Results 1 to 18 of 18

Thread: More CPU or More Ghz?

  1. #1
    Tester
    Nania Francesco's Avatar
    Join Date
    May 2008
    Location
    Italy
    Posts
    1,565
    Thanks
    220
    Thanked 146 Times in 83 Posts

    More CPU or More Ghz?

    For you the best choise of the future is More CPU or More Ghz Frequency?
    multi-threading or single-threading?

  2. #2
    Administrator Shelwien's Avatar
    Join Date
    May 2008
    Location
    Kharkov, Ukraine
    Posts
    3,134
    Thanks
    179
    Thanked 921 Times in 469 Posts
    The bottleneck is memory and storage since long time ago.
    Also a smart CPU would work faster than a dumb one, given the same clock.
    But increasing the freq is still the best way to gain more performance without software changes.
    And anyway, the choice is not up to us, and its parallel computing

  3. #3
    Member m^2's Avatar
    Join Date
    Sep 2008
    Location
    Ślůnsk, PL
    Posts
    1,612
    Thanks
    30
    Thanked 65 Times in 47 Posts
    Cores. Definitely. Great majority of programs that need CPU power is multithreaded already and thing will only be better.

  4. #4
    Programmer osmanturan's Avatar
    Join Date
    May 2008
    Location
    Mersin, Turkiye
    Posts
    651
    Thanks
    0
    Thanked 0 Times in 0 Posts
    If it would be up to me, I would vote for both of them: faster & multi-core. But, practical implementation is bounded by some rules. Let's have a look some basic electronics formula:

    P = C ? V^2 ? F

    This formula gives us power consumption (P) for a chip. C is the capacitance which is changed proportional to number of transistor's input changes per clock cycle. V is voltage and F is the chip frequency.

    As you see, both transistors count and running frequency are going to higher level day by day. As a result power consumption is being increased. So, actually we stuck in multi-core CPUs.

    Does anyone know Intel's cancellation about making 7 GHz CPUs prior to multi-core trend?
    BIT Archiver homepage: www.osmanturan.com

  5. #5
    Member m^2's Avatar
    Join Date
    Sep 2008
    Location
    Ślůnsk, PL
    Posts
    1,612
    Thanks
    30
    Thanked 65 Times in 47 Posts
    When starting NetBurst Intel claimed 10 Ghz in 2010.

    And getting more cores is more cost efficient than increasing frequency.
    Though myself I'm looking a bit higher, I can't wait to see the match between L3-less quad and more regular X3 from AMD.

  6. #6
    The Founder encode's Avatar
    Join Date
    May 2006
    Location
    Moscow, Russia
    Posts
    3,954
    Thanks
    359
    Thanked 332 Times in 131 Posts
    I'd prefer more Freq. Say 4 cores with 20 GHz each. And of course a huge L1 and L2 cache... And a faster memory and HDDs...

  7. #7
    Tester
    Nania Francesco's Avatar
    Join Date
    May 2008
    Location
    Italy
    Posts
    1,565
    Thanks
    220
    Thanked 146 Times in 83 Posts

    Hi! Multi-threading programmation is the solution !

    Multi-threading programmation in C++ is very difficult or not ? TBB of Intel is the definitive solution or OpenMP ?

  8. #8
    Tester
    Black_Fox's Avatar
    Join Date
    May 2008
    Location
    [CZE] Czechia
    Posts
    471
    Thanks
    26
    Thanked 9 Times in 8 Posts
    I heard on the lesson today that in the coming new C++ norm there will be added support for (multi)threading, it may help
    I am... Black_Fox... my discontinued benchmark
    "No one involved in computers would ever say that a certain amount of memory is enough for all time? I keep bumping into that silly quotation attributed to me that says 640K of memory is enough. There's never a citation; the quotation just floats like a rumor, repeated again and again." -- Bill Gates

  9. #9
    Member m^2's Avatar
    Join Date
    Sep 2008
    Location
    Ślůnsk, PL
    Posts
    1,612
    Thanks
    30
    Thanked 65 Times in 47 Posts
    OpenMP is so popular that I don't really see the need for expanding C++ standard...

  10. #10
    Member
    Join Date
    Sep 2007
    Location
    Denmark
    Posts
    856
    Thanks
    45
    Thanked 104 Times in 82 Posts
    More cores as i multithread a lot.

    i would just wish there was more multithreaded aware applications out there

    e.g. packjpg could easily decompress several files at the same time for an easy multithreading capatibility.

