How much ppd do you get?

Discussion in 'Folding@Home - Join Team Guru3D !' started by ra_alfaomega, Sep 24, 2016.

  1. ra_alfaomega

    ra_alfaomega Member Guru

    Messages:
    120
    Likes Received:
    0
    GPU:
    Gtx1070 MSI Gaming X
    Thanks for your reply :)
    When I said cpu:6 I meant how appears in fah control panel, so with cpu:6 fah uses half of the processing power of my 3930 k. 6 from 12 threads, and the ppd is 19-20 K. With cpu: 10 I have 36k. 1 thread is used by the gpu for fah, so 1 thread it's not used, but can be if I buy another gpu.
    So if in the future I buy a 1060 I don't have to link it by sli bridge to my 1070, I just have to put that card in a pci express slot and when entering fah control the card will be recognised?
     
  2. PantherX

    PantherX Folding@Home Team Leader

    Messages:
    1,380
    Likes Received:
    0
    GPU:
    Nvidia GTX 1080 Ti
    It's great to hear that jumping from 6 CPUs to 10 CPUs gives a nice boost in PPD!

    Yep, you pop-in the GTX 1060, hook it up to the PSU (if it needs to) and F@H should pick it up automatically without issues when you go to add another GPU Slot :) However, if you have angered the F@H gods, can always help you out :p
     
  3. ra_alfaomega

    ra_alfaomega Member Guru

    Messages:
    120
    Likes Received:
    0
    GPU:
    Gtx1070 MSI Gaming X
    Thanks a lot PantherX for your support and answers :)
     
  4. PantherX

    PantherX Folding@Home Team Leader

    Messages:
    1,380
    Likes Received:
    0
    GPU:
    Nvidia GTX 1080 Ti
    Not a problem, mate! Always happy to help out a fellow folder :D
     

  5. ra_alfaomega

    ra_alfaomega Member Guru

    Messages:
    120
    Likes Received:
    0
    GPU:
    Gtx1070 MSI Gaming X
  6. iancook221188

    iancook221188 Ancient Guru

    Messages:
    1,729
    Likes Received:
    20
    GPU:
    GTX 670 SLI / GTX 460 SLI
    yer ive just got them too 50% bump on my 2600k
     
  7. PantherX

    PantherX Folding@Home Team Leader

    Messages:
    1,380
    Likes Received:
    0
    GPU:
    Nvidia GTX 1080 Ti
    I have read elsewhere (unofficial) that FahCore_a7 is new for CPU and aims to make CPU folding more balanced PPD-wise. If that is indeed the case, we can expect official news soon :)
     
  8. ra_alfaomega

    ra_alfaomega Member Guru

    Messages:
    120
    Likes Received:
    0
    GPU:
    Gtx1070 MSI Gaming X
    Maybe this core is more efficient than a4 and it's faster, I think the ppd are in relation to the science being made.

    Me too>From 36k to 56K . more than 50% :)
     
  9. iancook221188

    iancook221188 Ancient Guru

    Messages:
    1,729
    Likes Received:
    20
    GPU:
    GTX 670 SLI / GTX 460 SLI
    i was gonna stop the cpu folding but this may have changed my mind
     
  10. PantherX

    PantherX Folding@Home Team Leader

    Messages:
    1,380
    Likes Received:
    0
    GPU:
    Nvidia GTX 1080 Ti
    I wonder if it uses the AVX/AVX2 for getting better performance. Here are my stats:
    CPU:6 when from 25K to 83K with TPF of 41 seconds. I would have to say, it's great news for my system :)
     

  11. ra_alfaomega

    ra_alfaomega Member Guru

    Messages:
    120
    Likes Received:
    0
    GPU:
    Gtx1070 MSI Gaming X
    Maybe, I don't have AVX/AVX2 on 3930k , and your ppd increased by more than 200%, that's incredible. It seems that the newer cpu's have a greater benefit. But I'm happy with 56k.So a 6950x maybe can get well over 200k.

