Results 1 to 11 of 11

Thread: Windows and Linux benching

  1. #1
    HEXUS.social member Agent's Avatar
    Join Date
    Jul 2003
    Location
    Internet
    Posts
    19,168
    Thanks
    735
    Thanked
    1,607 times in 1,045 posts

    Windows and Linux benching

    I've got some free time after next Wednesday and was thinking of benching the two against each other.

    I assume the data files it downloads from the server are universal and the client does all the leg work?

    Also, if these are universal, how much so? I'm guessing they will be slightly different types of file for the PS3 & GPU's.
    What about PPC & x86 data? Same file and work done on the client? If so, it should make benching across architectures easy too
    Quote Originally Posted by Saracen View Post
    And by trying to force me to like small pants, they've alienated me.

  2. #2
    Comfortably Numb directhex's Avatar
    Join Date
    Jul 2003
    Location
    /dev/urandom
    Posts
    17,074
    Thanks
    228
    Thanked
    1,027 times in 678 posts
    • directhex's system
      • Motherboard:
      • Asus ROG Strix B550-I Gaming
      • CPU:
      • Ryzen 5900x
      • Memory:
      • 64GB G.Skill Trident Z RGB
      • Storage:
      • 2TB Seagate Firecuda 520
      • Graphics card(s):
      • EVGA GeForce RTX 3080 XC3 Ultra
      • PSU:
      • EVGA SuperNOVA 850W G3
      • Case:
      • NZXT H210i
      • Operating System:
      • Ubuntu 20.04, Windows 10
      • Monitor(s):
      • LG 34GN850
      • Internet:
      • FIOS

    Re: Windows and Linux benching

    problem is, you can't guarantee which cores a job will require. well, you can make a pretty educated guess for SMP (a1), but not for the regular cpu client

    ps3, gpu, smp, and uniprocessor units will all have different input, but can be cross-compared processorwise (e.g. smp windows versus smp linux is okay, but not gpu versus smp as they use different input data)

  3. #3
    HEXUS.social member Agent's Avatar
    Join Date
    Jul 2003
    Location
    Internet
    Posts
    19,168
    Thanks
    735
    Thanked
    1,607 times in 1,045 posts

    Re: Windows and Linux benching

    Yeah, was 99% sure on the PS3/GPU one. Any idea on PPC / x86?

    Quote Originally Posted by directhex View Post
    problem is, you can't guarantee which cores a job will require. well, you can make a pretty educated guess for SMP (a1), but not for the regular cpu client
    Why would the core its running on make a difference? If they are all the same speed / cache level, performance differences should be minimal I would have thought?
    Or do you mean the architecture by that?

    I guess I can give it a shot - worst that can happen is an error (I hope )
    Quote Originally Posted by Saracen View Post
    And by trying to force me to like small pants, they've alienated me.

  4. #4
    Comfortably Numb directhex's Avatar
    Join Date
    Jul 2003
    Location
    /dev/urandom
    Posts
    17,074
    Thanks
    228
    Thanked
    1,027 times in 678 posts
    • directhex's system
      • Motherboard:
      • Asus ROG Strix B550-I Gaming
      • CPU:
      • Ryzen 5900x
      • Memory:
      • 64GB G.Skill Trident Z RGB
      • Storage:
      • 2TB Seagate Firecuda 520
      • Graphics card(s):
      • EVGA GeForce RTX 3080 XC3 Ultra
      • PSU:
      • EVGA SuperNOVA 850W G3
      • Case:
      • NZXT H210i
      • Operating System:
      • Ubuntu 20.04, Windows 10
      • Monitor(s):
      • LG 34GN850
      • Internet:
      • FIOS

    Re: Windows and Linux benching

    Quote Originally Posted by Agent View Post
    Why would the core its running on make a difference? If they are all the same speed / cache level, performance differences should be minimal I would have thought?
    Or do you mean the architecture by that?
    i mean the computation core, i.e. the version of GROMACS being used. you can't mix non-matching input files with FahCore_XX.exe files.