  11. #11
    Member m^2's Avatar
    Join Date
    Sep 2008
    Location
    Ślůnsk, PL
    Posts
    1,612
    Thanks
    30
    Thanked 65 Times in 47 Posts
    Quote Originally Posted by SvenBent View Post
    More cores as i multithread a lot.

    i would just wish there was more multithreaded aware applications out there

    e.g. packjpg could easily decompress several files at the same time for an easy multithreading capatibility.
    Run multiple instances at once.

  12. #12
    Member
    Join Date
    Sep 2007
    Location
    Denmark
    Posts
    856
    Thanks
    45
    Thanked 104 Times in 82 Posts
    Quote Originally Posted by m^2 View Post
    Run multiple instances at once.
    Hard to do in a automatic way runnig through a folder

  13. #13
    Member m^2's Avatar
    Join Date
    Sep 2008
    Location
    Ślůnsk, PL
    Posts
    1,612
    Thanks
    30
    Thanked 65 Times in 47 Posts
    Quote Originally Posted by SvenBent View Post
    Hard to do in a automatic way runnig through a folder
    This part is easy. for /f %%a in dir do start cmd /c "%~dp0" %%a.
    The hard part is implementing a process pool. You'd need to use SysInternals PsList and periodically check number of cmd.exe processes in your process tree. You need to identify own pid, but it's the cmd.exe instance, which started PsList. Ultimately you can randomly change PsList.exe name to get it unique.

    Overhead of this method will be big, but still worth using extra core(s).

    I was never desperate enough to create it, I run another instance for every file in the process tree. And use this carefully.

  14. #14
    Member
    Join Date
    Jun 2008
    Posts
    26
    Thanks
    0
    Thanked 0 Times in 0 Posts
    Id vote for more Ghz. Its obviously the hardest to do, so more would be learned trying to achieve it.

    I have i7 and the limiting factor is the memory/controller. this will change when we see 1.35v ddr3 benched.

    just remember the whole i7 is produced using same tech. but they spec mem controller at 1.65v. and cpu core's at 1.2v. That alone screams out dis-parity. but we will see in the next 6 months i guess.

    slight distraction from topic :

    In fact we may even see something in march when they change stepping to D0 on i7 920. Personally I think intel did not anticipate ddr3 getting faster and lower voltage so quickly. And as ppl are only really buying 920 (mine is 920 at 4ghz = +42% O/C) they selling bugger all mass profit 965 chips.

    This will only get worse as faster and lower power memory hits market, remember 920/940 are 100c or 100watts max before the cpu steps down. 965 is 135watts. less power for mem controller = more potencial O/C.

    I think Intel will "try" to break out the gimp stick, before i7/i5 becomes more mainstream. :P

    the thing i dont understand about threaded apps is simple stuff like.
    you want to create a dvd from whatever format.
    you have a keyframe every 3-5min thats a total seperate dataset for processing?. Why not run xx threads each working on 3-5min sections of the film?
    Last edited by Tribune; 23rd February 2009 at 22:28.

  15. #15
    Member d2k's Avatar
    Join Date
    Dec 2008
    Location
    dreams
    Posts
    12
    Thanks
    0
    Thanked 0 Times in 0 Posts

    where's my gpgpu

    give me a atom.and and a big GPGPU anyday


  16. #16
    Member
    Join Date
    Sep 2007
    Location
    Denmark
    Posts
    856
    Thanks
    45
    Thanked 104 Times in 82 Posts
    Quote Originally Posted by m^2 View Post
    This part is easy. for /f %%a in dir do start cmd /c "%~dp0" %%a.

    kinda works like crazy if you have several hundreds jpeg to do

    just above 20 threads of jpeg makes my machine slow to work with (evnne if started with /idle)

    and I'm not evne using half of my 8gb ram, so its not a memory problem

  17. #17
    Member
    Join Date
    Sep 2007
    Location
    Denmark
    Posts
    856
    Thanks
    45
    Thanked 104 Times in 82 Posts
    Quote Originally Posted by Tribune View Post
    you have a keyframe every 3-5min thats a total seperate dataset for processing?. Why not run xx threads each working on 3-5min sections of the film?
    because when you encode it to something else you don't have a specific time set for a keyframe and you have to analyse the data before you know where to insert a keyframe or a delta frame (and which ind of delta frame)

  18. #18
    Member m^2's Avatar
    Join Date
    Sep 2008
    Location
    Ślůnsk, PL
    Posts
    1,612
    Thanks
    30
    Thanked 65 Times in 47 Posts
    Quote Originally Posted by m^2 View Post
    And use this carefully.



    Another option would be first splitting the data into n parts, then running 1 instance for each part. Much simpler than a queue, but in most cases it will split the job unevenly.

Similar Threads

  1. RZM doesn't like to share cpu
    By SvenBent in forum Data Compression
    Replies: 2
    Last Post: 19th July 2008, 22:35

Posting Permissions

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