: 31.93GiB
17:56:03:WU03:FS00:0xa7:Free Memory: 21.14GiB
17:56:03:WU03:FS00:0xa7:    Threads: WINDOWS_THREADS
17:56:03:WU03:FS00:0xa7: OS Version: 6.2
17:56:03:WU03:FS00:0xa7:Has Battery: false
17:56:03:WU03:FS00:0xa7: On Battery: false
17:56:03:WU03:FS00:0xa7: UTC Offset: -4
17:56:03:WU03:FS00:0xa7:        PID: 21080
17:56:03:WU03:FS00:0xa7:        CWD: C:\Users\therm\AppData\Roaming\FAHClient\work
17:56:03:WU03:FS00:0xa7:******************************** Build - libFAH ********************************
17:56:03:WU03:FS00:0xa7:    Version: 0.0.18
17:56:03:WU03:FS00:0xa7:     Author: Joseph Coffland <joseph@cauldrondevelopment.com>
17:56:03:WU03:FS00:0xa7:  Copyright: 2019 foldingathome.org
17:56:03:WU03:FS00:0xa7:   Homepage: https://foldingathome.org/
17:56:03:WU03:FS00:0xa7:       Date: Oct 26 2019
17:56:03:WU03:FS00:0xa7:       Time: 01:52:30
17:56:03:WU03:FS00:0xa7:   Revision: c1e3513b1bc0c16013668f2173ee969e5995b38e
17:56:03:WU03:FS00:0xa7:     Branch: master
17:56:03:WU03:FS00:0xa7:   Compiler: Visual C++ 2008
17:56:03:WU03:FS00:0xa7:    Options: /TP /nologo /EHa /wd4297 /wd4103 /Ox /MT
17:56:03:WU03:FS00:0xa7:   Platform: win32 10
17:56:03:WU03:FS00:0xa7:       Bits: 64
17:56:03:WU03:FS00:0xa7:       Mode: Release
17:56:03:WU03:FS00:0xa7:************************************ Build *************************************
17:56:03:WU03:FS00:0xa7:       SIMD: avx_256
17:56:03:WU03:FS00:0xa7:********************************************************************************
17:56:03:WU03:FS00:0xa7:Project: 14621 (Run 166, Clone 0, Gen 56)
17:56:03:WU03:FS00:0xa7:Unit: 0x000000419bf7a4d55e89f1ba97b4800f
17:56:03:WU03:FS00:0xa7:Digital signatures verified
17:56:03:WU03:FS00:0xa7:Reducing thread count from 13 to 12 to avoid domain decomposition by a prime number > 3
17:56:03:WU03:FS00:0xa7:Calling: mdrun -s frame56.tpr -o frame56.trr -cpi state.cpt -cpt 15 -nt 12
17:56:03:WU03:FS00:0xa7:Steps: first=0 total=250000
17:56:04:WU03:FS00:0xa7:Completed 210877 out of 250000 steps (84%)
17:56:30:WU03:FS00:0xa7:Completed 212500 out of 250000 steps (85%)
17:56:34:Removing old file 'configs/config-20200426-182815.xml'
17:56:34:Saving configuration to config.xml
17:56:34:<config>
17:56:34:  <!-- Folding Core -->
17:56:34:  <core-priority v='low'/>
17:56:34:
17:56:34:  <!-- Network -->
17:56:34:  <proxy v=':8080'/>
17:56:34:
17:56:34:  <!-- Slot Control -->
17:56:34:  <pause-on-battery v='false'/>
17:56:34:  <power v='full'/>
17:56:34:
17:56:34:  <!-- User Information -->
17:56:34:  <passkey v='*****'/>
17:56:34:  <user v='Pellicle'/>
17:56:34:
17:56:34:  <!-- Folding Slots -->
17:56:34:  <slot id='0' type='CPU'>
17:56:34:    <cpus v='13'/>
17:56:34:    <paused v='False'/>
17:56:34:  </slot>
17:56:34:  <slot id='1' type='GPU'/>
17:56:34:</config>
17:57:09:WU03:FS00:0xa7:Completed 215000 out of 250000 steps (86%)
17:57:49:WU03:FS00:0xa7:Completed 217500 out of 250000 steps (87%)
17:58:29:WU03:FS00:0xa7:Completed 220000 out of 250000 steps (88%)
17:58:58:WU00:FS01:0x22:Completed 4100000 out of 5000000 steps (82%)
17:59:08:WU03:FS00:0xa7:Completed 222500 out of 250000 steps (89%)
17:59:48:WU03:FS00:0xa7:Completed 225000 out of 250000 steps (90%)
18:00:27:WU03:FS00:0xa7:Completed 227500 out of 250000 steps (91%)
18:01:07:WU03:FS00:0xa7:Completed 230000 out of 250000 steps (92%)
18:01:47:WU03:FS00:0xa7:Completed 232500 out of 250000 steps (93%)
18:02:26:WU03:FS00:0xa7:Completed 235000 out of 250000 steps (94%)
18:02:49:WU00:FS01:0x22:Completed 4150000 out of 5000000 steps (83%)
18:03:06:WU03:FS00:0xa7:Completed 237500 out of 250000 steps (95%)
18:03:45:WU03:FS00:0xa7:Completed 240000 out of 250000 steps (96%)
18:04:25:WU03:FS00:0xa7:Completed 242500 out of 250000 steps (97%)
18:05:05:WU03:FS00:0xa7:Completed 245000 out of 250000 steps (98%)
18:05:45:WU03:FS00:0xa7:Completed 247500 out of 250000 steps (99%)
18:05:45:WU02:FS00:Connecting to 65.254.110.245:80
18:05:45:WU02:FS00:Assigned to work server 168.245.198.125
18:05:45:WU02:FS00:Requesting new work unit for slot 00: RUNNING cpu:13 from 168.245.198.125
18:05:45:WU02:FS00:Connecting to 168.245.198.125:8080
18:05:46:WU02:FS00:Downloading 13.97MiB
18:05:51:WU02:FS00:Download complete
18:05:51:WU02:FS00:Received Unit: id:02 state:DOWNLOAD error:NO_ERROR project:16408 run:0 clone:1373 gen:57 core:0xa7 unit:0x00000041a8f5c67d5e801d790b21ab48
18:06:25:WU03:FS00:0xa7:Completed 250000 out of 250000 steps (100%)
18:06:26:WU03:FS00:0xa7:Saving result file ..\logfile_01.txt
18:06:26:WU03:FS00:0xa7:Saving result file dhdl.xvg
18:06:26:WU03:FS00:0xa7:Saving result file frame56.trr
18:06:26:WU03:FS00:0xa7:Saving result file md.log
18:06:26:WU03:FS00:0xa7:Saving result file pullf.xvg
18:06:26:WU03:FS00:0xa7:Saving result file pullx.xvg
18:06:26:WU03:FS00:0xa7:Saving result file science.log
18:06:26:WU03:FS00:0xa7:Saving result file traj_comp.xtc
18:06:26:WU03:FS00:0xa7:Folding@home Core Shutdown: FINISHED_UNIT
18:06:26:WU03:FS00:FahCore returned: FINISHED_UNIT (100 = 0x64)
18:06:26:WU03:FS00:Sending unit results: id:03 state:SEND error:NO_ERROR project:14621 run:166 clone:0 gen:56 core:0xa7 unit:0x000000419bf7a4d55e89f1ba97b4800f
18:06:26:WU03:FS00:Uploading 6.47MiB to 155.247.164.213
18:06:26:WU03:FS00:Connecting to 155.247.164.213:8080
18:06:26:WU02:FS00:Starting
18:06:26:WU02:FS00:Running FahCore: "C:\Program Files (x86)\FAHClient/FAHCoreWrapper.exe" C:\Users\therm\AppData\Roaming\FAHClient\cores/cores.foldingathome.org/v7/win/64bit/avx/Core_a7.fah/FahCore_a7.exe -dir 02 -suffix 01 -version 706 -lifeline 8780 -checkpoint 15 -np 13
18:06:26:WU02:FS00:Started FahCore on PID 21616
18:06:26:WU02:FS00:Core PID:21064
18:06:26:WU02:FS00:FahCore 0xa7 started
18:06:26:WU02:FS00:0xa7:*********************** Log Started 2020-05-02T18:06:26Z ***********************
18:06:26:WU02:FS00:0xa7:************************** Gromacs Folding@home Core ***************************
18:06:26:WU02:FS00:0xa7:       Type: 0xa7
18:06:26:WU02:FS00:0xa7:       Core: Gromacs
18:06:26:WU02:FS00:0xa7:       Args: -dir 02 -suffix 01 -version 706 -lifeline 21616 -checkpoint 15 -np
18:06:26:WU02:FS00:0xa7:             13
18:06:26:WU02:FS00:0xa7:************************************ CBang *************************************
18:06:26:WU02:FS00:0xa7:       Date: Oct 26 2019
18:06:26:WU02:FS00:0xa7:       Time: 01:38:25
18:06:26:WU02:FS00:0xa7:   Revision: c46a1a011a24143739ac7218c5a435f66777f62f
18:06:26:WU02:FS00:0xa7:     Branch: master
18:06:26:WU02:FS00:0xa7:   Compiler: Visual C++ 2008
18:06:26:WU02:FS00:0xa7:    Options: /TP /nologo /EHa /wd4297 /wd4103 /Ox /MT
18:06:26:WU02:FS00:0xa7:   Platform: win32 10
18:06:26:WU02:FS00:0xa7:       Bits: 64
18:06:26:WU02:FS00:0xa7:       Mode: Release
18:06:26:WU02:FS00:0xa7:************************************ System ************************************