    Fr'example, the last few WUs run on this single core box have required:
    [04:46:43] Core required: FahCore_78.exe
    [20:46:47] Core required: FahCore_81.exe
    [07:19:02] Core required: FahCore_81.exe
    [02:23:07] Core required: FahCore_81.exe
    [21:57:22] Core required: FahCore_81.exe
    [15:26:53] Core required: FahCore_81.exe
    [10:15:01] Core required: FahCore_81.exe
    [02:34:11] Core required: FahCore_81.exe
    [21:47:52] Core required: FahCore_81.exe

    what happens if on machine A i have access to 78, and on machine B i have access to 81? I can't just mix the input files, as they're incompatible. and only the master fah executable determines which cores to download when

    perhaps you could compile up a non-fah gromacs executable to test using common input files instead?

  5. #5
    Folding Flunkie Webby's Avatar
    Join Date
    Sep 2007
    Location
    Wiltshire
    Posts
    2,323
    Thanks
    12
    Thanked
    245 times in 229 posts
    • Webby's system
      • Motherboard:
      • Gigabyte G33M-DS2R, Swiftech MCW30 Northbridge Cooler
      • CPU:
      • Intel Core 2 Duo E6550 @ 3.5GHz, Cooling D-Tek Fuzion V2
      • Memory:
      • 2GB OCZ Flex DDR2 PC2-9200 5-5-5-15 @ 1000MHz 4-4-4-12
      • Storage:
      • 2x 250GB WD SataII
      • Graphics card(s):
      • Sapphire HD4870 512MB, Cooling Swiftech MCW60
      • PSU:
      • Corsair HX520w
      • Case:
      • Silverstone Tremjin TJ06 - Modded for Water Cooling Goodness
      • Operating System:
      • Windows XP Pro SP3
      • Monitor(s):
      • 22" Widescreen Cibox C2201 (with DVI input)
      • Internet:
      • 8Mb/s ADSL

    Re: Windows and Linux benching

    The f@H FAQ states that the are no performance gains to be had by running Linux or Windows at least for the standard CPU client and I assume the GPU client as well (which incidentally only runs one type of unit ever!) However this does not include the SMP client where we have seen big differences on different systems (mainly mine and Hex's) where Linux shaves off ~3.5 hours at similar clock speeds (2.2GHz to 2.33GHz), but saying that my speeds compared to Agent's (posted in another thread) are about 4 hours faster at 2.8GHz (Me) and 3GHz (Agent) so maybe its more a what else is running in the back ground rather than purely OS dependent.

  6. #6
    HEXUS.social member Agent's Avatar
    Join Date
    Jul 2003
    Location
    Internet
    Posts
    19,168
    Thanks
    735
    Thanked
    1,607 times in 1,045 posts

    Re: Windows and Linux benching

    Quote Originally Posted by directhex View Post
    what happens if on machine A i have access to 78, and on machine B i have access to 81? I can't just mix the input files, as they're incompatible. and only the master fah executable determines which cores to download when
    As I understand it, isn't the work unit tied to a core type?
    If machine B had a WU put on it that needed 78 and not 81, I thought it would try and download the 78 client or refuse to crunch it?
    Quote Originally Posted by Saracen View Post
    And by trying to force me to like small pants, they've alienated me.

