Results 1 to 10 of 10

Thread: P4 Hyper-Threading Benchmark??

  1. #1

    Post P4 Hyper-Threading Benchmark??

    Can anyone running DF with a hyper-threading-enabled P4 post a benchmark?

  2. #2
    Senior Member KWSN_Millennium2001Guy's Avatar
    Join Date
    Mar 2002
    Location
    Worked 2 years in Aliso Viejo, CA
    Posts
    205

    Re: P4 Hyper-Threading Benchmark??

    Originally posted by HaloJones
    Can anyone running DF with a hyper-threading-enabled P4 post a benchmark?
    Yes, I believe that there is nothing in the rules which would prevent them from posting.

    I hope this answers your question.

  3. #3
    Junior Member
    Join Date
    Apr 2002
    Location
    Florida
    Posts
    13

    Re: Re: P4 Hyper-Threading Benchmark??

    Originally posted by KWSN_Millennium2001Guy
    Yes, I believe that there is nothing in the rules which would prevent them from posting.

    I hope this answers your question.

    M2K+1Guy. You are so silly No wonder no one takes us serious

    Ni

  4. #4
    Junior Member
    Join Date
    Apr 2002
    Location
    The Knights Who Say Ni!
    Posts
    27
    Short version? It sucks.

    I don't have the actual numbers anymore, but when setting up a Dell PE6600 with four physical processors, I did a weekend worth of DF crunching as part of the burn in.

    Setup Info: Dell (Dude, your getting) PE6600. 4 Xeon 1.4 cpus, 4Gb RAM, bunch O' disks, Win2K Advanced Server, SQL Ent, etc.

    The four virtual processors produced hardly any results (compared to the four "real" processors) and if I remember correctly the total output while running DF across all eight processors was worse than running just four instances. There may be a "happy medium" (say 6 instances of DF in this case) that produces well, but I wouldn't bet on it...

    It kind of makes sense. Hyperthreading isn't going to be able to do much with a CPU slammer like DF. It's just not what hyperthreading was designed for. Though, on the upside, we are seeing a performance improvement in MS-SQL of around 30-35% with hyperthreading turned on...

  5. #5
    Junior Member
    Join Date
    May 2002
    Location
    Singapore
    Posts
    14
    Here's an interesting article from Sudhian worth taking a look. Though DF wasn't run in those tests, there are some hard figures and analysis on F@H and Seti@H.

  6. #6

    Re: Re: P4 Hyper-Threading Benchmark??

    Originally posted by KWSN_Millennium2001Guy
    Yes, I believe that there is nothing in the rules which would prevent them from posting.

    I hope this answers your question.



    Go back to your shrubbery you silly kniggit!

  7. #7
    25/25Mbit is nearly enough :p pointwood's Avatar
    Join Date
    Dec 2001
    Location
    Denmark
    Posts
    831
    What kind of benchmark do you want?

    I have a dual 2Ghz Xeon (512Kb cache) server, sadly only temporarily

    Right now I'm running just a single processor to see what kind of performance that gives.

    I will try to run a quick benchmark with 4 clients running too.

    I have already tried to start 4 clients and I experienced something strange - they all started normally, using ~100MB memory (using the -rt switch), but when I checked it a bit later (using Windows Terminal Server), they only used ~50-60MB each
    Pointwood
    Jabber ID: pointwood@jabber.shd.dk
    irc.arstechnica.com, #distributed

  8. #8
    25/25Mbit is nearly enough :p pointwood's Avatar
    Join Date
    Dec 2001
    Location
    Denmark
    Posts
    831
    A single client:

    ------------------------------------------------------------
    Distributed Folding Windows dfGUI v1.8 Benchmark

    Sample Size : 1240 structures over 1133 seconds.
    Protein Size: 129AA

    Structures Per Second: 1.09
    Structures Per Minute: 65.7
    Structures Per Hour : 3940
    Structures Per Day : 94560

    OS : Windows 2000 MHz: 1995
    CPU: Intel(R) XEON(TM) CPU 2.00GHz
    Client Switches: -df -rt -s 1000
    ------------------------------------------------------------

    I thought it would be faster

    4 Clients running:

    ------------------------------------------------------------
    Distributed Folding Windows dfGUI v1.8 Benchmark