    You are right, now I want to fold more on the cpu, than before ;) It was a very big difference between cpu's and gpu's ppd wise. So I understand the sofware behind the core is very important, to get the best of our cpu's.
     
  12. PantherX

    PantherX Folding@Home Team Leader

    Messages:
    1,380
    Likes Received:
    0
    GPU:
    Nvidia GTX 1080 Ti
    I do like the PPD boost as long as it is related to the science (not an easy thing). My best guess is that they might have an initial trial, see what PPD range is there and may optimize the PPD to reflect the actual science being done. They tend to do it when a new FahCore is released so let's see what happens.

    This is another thread from another team discussing FahCore_a7 (http://www.overclock.net/t/1613533/new-cpu-core-ax7-not-sure-if-this-is-news). Seems to me that it is very new so hopefully, can get more details, would be a nice Christmas gift for my CPU [​IMG]
     
  13. ra_alfaomega

    ra_alfaomega Member Guru

    Messages:
    120
    Likes Received:
    0
    GPU:
    Gtx1070 MSI Gaming X
    You are right, the ppd must reflect the actual science, anyway if a7 is more efficient, the ppd also have to increase.

    So, this new core uses, besides other, the avx, which it seems to have a big impact in the simulation of proteins. In the description of the 11920 project they say are using an old 9752 a4 core project. I remember this project. It was one with the biggest ppd , except bigadv projects. If I remember correctly with cpu:10 at @4,4 gave me in the range of 50k. Now with cpu:10 this "new" project 55-56k. So it's a bit faster than a4, but I think the biggest gain is the use of avx, which I don't have at this cpu :( .

    Thanks for the thread, maybe somebody with avx and 12, 16 or 20 threads can say how much ppd they have. Anyway, my next cpu will be with at least 8/16, and must have avx. I'm wondering if Zen has avx?:3eyes:
     
  14. ra_alfaomega

    ra_alfaomega Member Guru

    Messages:
    120
    Likes Received:
    0
    GPU:
    Gtx1070 MSI Gaming X
    I have to correct myself:bang:, it seems I have avx, but I don't have avx 2, that's why I don't see a big gain in ppd?
     
  15. PantherX

    PantherX Folding@Home Team Leader

    Messages:
    1,380
    Likes Received:
    0
    GPU:
    Nvidia GTX 1080 Ti
    I did remember that PG was looking into AVX support but that was a while back (few years). However, the good news is that "stuff" is happening in the background (https://foldingforum.org/viewtopic.php?f=72&t=28119) so until they officially announce what changes are in FahCore_a7 and/or FahCore_a8, your guess is as good as mine. Given that I got it while not using any flags and there wasn't any blog post about misconfigured server(s), I would say that this is intention so a public announcement could be soonish :)

    I did compare our CPUs (http://ark.intel.com/compare/88195,63697) and have a few thoughts:
    1) There is 3 Generation's "performance gap". I have briefly read articles that performance jump between each generation is 5% to 10% but not sure how that relates to F@H performance.
    2) i7-6700K has SSE4.1 and AVX2 instructions which yours don't. Not sure if the PPD increase is from SSE4.1 and/or AVX2 instructions.
    3) While you use DDR3 and I use DDR4, the difference between RAM timings can't really account for that difference. May be a percentage or two but not more than that unless there's some dramatic changes that I am unaware of.

    AMD Zen I do believe might have AVX support natively but AVX2 isn't nativly supported... Until it is released, it might be unclear:
    Source (http://www.anandtech.com/show/10591...t-2-extracting-instructionlevel-parallelism/2)
    Source (http://www.anandtech.com/show/10591...t-2-extracting-instructionlevel-parallelism/4)

    I guess 2017 might be an awesome year for CPU folding [​IMG]
     

  16. ra_alfaomega

    ra_alfaomega Member Guru

    Messages:
    120
    Likes Received:
    0
    GPU:
    Gtx1070 MSI Gaming X
    Interesting information regarding zen but we have almost no clue regarding how zen will do with fah.I hope this new architecture will be a big step for amd to close the distance between them and intel. If on the gpu side we have tons of information regarding the performance, the cpu segment is lacking ppd numbers. That's why is important to inform others about the cpu ppd we have. The cpu folding is important too...
    For now I am still happy with my 3930k regardless of lacking SSE4.1 and AVX2. And the fact that intel gains in ipc are very small 5-10% per generation, and the amount of money they charge, in the near future, I don't plan to upgrade. But Zen is interesting and like you said 2017 could be a very exciting year for cpu's...
    So, there is an a8 core also?
     
  17. ra_alfaomega

    ra_alfaomega Member Guru

    Messages:
    120
    Likes Received:
    0
    GPU:
    Gtx1070 MSI Gaming X
    Project 11920, cpu:12 - 67k tpf 0:47. In the near future I will try at @4,4.
     
  18. PantherX

    PantherX Folding@Home Team Leader

    Messages:
    1,380
    Likes Received:
    0
    GPU:
    Nvidia GTX 1080 Ti
    I did dig into the F@H log and found the first time I was assigned a FahCore_a7 WU:
    Code:
    20:42:36:WU01:FS00:Connecting to 171.67.108.45:8080
    20:42:37:WU01:FS00:Assigned to work server 171.64.65.41
    20:42:37:WU01:FS00:Requesting new work unit for slot 00: RUNNING cpu:6 from 171.64.65.41
    20:42:37:WU01:FS00:Connecting to 171.64.65.41:8080
    20:42:38:WU01:FS00:Downloading 20.96MiB
    20:42:43:WU01:FS00:Download complete
    20:42:43:WU01:FS00:Received Unit: id:01 state:DOWNLOAD error:NO_ERROR project:11920 run:194 clone:0 gen:1 core:0xa7 unit:0x00000001ab40412957bff714d8dbe4fa
    20:42:43:WU01:FS00:Downloading core from http://web.stanford.edu/~pande/Win32/AMD64/AVX/Core_a7.fah
    20:42:43:WU01:FS00:Connecting to web.stanford.edu:80
    20:42:43:WU01:FS00:FahCore a7: Downloading 7.35MiB
    20:42:49:WU01:FS00:FahCore a7: 27.22%
    20:42:55:WU01:FS00:FahCore a7: 59.55%
    20:43:01:WU01:FS00:FahCore a7: 91.87%
    20:43:02:WU01:FS00:FahCore a7: Download complete
    20:43:02:WU01:FS00:Valid core signature
    20:43:02:WU01:FS00:Unpacked 13.14MiB to cores/web.stanford.edu/~pande/Win32/AMD64/AVX/Core_a7.fah/FahCore_a7.exe
    20:43:02:WU01:FS00:Unpacked 72.16KiB to cores/web.stanford.edu/~pande/Win32/AMD64/AVX/Core_a7.fah/libbz2-1.dll
    20:43:02:WU01:FS00:Unpacked 2.17MiB to cores/web.stanford.edu/~pande/Win32/AMD64/AVX/Core_a7.fah/libeay32.dll
    20:43:02:WU01:FS00:Unpacked 154.79KiB to cores/web.stanford.edu/~pande/Win32/AMD64/AVX/Core_a7.fah/libexpat-1.dll
    20:43:02:WU01:FS00:Unpacked 2.11MiB to cores/web.stanford.edu/~pande/Win32/AMD64/AVX/Core_a7.fah/libfftw3f-3.dll
    20:43:02:WU01:FS00:Unpacked 406.90KiB to cores/web.stanford.edu/~pande/Win32/AMD64/AVX/Core_a7.fah/ssleay32.dll
    20:43:02:WU01:FS00:Unpacked 88.13KiB to cores/web.stanford.edu/~pande/Win32/AMD64/AVX/Core_a7.fah/zlib1.dll
    20:43:02:WU01:FS00:Unpacked 81.28KiB to cores/web.stanford.edu/~pande/Win32/AMD64/AVX/Core_a7.fah/libgcc_s_seh-1.dll
    20:43:02:WU01:FS00:Unpacked 55.64KiB to cores/web.stanford.edu/~pande/Win32/AMD64/AVX/Core_a7.fah/libwinpthread-1.dll
    20:43:02:WU01:FS00:Unpacked 1.35MiB to cores/web.stanford.edu/~pande/Win32/AMD64/AVX/Core_a7.fah/libstdc++-6.dll
    20:43:02:WU01:FS00:Starting
    20:43:02:WU01:FS00:Running FahCore: "C:\Program Files (x86)\FAHClient/FAHCoreWrapper.exe" C:/Users/Tweaker/AppData/Roaming/FAHClient/cores/web.stanford.edu/~pande/Win32/AMD64/AVX/Core_a7.fah/FahCore_a7.exe -dir 01 -suffix 01 -version 704 -lifeline 3012 -checkpoint 15 -np 6
    20:43:02:WU01:FS00:Started FahCore on PID 140
    20:43:02:WU01:FS00:Core PID:5712
    20:43:02:WU01:FS00:FahCore 0xa7 started
    20:43:03:WU01:FS00:0xa7:*********************** Log Started 2016-10-12T20:43:02Z ***********************
    20:43:03:WU01:FS00:0xa7:************************** Gromacs Folding@home Core ***************************
    20:43:03:WU01:FS00:0xa7:       Type: 0xa7
    20:43:03:WU01:FS00:0xa7:       Core: Gromacs
    20:43:03:WU01:FS00:0xa7:    Website: http://folding.stanford.edu/
    20:43:03:WU01:FS00:0xa7:  Copyright: (c) 2009-2016 Stanford University
    20:43:03:WU01:FS00:0xa7:     Author: Joseph Coffland <joseph@cauldrondevelopment.com>
    20:43:03:WU01:FS00:0xa7:       Args: -dir 01 -suffix 01 -version 704 -lifeline 140 -checkpoint 15 -np 6
    20:43:03:WU01:FS00:0xa7:     Config: <none>
    20:43:03:WU01:FS00:0xa7:************************************ Build *************************************
    20:43:03:WU01:FS00:0xa7:    Version: 0.0.11
    20:43:03:WU01:FS00:0xa7:       Date: Sep 21 2016
    20:43:03:WU01:FS00:0xa7:       Time: 01:43:48
    20:43:03:WU01:FS00:0xa7: Repository: Git
    20:43:03:WU01:FS00:0xa7:   Revision: 957bd90e68d95ddcf1594dc15ff6c64cc4555146
    20:43:03:WU01:FS00:0xa7:     Branch: master
    20:43:03:WU01:FS00:0xa7:   Compiler: GNU 4.2.1 Compatible Clang 3.9.0 (trunk 274080)
    20:43:03:WU01:FS00:0xa7:    Options: -std=gnu++98 -O3 -funroll-loops -ffast-math -mfpmath=sse
    20:43:03:WU01:FS00:0xa7:             -fno-unsafe-math-optimizations -msse2 -I/mingw64/include
    20:43:03:WU01:FS00:0xa7:             -Wno-inconsistent-dllimport -Wno-parentheses-equality
    20:43:03:WU01:FS00:0xa7:             -Wno-deprecated-register -Wno-unused-local-typedef
    20:43:03:WU01:FS00:0xa7:   Platform: linux2 4.6.0-1-amd64
    20:43:03:WU01:FS00:0xa7:       Bits: 64
    20:43:03:WU01:FS00:0xa7:       Mode: Release
    20:43:03:WU01:FS00:0xa7:       SIMD: avx_256
    20:43:03:WU01:FS00:0xa7:************************************ System ************************************
    20:43:03:WU01:FS00:0xa7:        CPU: Intel(R) Core(TM) i7-6700K CPU @ 4.00GHz
    20:43:03:WU01:FS00:0xa7:     CPU ID: GenuineIntel Family 6 Model 94 Stepping 3
    20:43:03:WU01:FS00:0xa7:       CPUs: 8
    20:43:03:WU01:FS00:0xa7:     Memory: 31.93GiB
    20:43:03:WU01:FS00:0xa7:Free Memory: 28.83GiB
    20:43:03:WU01:FS00:0xa7:    Threads: WINDOWS_THREADS
    20:43:03:WU01:FS00:0xa7: OS Version: 6.2
    20:43:03:WU01:FS00:0xa7:Has Battery: false
    20:43:03:WU01:FS00:0xa7: On Battery: false
    20:43:03:WU01:FS00:0xa7: UTC Offset: 13
    20:43:03:WU01:FS00:0xa7:        PID: 5712
    20:43:03:WU01:FS00:0xa7:        CWD: C:\Users\Tweaker\AppData\Roaming\FAHClient\work
    20:43:03:WU01:FS00:0xa7:         OS: Windows 10 Pro
    20:43:03:WU01:FS00:0xa7:    OS Arch: AMD64
    20:43:03:WU01:FS00:0xa7:********************************************************************************
    20:43:03:WU01:FS00:0xa7:Project: 11920 (Run 194, Clone 0, Gen 1)
    20:43:03:WU01:FS00:0xa7:Unit: 0x00000001ab40412957bff714d8dbe4fa
    20:43:03:WU01:FS00:0xa7:Reading tar file core.xml
    20:43:03:WU01:FS00:0xa7:Reading tar file frame1.tpr
    20:43:03:WU01:FS00:0xa7:Digital signatures verified
    20:43:03:WU01:FS00:0xa7:Calling: mdrun -s frame1.tpr -o frame1.trr -cpt 15 -nt 6
    20:43:03:WU01:FS00:0xa7:Steps: first=80000 total=80000
    20:43:05:WU01:FS00:0xa7:Completed 1 out of 80000 steps (0%)
    20:43:46:WU01:FS00:0xa7:Completed 800 out of 80000 steps (1%)
    20:44:27:WU01:FS00:0xa7:Completed 1600 out of 80000 steps (2%)
    20:45:09:WU01:FS00:0xa7:Completed 2400 out of 80000 steps (3%)
    20:45:49:WU01:FS00:0xa7:Completed 3200 out of 80000 steps (4%)
    20:46:30:WU01:FS00:0xa7:Completed 4000 out of 80000 steps (5%)
    
    While I haven't gotten the knowledge to decode the flags of the compiler, I did come across this:
    SIMD: avx_256
    Now I am unsure if this is AVX or AVX2 as the Wikipedia isn't very clear about that and the few threads that I have read about, aren't very clear about it :/

    Moreover, this FahCore is under a month old which is great:
    Date: Sep 21 2016

    Hopefully, they can give us an early Christmas gift by announcing an AVX and/or AVX2 FahCore along with a new FahClient (currently it is undergoing open Public Beta).

    According to the F@H thread and the project information (http://fah-web.stanford.edu/cgi-bin/fahproject.overusingIPswillbebanned?p=11900), yep but that could be an internal build or a typo. My best guess is that FahCore_a7 is AVX while FahCore_a8 is AVX2. However, it is just a guess :)
     
  19. Athlonite

    Athlonite Maha Guru

    Messages:
    1,358
    Likes Received:
    52
    GPU:
    Pulse RX5700 8GB
    I'm only a couple of WU's in on my RX480 and it's getting 346174 PPD and my cpu is getting 10170 PPD

    fairly big step up from my R9 285 on 0x21 core

    @ PantherX
    I've never seen an 0xa7 or a8 core WU always with the 0xa4
     
    Last edited: Dec 22, 2016
  20. ra_alfaomega

    ra_alfaomega Member Guru

    Messages:
    120
    Likes Received:
    0
    GPU:
    Gtx1070 MSI Gaming X
    That's very good. That's more than 3x what I used to get with 270x. And it's half of my 1070.
    You will have a great contribution to fah with your RX480 :)

    I don't know why, but I know that for several days I didn't received an a7 core project, maybe because I fold just 1-2 projects per day with my cpu, and the chances to get one are slim. Regarding a8 I never received this kind of WU.
     

Share This Page