  7. #7
    Comfortably Numb directhex's Avatar
    Join Date
    Jul 2003
    Location
    /dev/urandom
    Posts
    17,074
    Thanks
    228
    Thanked
    1,027 times in 678 posts
    • directhex's system
      • Motherboard:
      • Asus ROG Strix B550-I Gaming
      • CPU:
      • Ryzen 5900x
      • Memory:
      • 64GB G.Skill Trident Z RGB
      • Storage:
      • 2TB Seagate Firecuda 520
      • Graphics card(s):
      • EVGA GeForce RTX 3080 XC3 Ultra
      • PSU:
      • EVGA SuperNOVA 850W G3
      • Case:
      • NZXT H210i
      • Operating System:
      • Ubuntu 20.04, Windows 10
      • Monitor(s):
      • LG 34GN850
      • Internet:
      • FIOS

    Re: Windows and Linux benching

    Quote Originally Posted by Agent View Post
    As I understand it, isn't the work unit tied to a core type?
    If machine B had a WU put on it that needed 78 and not 81, I thought it would try and download the 78 client or refuse to crunch it?
    indeed

    but that's the problem - how can you benchmark it, when assignment of projects is random? you could be waiting for weeks for the same WU to happen on both OSes

  8. #8
    HEXUS.social member Agent's Avatar
    Join Date
    Jul 2003
    Location
    Internet
    Posts
    19,168
    Thanks
    735
    Thanked
    1,607 times in 1,045 posts

    Re: Windows and Linux benching

    I was hoping to just copy the WU from one machine to the other?

    Surely the data that is sent to the client to crunch is the same, provided its not going across the CPU/GPU/PS3 boundary's? Or are the WU's different for every OS and architecture they are running on?
    Quote Originally Posted by Saracen View Post
    And by trying to force me to like small pants, they've alienated me.

  9. #9
    Comfortably Numb directhex's Avatar
    Join Date
    Jul 2003
    Location
    /dev/urandom
    Posts
    17,074
    Thanks
    228
    Thanked
    1,027 times in 678 posts
    • directhex's system
      • Motherboard:
      • Asus ROG Strix B550-I Gaming
      • CPU:
      • Ryzen 5900x
      • Memory:
      • 64GB G.Skill Trident Z RGB
      • Storage:
      • 2TB Seagate Firecuda 520
      • Graphics card(s):
      • EVGA GeForce RTX 3080 XC3 Ultra
      • PSU:
      • EVGA SuperNOVA 850W G3
      • Case:
      • NZXT H210i
      • Operating System:
      • Ubuntu 20.04, Windows 10
      • Monitor(s):
      • LG 34GN850
      • Internet:
      • FIOS

    Re: Windows and Linux benching

    Quote Originally Posted by Agent View Post
    I was hoping to just copy the WU from one machine to the other?

    Surely the data that is sent to the client to crunch is the same, provided its not going across the CPU/GPU/PS3 boundary's? Or are the WU's different for every OS and architecture they are running on?
    it might work. worth a shot. as long as you don't mind not using your pc for a day whilst it works on a WU

  10. #10
    HEXUS.social member Agent's Avatar
    Join Date
    Jul 2003
    Location
    Internet
    Posts
    19,168
    Thanks
    735
    Thanked
    1,607 times in 1,045 posts

    Re: Windows and Linux benching

    Aye, that's the idea - set them both up and leave them on till completion with no user interaction on the machine

    Cheers for the help
    Quote Originally Posted by Saracen View Post
    And by trying to force me to like small pants, they've alienated me.

  11. #11

Thread Information

Users Browsing this Thread

There are currently 1 users browsing this thread. (0 members and 1 guests)

Similar Threads

  1. Windows and Linux on separate drives?
    By Ricardo in forum Software
    Replies: 21
    Last Post: 10-01-2007, 01:19 PM
  2. GRUB, Linux, and reinstalling windows
    By Agent in forum Software
    Replies: 20
    Last Post: 20-02-2006, 08:20 PM
  3. Run Windows software on Linspire desktop Linux
    By Bob Crabtree in forum HEXUS News
    Replies: 0
    Last Post: 09-02-2006, 01:16 AM
  4. Replies: 0
    Last Post: 24-12-2005, 12:58 PM
  5. audio and video problems, and windows app in linux
    By fedoracore in forum Software
    Replies: 0
    Last Post: 28-01-2004, 05:22 PM

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •