FahCore_a8; over than 6 cores reduces speed

[mmm]

Console client 7.6.13; core A8 0.0.6; 2xE5-2609 (2x4Core); win 2008r2

Last time I have noticed many tasks for Core_A8 and very low PPD for it. I have been testing it by reducing cores and I have seen working speed with 8 cores was less than with 3 cores. So how I can fix it?

[PantherX]

Welcome to the F@H Forum mmm,

Can you please post the log file? Ensure you include the first 100 lines which will inform us of what the system configuration is and what the client settings are. If you require guidance, please view this topic: viewtopic.php?f=24&t=26036

Please note that CPU folding is sensitive to background tasks too. If you're using applications which use 1 or 2 CPUs, then folding with 8 CPUs would produce less PPD than folding with 6 CPUs since the other 2 CPUs are for application usage. Once we see how your client is configured, we might be able to provide some suggestions

[mmm]

Hi. Before posting my question I have backed up a client and calculate the same steps with different CPU quantity (--cpus=N) from 3 to 8. I have used two identical machines for the tests to reduce the time. Below logs for 3CPUs and full 8 CPUs. As you can see in each case time for one step is about equal.

3 cores

Code: Select all

*********************** Log Started 2020-09-12T21:18:13Z ***********************
21:18:13:Trying to access database...
21:18:13:Successfully acquired database lock
21:18:13:Read GPUs.txt
21:18:14:Enabled folding slot 04: READY cpu:3
21:18:14:****************************** FAHClient ******************************
21:18:14:      Version: 7.6.13
21:18:14:       Author: Joseph Coffland <joseph@cauldrondevelopment.com>
21:18:14:    Copyright: 2020 foldingathome.org
21:18:14:     Homepage: https://foldingathome.org/
21:18:14:         Date: Apr 27 2020
21:18:14:         Time: 21:21:01
21:18:14:     Revision: 5a652817f46116b6e135503af97f18e094414e3b
21:18:14:       Branch: master
21:18:14:     Compiler: Visual C++ 2008
21:18:14:      Options: /TP /nologo /EHa /wd4297 /wd4103 /Ox /MT
21:18:14:     Platform: win32 10
21:18:14:         Bits: 32
21:18:14:         Mode: Release
21:18:14:         Args: --cpus=3
21:18:14:       Config: ******\test\FAHClient\config.xml
21:18:14:******************************** CBang ********************************
21:18:14:         Date: Apr 24 2020
21:18:14:         Time: 17:07:55
21:18:14:     Revision: ea081a3b3b0f4a37c4d0440b4f1bc184197c7797
21:18:14:       Branch: master
21:18:14:     Compiler: Visual C++ 2008
21:18:14:      Options: /TP /nologo /EHa /wd4297 /wd4103 /Ox /MT
21:18:14:     Platform: win32 10
21:18:14:         Bits: 32
21:18:14:         Mode: Release
21:18:14:******************************* System ********************************
21:18:14:          CPU: Intel(R) Xeon(R) CPU E5-2609 0 @ 2.40GHz
21:18:14:       CPU ID: GenuineIntel Family 6 Model 45 Stepping 7
21:18:14:         CPUs: 8
21:18:14:       Memory: 31.94GiB
21:18:14:  Free Memory: 23.06GiB
21:18:14:      Threads: WINDOWS_THREADS
21:18:14:   OS Version: 6.1
21:18:14:  Has Battery: true
21:18:14:   On Battery: false
21:18:14:   UTC Offset: 3
21:18:14:          PID: 16136
21:18:14:          CWD: ******\test\FAHClient
21:18:14:Win32 Service: false
21:18:14:           OS: Windows Server 2008 R2 Standard
21:18:14:      OS Arch: AMD64
21:18:14:         GPUs: 0
21:18:14:         CUDA: Not detected: Failed to open dynamic library 'nvcuda.dll': Не
21:18:14:               найден указанный модуль.
21:18:14:
21:18:14:       OpenCL: Not detected: Failed to open dynamic library 'OpenCL.dll': Не
21:18:14:               найден указанный модуль.
21:18:14:
21:18:14:******************************* libFAH ********************************
21:18:14:         Date: Apr 15 2020
21:18:14:         Time: 14:53:14
21:18:14:     Revision: 216968bc7025029c841ed6e36e81a03a316890d3
21:18:14:       Branch: master
21:18:14:     Compiler: Visual C++ 2008
21:18:14:      Options: /TP /nologo /EHa /wd4297 /wd4103 /Ox /MT
21:18:14:     Platform: win32 10
21:18:14:         Bits: 32
21:18:14:         Mode: Release
21:18:14:***********************************************************************
21:18:14:<config>
21:18:14:  <!-- HTTP Server -->
21:18:14:  <allow v='******'/>
21:18:14:
21:18:14:  <!-- Network -->
21:18:14:  <proxy v='******'/>
21:18:14:  <proxy-enable v='true'/>
21:18:14:
21:18:14:  <!-- Remote Command Server -->
21:18:14:  <command-allow-no-pass v='******'/>
21:18:14:  <password v='*****'/>
21:18:14:
21:18:14:  <!-- Slot Control -->
21:18:14:  <power v='full'/>
21:18:14:
21:18:14:  <!-- User Information -->
21:18:14:  <passkey v='*****'/>
21:18:14:  <team v='******'/>
21:18:14:  <user v='******'/>
21:18:14:
21:18:14:  <!-- Folding Slots -->
21:18:14:  <slot id='4' type='CPU'/>
21:18:14:</config>
21:18:14:WU01:FS04:Starting
21:18:14:WARNING:WU01:FS04:Changed SMP threads from 8 to 3 this can cause some work units to fail
21:18:14:WU01:FS04:Running FahCore: ******\test\FAHClient/FAHCoreWrapper.exe ******\test\FAHClient\cores/cores.foldingathome.org/win/64bit-avx-256/a8-0.0.6/Core_a8.fah/FahCore_a8.exe -dir 01 -suffix 01 -version 706 -lifeline 16136 -checkpoint 15 -np 3
21:18:15:WU01:FS04:Started FahCore on PID 9684
21:18:15:WU01:FS04:Core PID:9588
21:18:15:WU01:FS04:FahCore 0xa8 started
21:18:15:WU01:FS04:0xa8:*********************** Log Started 2020-09-12T21:18:15Z ***********************
21:18:15:WU01:FS04:0xa8:************************** Gromacs Folding@home Core ***************************
21:18:15:WU01:FS04:0xa8:       Core: Gromacs
21:18:15:WU01:FS04:0xa8:       Type: 0xa8
21:18:15:WU01:FS04:0xa8:    Version: 0.0.6
21:18:15:WU01:FS04:0xa8:     Author: Joseph Coffland <joseph@cauldrondevelopment.com>
21:18:15:WU01:FS04:0xa8:  Copyright: 2020 foldingathome.org
21:18:15:WU01:FS04:0xa8:   Homepage: https://foldingathome.org/
21:18:15:WU01:FS04:0xa8:       Date: Jul 25 2020
21:18:15:WU01:FS04:0xa8:       Time: 06:59:38
21:18:15:WU01:FS04:0xa8:   Revision: bb73a0bb4b90096d0c4f404ba28d26418056a589
21:18:15:WU01:FS04:0xa8:     Branch: master
21:18:15:WU01:FS04:0xa8:   Compiler: Visual C++ 2019 16.6
21:18:15:WU01:FS04:0xa8:    Options: /TP /std:c++14 /nologo /EHa /wd4297 /wd4103 /O2 /Ob3
21:18:15:WU01:FS04:0xa8:             /Zc:throwingNew /MT
21:18:15:WU01:FS04:0xa8:   Platform: win32 10
21:18:15:WU01:FS04:0xa8:       Bits: 64
21:18:15:WU01:FS04:0xa8:       Mode: Release
21:18:15:WU01:FS04:0xa8:       SIMD: avx_256
21:18:15:WU01:FS04:0xa8:     OpenMP: ON
21:18:15:WU01:FS04:0xa8:       CUDA: OFF
21:18:15:WU01:FS04:0xa8:       Args: -dir 01 -suffix 01 -version 706 -lifeline 9684 -checkpoint 15 -np 3
21:18:15:WU01:FS04:0xa8:************************************ libFAH ************************************
21:18:15:WU01:FS04:0xa8:       Date: Jul 25 2020
21:18:15:WU01:FS04:0xa8:       Time: 06:58:32
21:18:15:WU01:FS04:0xa8:   Revision: bb73a0bb4b90096d0c4f404ba28d26418056a589
21:18:15:WU01:FS04:0xa8:     Branch: master
21:18:15:WU01:FS04:0xa8:   Compiler: Visual C++ 2019 16.6
21:18:15:WU01:FS04:0xa8:    Options: /TP /std:c++14 /nologo /EHa /wd4297 /wd4103 /O2 /Ob3
21:18:15:WU01:FS04:0xa8:             /Zc:throwingNew /MT
21:18:15:WU01:FS04:0xa8:   Platform: win32 10
21:18:15:WU01:FS04:0xa8:       Bits: 64
21:18:15:WU01:FS04:0xa8:       Mode: Release
21:18:15:WU01:FS04:0xa8:************************************ CBang *************************************
21:18:15:WU01:FS04:0xa8:       Date: Jul 25 2020
21:18:15:WU01:FS04:0xa8:       Time: 06:58:21
21:18:15:WU01:FS04:0xa8:   Revision: bb73a0bb4b90096d0c4f404ba28d26418056a589
21:18:15:WU01:FS04:0xa8:     Branch: master
21:18:15:WU01:FS04:0xa8:   Compiler: Visual C++ 2019 16.6
21:18:15:WU01:FS04:0xa8:    Options: /TP /std:c++14 /nologo /EHa /wd4297 /wd4103 /O2 /Ob3
21:18:15:WU01:FS04:0xa8:             /Zc:throwingNew /MT
21:18:15:WU01:FS04:0xa8:   Platform: win32 10
21:18:15:WU01:FS04:0xa8:       Bits: 64
21:18:15:WU01:FS04:0xa8:       Mode: Release
21:18:15:WU01:FS04:0xa8:************************************ System ************************************
21:18:15:WU01:FS04:0xa8:        CPU: Intel(R) Xeon(R) CPU E5-2609 0 @ 2.40GHz
21:18:15:WU01:FS04:0xa8:     CPU ID: GenuineIntel Family 6 Model 45 Stepping 7
21:18:15:WU01:FS04:0xa8:       CPUs: 8
21:18:15:WU01:FS04:0xa8:     Memory: 31.94GiB
21:18:15:WU01:FS04:0xa8:Free Memory: 23.05GiB
21:18:15:WU01:FS04:0xa8:    Threads: WINDOWS_THREADS
21:18:15:WU01:FS04:0xa8: OS Version: 6.1
21:18:15:WU01:FS04:0xa8:Has Battery: true
21:18:15:WU01:FS04:0xa8: On Battery: false
21:18:15:WU01:FS04:0xa8: UTC Offset: 3
21:18:15:WU01:FS04:0xa8:        PID: 9588
21:18:15:WU01:FS04:0xa8:        CWD: ******\test\FAHClient\work
21:18:15:WU01:FS04:0xa8:********************************************************************************
21:18:15:WU01:FS04:0xa8:Project: 16810 (Run 0, Clone 641, Gen 21)
21:18:15:WU01:FS04:0xa8:Unit: 0x00000016b2aec48a5eff2c95e9fe38b9
21:18:15:WU01:FS04:0xa8:Digital signatures verified
21:18:15:WU01:FS04:0xa8:Calling: mdrun -c frame21.gro -s frame21.tpr -x frame21.xtc -cpi state.cpt -cpt 15 -nt 3 -ntmpi 1
21:18:15:WU01:FS04:0xa8:Steps: first=10500000 total=11000000
21:18:17:WU01:FS04:0xa8:Completed 179942 out of 500000 steps (35%)
21:18:22:WU01:FS04:0xa8:Completed 180000 out of 500000 steps (36%)
21:25:40:WU01:FS04:0xa8:Completed 185000 out of 500000 steps (37%)
21:33:04:WU01:FS04:0xa8:Completed 190000 out of 500000 steps (38%)
21:40:21:WU01:FS04:0xa8:Completed 195000 out of 500000 steps (39%)
21:47:29:WU01:FS04:0xa8:Completed 200000 out of 500000 steps (40%)
21:54:40:WU01:FS04:0xa8:Completed 205000 out of 500000 steps (41%)
21:56:05:WARNING:Console control signal 0 on PID 16136
21:56:05:Exiting, please wait. . .
21:56:07:FS04:Shutting core down
21:56:07:WU01:FS04:0xa8:WARNING:Console control signal 1 on PID 9588
21:56:07:WU01:FS04:0xa8:Exiting, please wait. . .
21:56:10:Clean exit

8 cores

Code: Select all

*********************** Log Started 2020-09-12T20:24:01Z ***********************
20:24:01:Trying to access database...
20:24:01:Successfully acquired database lock
20:24:01:Read GPUs.txt
20:24:02:Enabled folding slot 04: READY cpu:8
20:24:02:****************************** FAHClient ******************************
20:24:02:      Version: 7.6.13
20:24:02:       Author: Joseph Coffland <joseph@cauldrondevelopment.com>
20:24:02:    Copyright: 2020 foldingathome.org
20:24:02:     Homepage: https://foldingathome.org/
20:24:02:         Date: Apr 27 2020
20:24:02:         Time: 21:21:01
20:24:02:     Revision: 5a652817f46116b6e135503af97f18e094414e3b
20:24:02:       Branch: master
20:24:02:     Compiler: Visual C++ 2008
20:24:02:      Options: /TP /nologo /EHa /wd4297 /wd4103 /Ox /MT
20:24:02:     Platform: win32 10
20:24:02:         Bits: 32
20:24:02:         Mode: Release
20:24:02:       Config: ******\FAHClient\config.xml
20:24:02:******************************** CBang ********************************
20:24:02:         Date: Apr 24 2020
20:24:02:         Time: 17:07:55
20:24:02:     Revision: ea081a3b3b0f4a37c4d0440b4f1bc184197c7797
20:24:02:       Branch: master
20:24:02:     Compiler: Visual C++ 2008
20:24:02:      Options: /TP /nologo /EHa /wd4297 /wd4103 /Ox /MT
20:24:02:     Platform: win32 10
20:24:02:         Bits: 32
20:24:02:         Mode: Release
20:24:02:******************************* System ********************************
20:24:02:          CPU: Intel(R) Xeon(R) CPU E5-2609 0 @ 2.40GHz
20:24:02:       CPU ID: GenuineIntel Family 6 Model 45 Stepping 7
20:24:02:         CPUs: 8
20:24:02:       Memory: 31.94GiB
20:24:02:  Free Memory: 8.27GiB
20:24:02:      Threads: WINDOWS_THREADS
20:24:02:   OS Version: 6.1
20:24:02:  Has Battery: false
20:24:02:   On Battery: false
20:24:02:   UTC Offset: 3
20:24:02:          PID: 5896
20:24:02:          CWD: ******\FAHClient
20:24:02:Win32 Service: false
20:24:02:           OS: Windows Server 2008 R2 Standard
20:24:02:      OS Arch: AMD64
20:24:02:         GPUs: 0
20:24:02:         CUDA: Not detected: Failed to open dynamic library 'nvcuda.dll': Не
20:24:02:               найден указанный модуль.
20:24:02:
20:24:02:       OpenCL: Not detected: Failed to open dynamic library 'OpenCL.dll': Не
20:24:02:               найден указанный модуль.
20:24:02:
20:24:02:******************************* libFAH ********************************
20:24:02:         Date: Apr 15 2020
20:24:02:         Time: 14:53:14
20:24:02:     Revision: 216968bc7025029c841ed6e36e81a03a316890d3
20:24:02:       Branch: master
20:24:02:     Compiler: Visual C++ 2008
20:24:02:      Options: /TP /nologo /EHa /wd4297 /wd4103 /Ox /MT
20:24:02:     Platform: win32 10
20:24:02:         Bits: 32
20:24:02:         Mode: Release
20:24:02:***********************************************************************
20:24:02:<config>
20:24:02:  <!-- HTTP Server -->
20:24:02:  <allow v='******'/>
20:24:02:
20:24:02:  <!-- Network -->
20:24:02:  <proxy v='******'/>
20:24:02:  <proxy-enable v='true'/>
20:24:02:
20:24:02:  <!-- Remote Command Server -->
20:24:02:  <command-allow-no-pass v='******'/>
20:24:02:  <password v='*****'/>
20:24:02:
20:24:02:  <!-- Slot Control -->
20:24:02:  <power v='full'/>
20:24:02:
20:24:02:  <!-- User Information -->
20:24:02:  <passkey v='*****'/>
20:24:02:  <team v='******'/>
20:24:02:  <user v='******'/>
20:24:02:
20:24:02:  <!-- Folding Slots -->
20:24:02:  <slot id='4' type='CPU'/>
20:24:02:</config>
20:24:02:WU01:FS04:Starting
20:24:02:WU01:FS04:Running FahCore: ******\FAHClient/FAHCoreWrapper.exe ******\FAHClient\cores/cores.foldingathome.org/win/64bit-avx-256/a8-0.0.6/Core_a8.fah/FahCore_a8.exe -dir 01 -suffix 01 -version 706 -lifeline 5896 -checkpoint 15 -np 8
20:24:02:WU01:FS04:Started FahCore on PID 7136
20:24:02:WU01:FS04:Core PID:8408
20:24:02:WU01:FS04:FahCore 0xa8 started
20:24:02:WU01:FS04:0xa8:*********************** Log Started 2020-09-12T20:24:02Z ***********************
20:24:02:WU01:FS04:0xa8:************************** Gromacs Folding@home Core ***************************
20:24:02:WU01:FS04:0xa8:       Core: Gromacs
20:24:02:WU01:FS04:0xa8:       Type: 0xa8
20:24:02:WU01:FS04:0xa8:    Version: 0.0.6
20:24:02:WU01:FS04:0xa8:     Author: Joseph Coffland <joseph@cauldrondevelopment.com>
20:24:02:WU01:FS04:0xa8:  Copyright: 2020 foldingathome.org
20:24:02:WU01:FS04:0xa8:   Homepage: https://foldingathome.org/
20:24:02:WU01:FS04:0xa8:       Date: Jul 25 2020
20:24:02:WU01:FS04:0xa8:       Time: 06:59:38
20:24:02:WU01:FS04:0xa8:   Revision: bb73a0bb4b90096d0c4f404ba28d26418056a589
20:24:02:WU01:FS04:0xa8:     Branch: master
20:24:02:WU01:FS04:0xa8:   Compiler: Visual C++ 2019 16.6
20:24:02:WU01:FS04:0xa8:    Options: /TP /std:c++14 /nologo /EHa /wd4297 /wd4103 /O2 /Ob3
20:24:03:WU01:FS04:0xa8:             /Zc:throwingNew /MT
20:24:03:WU01:FS04:0xa8:   Platform: win32 10
20:24:03:WU01:FS04:0xa8:       Bits: 64
20:24:03:WU01:FS04:0xa8:       Mode: Release
20:24:03:WU01:FS04:0xa8:       SIMD: avx_256
20:24:03:WU01:FS04:0xa8:     OpenMP: ON
20:24:03:WU01:FS04:0xa8:       CUDA: OFF
20:24:03:WU01:FS04:0xa8:       Args: -dir 01 -suffix 01 -version 706 -lifeline 7136 -checkpoint 15 -np 8
20:24:03:WU01:FS04:0xa8:************************************ libFAH ************************************
20:24:03:WU01:FS04:0xa8:       Date: Jul 25 2020
20:24:03:WU01:FS04:0xa8:       Time: 06:58:32
20:24:03:WU01:FS04:0xa8:   Revision: bb73a0bb4b90096d0c4f404ba28d26418056a589
20:24:03:WU01:FS04:0xa8:     Branch: master
20:24:03:WU01:FS04:0xa8:   Compiler: Visual C++ 2019 16.6
20:24:03:WU01:FS04:0xa8:    Options: /TP /std:c++14 /nologo /EHa /wd4297 /wd4103 /O2 /Ob3
20:24:03:WU01:FS04:0xa8:             /Zc:throwingNew /MT
20:24:03:WU01:FS04:0xa8:   Platform: win32 10
20:24:03:WU01:FS04:0xa8:       Bits: 64
20:24:03:WU01:FS04:0xa8:       Mode: Release
20:24:03:WU01:FS04:0xa8:************************************ CBang *************************************
20:24:03:WU01:FS04:0xa8:       Date: Jul 25 2020
20:24:03:WU01:FS04:0xa8:       Time: 06:58:21
20:24:03:WU01:FS04:0xa8:   Revision: bb73a0bb4b90096d0c4f404ba28d26418056a589
20:24:03:WU01:FS04:0xa8:     Branch: master
20:24:03:WU01:FS04:0xa8:   Compiler: Visual C++ 2019 16.6
20:24:03:WU01:FS04:0xa8:    Options: /TP /std:c++14 /nologo /EHa /wd4297 /wd4103 /O2 /Ob3
20:24:03:WU01:FS04:0xa8:             /Zc:throwingNew /MT
20:24:03:WU01:FS04:0xa8:   Platform: win32 10
20:24:03:WU01:FS04:0xa8:       Bits: 64
20:24:03:WU01:FS04:0xa8:       Mode: Release
20:24:03:WU01:FS04:0xa8:************************************ System ************************************
20:24:03:WU01:FS04:0xa8:        CPU: Intel(R) Xeon(R) CPU E5-2609 0 @ 2.40GHz
20:24:03:WU01:FS04:0xa8:     CPU ID: GenuineIntel Family 6 Model 45 Stepping 7
20:24:03:WU01:FS04:0xa8:       CPUs: 8
20:24:03:WU01:FS04:0xa8:     Memory: 31.94GiB
20:24:03:WU01:FS04:0xa8:Free Memory: 8.26GiB
20:24:03:WU01:FS04:0xa8:    Threads: WINDOWS_THREADS
20:24:03:WU01:FS04:0xa8: OS Version: 6.1
20:24:03:WU01:FS04:0xa8:Has Battery: false
20:24:03:WU01:FS04:0xa8: On Battery: false
20:24:03:WU01:FS04:0xa8: UTC Offset: 3
20:24:03:WU01:FS04:0xa8:        PID: 8408
20:24:03:WU01:FS04:0xa8:        CWD: ******\FAHClient\work
20:24:03:WU01:FS04:0xa8:********************************************************************************
20:24:03:WU01:FS04:0xa8:Project: 16810 (Run 0, Clone 641, Gen 21)
20:24:03:WU01:FS04:0xa8:Unit: 0x00000016b2aec48a5eff2c95e9fe38b9
20:24:03:WU01:FS04:0xa8:Digital signatures verified
20:24:03:WU01:FS04:0xa8:Calling: mdrun -c frame21.gro -s frame21.tpr -x frame21.xtc -cpi state.cpt -cpt 15 -nt 8 -ntmpi 1
20:24:03:WU01:FS04:0xa8:Steps: first=10500000 total=11000000
20:24:04:WU01:FS04:0xa8:Completed 179942 out of 500000 steps (35%)
20:24:09:WU01:FS04:0xa8:Completed 180000 out of 500000 steps (36%)
20:31:33:WU01:FS04:0xa8:Completed 185000 out of 500000 steps (37%)
20:38:47:WU01:FS04:0xa8:Completed 190000 out of 500000 steps (38%)
20:46:16:WU01:FS04:0xa8:Completed 195000 out of 500000 steps (39%)
20:53:38:WU01:FS04:0xa8:Completed 200000 out of 500000 steps (40%)
21:01:03:WU01:FS04:0xa8:Completed 205000 out of 500000 steps (41%)
21:02:25:WARNING:Console control signal 0 on PID 5896
21:02:25:Exiting, please wait. . .
21:02:26:FS04:Shutting core down
21:02:26:WU01:FS04:0xa8:WARNING:Console control signal 1 on PID 8408
21:02:26:WU01:FS04:0xa8:Exiting, please wait. . .
21:02:28:WU01:FS04:0xa8:Folding@home Core Shutdown: INTERRUPTED
21:02:28:Clean exit

[JimboPalmer]

Welcome to Folding@Home!

https://ark.intel.com/content/www/us/en ... l-qpi.html

This is a 4 core, 4 thread, Sandy Bridge CPU. You must have two of them. There will be some slowdown communicating between them.

Have you tried setting two CPU slots with 4 CPUs each?

[Joe_H]

Since there are two separate processors in these servers, communication between them will be affected by the NUMA settings in the BIOS and Windows. However changing them may or may not make a difference with Core_A8. The currently released version of the core has one of the NUMA related settings turned off due to a problem encountered during testing before release. A future revision may be released with that enabled again. So the current core will work well on multi-core single processor systems, but not quite as well on multi-processor systems.

[Neil-B]

May not be "quite as well" but still a significant increase on A7 for my 24/56 and 32/56 slots (system 1 - twin xeons) ... even more so when tested (without client) on a 54/56 slot - which was blindingly quick

[JimboPalmer]

Neil-B Your Xeons may be new enough to use AVX2, mmm's CPU is older and just does AVX.

[mmm]

Disabled NUMA gives +57% boosts on Core_A8 and 5-7% on Core_A7

[gunnarre]

Does the old "cpu-affinity" config setting do anything?

[mmm]

What do you mean under "old"?
I ran Core_A8 on 4 cores (--cpus=4). Win2008r2 shared by itself all the workload mostly on 4 last cores (4-7) and rarely slight workload was on the 3rd core.
After by hand binding to the cores 4-7 I noticed no changes. The results I have not written down.

[PantherX]

Does the old "cpu-affinity" config setting do anything?

Nope, it doesn't do anything.

[PantherX]

...Win2008r2 shared by itself all the workload mostly on 4 last cores (4-7) and rarely slight workload was on the 3rd core.
After by hand binding to the cores 4-7 I noticed no changes...

The OS is responsible for scheduling and can choose what CPUs to use. It is good to know that it works on your system. On my system, (Windows 10 64-bit 2004) it is all over the place and ended up using process lasso to ensure that it uses those specific CPUs.