18:06:26:WU02:FS00:0xa7:        CPU: AMD Ryzen 7 3800X 8-Core Processor
18:06:26:WU02:FS00:0xa7:     CPU ID: AuthenticAMD Family 23 Model 113 Stepping 0
18:06:26:WU02:FS00:0xa7:       CPUs: 16
18:06:26:WU02:FS00:0xa7:     Memory: 31.93GiB
18:06:26:WU02:FS00:0xa7:Free Memory: 21.13GiB
18:06:26:WU02:FS00:0xa7:    Threads: WINDOWS_THREADS
18:06:26:WU02:FS00:0xa7: OS Version: 6.2
18:06:26:WU02:FS00:0xa7:Has Battery: false
18:06:26:WU02:FS00:0xa7: On Battery: false
18:06:26:WU02:FS00:0xa7: UTC Offset: -4
18:06:26:WU02:FS00:0xa7:        PID: 21064
18:06:26:WU02:FS00:0xa7:        CWD: C:\Users\therm\AppData\Roaming\FAHClient\work
18:06:26:WU02:FS00:0xa7:******************************** Build - libFAH ********************************
18:06:26:WU02:FS00:0xa7:    Version: 0.0.18
18:06:26:WU02:FS00:0xa7:     Author: Joseph Coffland <joseph@cauldrondevelopment.com>
18:06:27:WU02:FS00:0xa7:  Copyright: 2019 foldingathome.org
18:06:27:WU02:FS00:0xa7:   Homepage: https://foldingathome.org/
18:06:27:WU02:FS00:0xa7:       Date: Oct 26 2019
18:06:27:WU02:FS00:0xa7:       Time: 01:52:30
18:06:27:WU02:FS00:0xa7:   Revision: c1e3513b1bc0c16013668f2173ee969e5995b38e
18:06:27:WU02:FS00:0xa7:     Branch: master
18:06:27:WU02:FS00:0xa7:   Compiler: Visual C++ 2008
18:06:27:WU02:FS00:0xa7:    Options: /TP /nologo /EHa /wd4297 /wd4103 /Ox /MT
18:06:27:WU02:FS00:0xa7:   Platform: win32 10
18:06:27:WU02:FS00:0xa7:       Bits: 64
18:06:27:WU02:FS00:0xa7:       Mode: Release
18:06:27:WU02:FS00:0xa7:************************************ Build *************************************
18:06:27:WU02:FS00:0xa7:       SIMD: avx_256
18:06:27:WU02:FS00:0xa7:********************************************************************************
18:06:27:WU02:FS00:0xa7:Project: 16408 (Run 0, Clone 1373, Gen 57)
18:06:27:WU02:FS00:0xa7:Unit: 0x00000041a8f5c67d5e801d790b21ab48
18:06:27:WU02:FS00:0xa7:Reading tar file core.xml
18:06:27:WU02:FS00:0xa7:Reading tar file frame57.tpr
18:06:27:WU02:FS00:0xa7:Digital signatures verified
18:06:27:WU02:FS00:0xa7:Reducing thread count from 13 to 12 to avoid domain decomposition by a prime number > 3
18:06:27:WU02:FS00:0xa7:Calling: mdrun -s frame57.tpr -o frame57.trr -x frame57.xtc -cpt 15 -nt 12
18:06:27:WU02:FS00:0xa7:Steps: first=14250000 total=250000
18:06:28:WU02:FS00:0xa7:Completed 1 out of 250000 steps (0%)
18:06:31:WU03:FS00:Upload complete
18:06:31:WU03:FS00:Server responded WORK_ACK (400)
18:06:31:WU03:FS00:Final credit estimate, 6419.00 points
18:06:31:WU03:FS00:Cleaning up
18:06:38:WU00:FS01:0x22:Completed 4200000 out of 5000000 steps (84%)
18:08:23:WU02:FS00:0xa7:Completed 2500 out of 250000 steps (1%)
18:10:17:WU02:FS00:0xa7:Completed 5000 out of 250000 steps (2%)
18:10:33:WU00:FS01:0x22:Completed 4250000 out of 5000000 steps (85%)
18:12:12:WU02:FS00:0xa7:Completed 7500 out of 250000 steps (3%)
18:14:06:WU02:FS00:0xa7:Completed 10000 out of 250000 steps (4%)
18:14:28:WU00:FS01:0x22:Completed 4300000 out of 5000000 steps (86%)
18:16:01:WU02:FS00:0xa7:Completed 12500 out of 250000 steps (5%)
PantherX
Site Moderator
Posts: 6986
Joined: Wed Dec 23, 2009 9:33 am
Hardware configuration: V7.6.21 -> Multi-purpose 24/7
Windows 10 64-bit
CPU:2/3/4/6 -> Intel i7-6700K
GPU:1 -> Nvidia GTX 1080 Ti
§
Retired:
2x Nvidia GTX 1070
Nvidia GTX 675M
Nvidia GTX 660 Ti
Nvidia GTX 650 SC
Nvidia GTX 260 896 MB SOC
Nvidia 9600GT 1 GB OC
Nvidia 9500M GS
Nvidia 8800GTS 320 MB

Intel Core i7-860
Intel Core i7-3840QM
Intel i3-3240
Intel Core 2 Duo E8200
Intel Core 2 Duo E6550
Intel Core 2 Duo T8300
Intel Pentium E5500
Intel Pentium E5400
Location: Land Of The Long White Cloud
Contact:

Re: Optimize GPU vs CPU

Post by PantherX »

Pellicle wrote:Have set to 12 and 2 CPU threads using 2 slots....
17:56:01: <!-- Folding Slots -->
17:56:01: <slot id='0' type='CPU'>
17:56:01: <cpus v='13'/>
17:56:01: <paused v='False'/>
17:56:01: </slot>
17:56:01: <slot id='1' type='GPU'/>
17:56:01:</config>...
18:06:27:WU02:FS00:0xa7:Reducing thread count from 13 to 12 to avoid domain decomposition by a prime number > 3...
Please note that while the configuration has 13 CPUs in it, FahCore will automatically change it to 12. It would be better to simply change it to 12 in FAHControl to stop that message form being written. Also, you have a single CPU Slot using 12 CPUs and a GPU slot. I am not sure what you mean by "2 CPU threads using 2 slots"

I would say that this is the optimal setup. Sure you may have 3 CPUs unused but I won't suggest adding another CPU slot since folding performance is largely impacted by the physical cores which in this case are fully used. HT/SMT does provide an increased benefit but isn't as good as physical ones.
ETA:
Now ↞ Very Soon ↔ Soon ↔ Soon-ish ↔ Not Soon ↠ End Of Time

Welcome To The F@H Support Forum Ӂ Troubleshooting Bad WUs Ӂ Troubleshooting Server Connectivity Issues
Pellicle
Posts: 5
Joined: Sat May 02, 2020 2:47 pm

Re: Optimize GPU vs CPU

Post by Pellicle »

I have set 2 GPU slots
One with 12 threads and one with 2 threads
My CPU has 8 cores and 16 thread capability.
Not sure why the reduced count message unless log was not cleared when I reset the configuration.
Reverified setup in config is as noted above
The reason I am splitting the 1 physical CPU to 2 slots is to increase utilization of the GPU which seems to be limited when all 16 CPUs are in one slot.
I know that 1 is reserved to service the GPU but what I am observing is the GPU is waiting for the CPU to "load" it.
This is not a thermal throttling issue as both are cool and not throttled.
When I reduce Cores allocated to the CPU the GPU utilization increases significantly.
So the thought was to manually assign core counts. Adhere to the Prime number limitation and set 12 and 2 leaving 2 free. If there were not issues with prime I would set to 14 or 15 but have been advised against this.
Running this way I seem to be getting more utilization from the GPU which is more efficient a number crunching while sacrificing 1 thread of CPU capacity.
I am open to any input on the best way to optimize but default does not seem to be as effective in my system to maximize GPU output.
Alternate ideas welcome.
PantherX
Site Moderator
Posts: 6986
Joined: Wed Dec 23, 2009 9:33 am
Hardware configuration: V7.6.21 -> Multi-purpose 24/7
Windows 10 64-bit
CPU:2/3/4/6 -> Intel i7-6700K
GPU:1 -> Nvidia GTX 1080 Ti
§
Retired:
2x Nvidia GTX 1070
Nvidia GTX 675M
Nvidia GTX 660 Ti
Nvidia GTX 650 SC
Nvidia GTX 260 896 MB SOC
Nvidia 9600GT 1 GB OC
Nvidia 9500M GS
Nvidia 8800GTS 320 MB

Intel Core i7-860
Intel Core i7-3840QM
Intel i3-3240
Intel Core 2 Duo E8200
Intel Core 2 Duo E6550
Intel Core 2 Duo T8300
Intel Pentium E5500
Intel Pentium E5400
Location: Land Of The Long White Cloud
Contact:

Re: Optimize GPU vs CPU

Post by PantherX »

I understand your POV to have a system that fully utilized. However, there are two options:

Option 1 - Best utilization with maximum scientific contribution
You will have 1 CPU Slot with 12 CPUs assigned to it.
You will have 1 GPU Slot for your GPU.
You will notice that the overall system usage will be; CPU around 75% with bursts to 100% (this happens when the GPU writes checkpoints), GPU around 90% or higher.
The reason is that F@H desires that WUs are returned as quickly as possible. It isn't about the number of WUs your system can fold simultaneously, instead, it is about how quickly an assigned WU your system folds.

Option 2 - Highest utilization with negative impact on scientific contribution
You will have a CPU Slot with 12 CPUs assigned to it.
You will have another CPU Slot with 3 CPUs assigned to it.
You will have a GPU Slot for your GPU.
You will notice that the overall system usage will be; CPU around 94% with bursts to 100% (this happens when the GPU writes checkpoints), GPU around 90% or higher.
The reason this has a negative impact on scientific contribution is because your CPU is processing 2 WUs, not 1 thus, there's resource contention.

Please note that while you may have 16 threads, it is the physical 8 cores that folding fully utilize. The SMT/HT does provide some additional gains but not comparable to a physical core. Feel free to post your current configuration if you would like any additional clarification :)
ETA:
Now ↞ Very Soon ↔ Soon ↔ Soon-ish ↔ Not Soon ↠ End Of Time

Welcome To The F@H Support Forum Ӂ Troubleshooting Bad WUs Ӂ Troubleshooting Server Connectivity Issues
kiore
Posts: 924
Joined: Fri Jan 16, 2009 5:45 pm
Location: USA

Re: Optimize GPU vs CPU

Post by kiore »

I am running a similar rig but a Ryzen 3700x rather than a 3800 but still 8/16 cores, I am using what was recommended CPU=12 and the rest for the GPU which is giving me the best output for the project CPU wise without throttling the GPU which in windows needs one core/thread assigned and at times uses some more resources for checking. I know that it seems I am wasting 3 cores but overall this works better. You can experiment the mix yourself but you have received (above) good advice from people who know how this project uses resources.
Image
i7 7800x RTX 3070 OS= win10. AMD 3700x RTX 2080ti OS= win10 .

Team page: https://www.rationalskepticism.org/viewtopic.php?t=616
Neil-B
Posts: 1996
Joined: Sun Mar 22, 2020 5:52 pm
Hardware configuration: 1: 2x Xeon E5-2697v3@2.60GHz, 512GB DDR4 LRDIMM, SSD Raid, Win10 Ent 20H2, Quadro K420 1GB, FAH 7.6.21
2: Xeon E3-1505Mv5@2.80GHz, 32GB DDR4, NVME, Win10 Pro 20H2, Quadro M1000M 2GB, FAH 7.6.21 (actually have two of these)
3: i7-960@3.20GHz, 12GB DDR3, SSD, Win10 Pro 20H2, GTX 750Ti 2GB, GTX 1080Ti 11GB, FAH 7.6.21
Location: UK

Re: Optimize GPU vs CPU

Post by Neil-B »

To further expand the discussion to try and clarify my own understanding about resource contention:

As a matter of interest where does the resource contention break even? .. I haven't noticed significant (although I haven't looked too hard) differences in the PPD of either of my slots (32/56 and 24/56) when the other isn't folding (which is relatively rare) ... and I guess dropping the 24/56 might not be best for science as I would hope that the extra value of the 24/56 slot outweighs the relatively (from what I have seen) minimal impact from an resource contention ... but is there any rule of thumb? ... should a 2nd CPU slot only be used when say it is 4core, or 6core, or 8core? ... does the amount of contention vary dependant upon how many spare cores there are? ... is CPU slot to CPU slot resource contention as much of an issue as CPU slot to GPU slot resource contention? ... and I would guess it might also be reasonably kit dependant - would a multi CPU Xeon server handle CPU slot to CPU slot resource contention in a nicer manner than say a single CPU desktop/laptop? ... does the OS play into this Linux vs Win, Desktop vs Server? ... and type of CPU - Do Xeons handle multi-threaded cores more efficiently than hyperthreading - Is there an AMD equivalent of this?

I guess there won't be a simple answers to a few of the above questions - and any rule of thumb might be a bit long and have caveats - but hey, this is an interesting area of complexity that probably impacts most folders setups and an attempt at coalescing some answers in this thread will provide a reference thread when this topic next comes up ... as I couldn't find any previous thread with much more than passing mention.
2x Xeon E5-2697v3, 512GB DDR4 LRDIMM, SSD Raid, W10-Ent, Quadro K420
Xeon E3-1505Mv5, 32GB DDR4, NVME, W10-Pro, Quadro M1000M
i7-960, 12GB DDR3, SSD, W10-Pro, GTX1080Ti
i9-10850K, 64GB DDR4, NVME, W11-Pro, RTX3070

(Green/Bold = Active)
PantherX
Site Moderator
Posts: 6986
Joined: Wed Dec 23, 2009 9:33 am
Hardware configuration: V7.6.21 -> Multi-purpose 24/7
Windows 10 64-bit
CPU:2/3/4/6 -> Intel i7-6700K
GPU:1 -> Nvidia GTX 1080 Ti
§
Retired:
2x Nvidia GTX 1070
Nvidia GTX 675M
Nvidia GTX 660 Ti
Nvidia GTX 650 SC
Nvidia GTX 260 896 MB SOC
Nvidia 9600GT 1 GB OC
Nvidia 9500M GS
Nvidia 8800GTS 320 MB

Intel Core i7-860
Intel Core i7-3840QM
Intel i3-3240
Intel Core 2 Duo E8200
Intel Core 2 Duo E6550
Intel Core 2 Duo T8300
Intel Pentium E5500
Intel Pentium E5400
Location: Land Of The Long White Cloud
Contact:

Re: Optimize GPU vs CPU

Post by PantherX »

AFAIK, FahCores are not NUMA aware simply because at that time, multi-socket systems were not common. Fast forward a decade, Threadripper has used infinity fabric for HEDT. At this stage, I am unsure if FahCore will adapt to it or not. Moreover importantly, will Windows adapt to it or not since that could pay a part in adding those features to a future version of FahCore.

2 CPU slots "fighting" over a single CPU would be a higher resource contention then a CPU and GPU slot fighting over it. My rational is that CPU folding requires AVX instruction sets whereas the GPU just needs data and sanity checks every X% but it is sensitive to contention as GPUs are really fast. Hence, an interruption to data streaming from the CPU will cause the GPU usage to plummet.

If anyone is keen to test, see what the TPF is when using only physical cores and then physical plus virtual cores. The last time I did that on an i7-860, I found that I had about 12% to 23% reduction in TPF when I went from 4 physical cores to 8 CPUs so not the 50% reduction that one may expect when moving from 4 CPUs to 8 CPUs. For this usage, I suggest TPF rather than PPD as I am keen on seeing how the performance of the WU maps out as opposed to the change in PPD.
ETA:
Now ↞ Very Soon ↔ Soon ↔ Soon-ish ↔ Not Soon ↠ End Of Time

Welcome To The F@H Support Forum Ӂ Troubleshooting Bad WUs Ӂ Troubleshooting Server Connectivity Issues
Post Reply