    Sample Size : 1720 structures over 2849 seconds.
    Protein Size: 129AA

    Structures Per Second: 0.60
    Structures Per Minute: 36.2
    Structures Per Hour : 2173
    Structures Per Day : 52161

    OS : Windows 2000 MHz: 1995
    CPU: Intel(R) XEON(TM) CPU 2.00GHz
    Client Switches: -df -rt -s 1000
    ------------------------------------------------------------

    ------------------------------------------------------------
    Distributed Folding Windows dfGUI v1.8 Benchmark

    Sample Size : 1700 structures over 2843 seconds.
    Protein Size: 129AA

    Structures Per Second: 0.60
    Structures Per Minute: 36.0
    Structures Per Hour : 2159
    Structures Per Day : 51810

    OS : Windows 2000 MHz: 1995
    CPU: Intel(R) XEON(TM) CPU 2.00GHz
    Client Switches: -df -rt -s 1000
    ------------------------------------------------------------

    ------------------------------------------------------------
    Distributed Folding Windows dfGUI v1.8 Benchmark

    Sample Size : 1570 structures over 2846 seconds.
    Protein Size: 129AA

    Structures Per Second: 0.55
    Structures Per Minute: 33.1
    Structures Per Hour : 1986
    Structures Per Day : 47663

    OS : Windows 2000 MHz: 1995
    CPU: Intel(R) XEON(TM) CPU 2.00GHz
    Client Switches: -df -rt -s 1000
    ------------------------------------------------------------

    ------------------------------------------------------------
    Distributed Folding Windows dfGUI v1.8 Benchmark

    Sample Size : 1735 structures over 2847 seconds.
    Protein Size: 129AA

    Structures Per Second: 0.61
    Structures Per Minute: 36.6
    Structures Per Hour : 2194
    Structures Per Day : 52653

    OS : Windows 2000 MHz: 1995
    CPU: Intel(R) XEON(TM) CPU 2.00GHz
    Client Switches: -df -rt -s 1000
    ------------------------------------------------------------

    It seems that HT gives little performance boost, but it's not much (compared to the benchmark when running one client).

    I'm running 2 clients right now to see what kind of performance that gives.

    2 clients:

    Wierd - they are slow...

    ------------------------------------------------------------
    Distributed Folding Windows dfGUI v1.8 Benchmark

    Sample Size : 1350 structures over 1750 seconds.
    Protein Size: 129AA

    Structures Per Second: 0.77
    Structures Per Minute: 46.3
    Structures Per Hour : 2780
    Structures Per Day : 66728

    OS : Windows 2000 MHz: 1995
    CPU: Intel(R) XEON(TM) CPU 2.00GHz
    Client Switches: -df -rt -s 1000
    ------------------------------------------------------------


    ------------------------------------------------------------
    Distributed Folding Windows dfGUI v1.8 Benchmark

    Sample Size : 1490 structures over 1756 seconds.
    Protein Size: 129AA

    Structures Per Second: 0.85
    Structures Per Minute: 50.9
    Structures Per Hour : 3055
    Structures Per Day : 73312

    OS : Windows 2000 MHz: 1995
    CPU: Intel(R) XEON(TM) CPU 2.00GHz
    Client Switches: -df -rt -s 1000
    ------------------------------------------------------------
    Pointwood
    Jabber ID: pointwood@jabber.shd.dk
    irc.arstechnica.com, #distributed

  9. #9
    Senior Member KWSN_Millennium2001Guy's Avatar
    Join Date
    Mar 2002
    Location
    Worked 2 years in Aliso Viejo, CA
    Posts
    205
    How much memory and of what type on this dual XEON box? It looks as though you have a severe memory bottleneck or not enough RAM.

    Ni

  10. #10
    25/25Mbit is nearly enough :p pointwood's Avatar
    Join Date
    Dec 2001
    Location
    Denmark
    Posts
    831
    It's one of these: http://www.pc.ibm.com/support?page=8...e=&subtype=Cat

    But with 1GB memory.
    Pointwood
    Jabber ID: pointwood@jabber.shd.dk
    irc.arstechnica.com, #distributed

Posting Permissions

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