Results 1 to 13 of 13
  1. #1
    Joined
    Jan 2008
    Posts
    5

    Matching base core speeds of DDR2-800 and E6600 for Folding@Home

    Haven't overclocked since using Northwood Core Pentiums.

    Forgot all about CPU FSB matching Memory Core speeds for optimal memory access.

    To improve Folding@Home results with a Intel E6600 on a Gigabyte 965P-DS3 with Kingston DDR2-800 Value Ram, I changed default base frequency of the CPU to 400 Hhz (from 266) to match DDR2-800's core speed of 400 MHz. Changed "CPU clock ratio" to 6 (from 9) to achieve same 2.4 GHz rating.
    Folding@Home performance improved.
    Eventually went to a CPU clock ratio of 7 for 2.8 GHz and sometimes 8 for 3.2 GHz to get even better performance.

    On the otherhand, I didn't see much improvement if I left CPU multiplier at 9 and merely moved the CPU base frequency to a bit above 266 (to about 320).

    This leads to the question : Have other people seen any benefits of matching FSB base speed with DD2 base speed with a Core 2 Duo ?

    Any feedback would be appreciated.

    (down side of the core speed increase is that X1300 Graphics card will not redisplay on simple reboots ... have to shutdown/power off/power on/boot to get display back ... whimpy VisonTek X1300 PCI card ?)

    Thanks !

    HerbVista

  2. #2
    Joined
    Feb 2003
    Posts
    8,739

    Re: Matching base core speeds of DDR2-800 and E6600 for Folding@Home




    ud

  3. #3
    Joined
    Aug 2004
    Posts
    10,610

    Re: Matching base core speeds of DDR2-800 and E6600 for Folding@Home

    You should bet the best results with your cpu at or about the highest frequency it can run folding stabily. This is often faster than the memory can be made to run.
    When you say
    On the otherhand, I didn't see much improvement if I left CPU multiplier at 9 and merely moved the CPU base frequency to a bit above 266 (to about 320).
    (note: 2880) in comparison to what? Why not 355x9 (3200) or higher??


    "The significant problems we face cannot be solved at the same level of thinking we were at when we created them."
    - Albert Einstein (1879-1955)

  4. #4
    Joined
    Aug 2000
    Location
    Vvardenfell
    Age
    58
    Posts
    10,924

    Re: Matching base core speeds of DDR2-800 and E6600 for Folding@Home

    Er - aren't folding units different sizes? Or should they always take the same time if all hardware stays set the same? Because if the units vary then you've proved nothing except that units vary. Can you replicate this result with any other benches? Benches at each setting must be run back-to-back to eliminate differences due to different drivers etc.


    M

  5. #5
    Joined
    Aug 2004
    Posts
    10,610

    Re: Matching base core speeds of DDR2-800 and E6600 for Folding@Home

    There are variations, but you can get a good idea from knowing times of previous same # WU's. You have to have a fair idea of what is going on as the calculation of one frame (1 ns of a 100ns WU) can change quite a bit within the WU. Kind of a fast, slow and normal ranking depending on how close the atoms are to each other at that time.

    I can't answer how bandwidth effect intels on F@H, the general rule is not much, but in a specific application I won't guess. The large caches tend to mask the problems of p3-4's.

    HerbVista, we have plenty of intel experience in the folding forum, to get the specific answer to bandwidth effect on folding, better to ask there.
    My feeling is at 3.2+ and 355+ he will get the best performance, but he hasn't pushed his hardware, and hasn't differentiated between cpu speed and bandwidth yet.

    By the way, Herb, welcome to the Forums, and check the Folding forum for your folding needs. We're glad to help out any folder, no need to join unless you want to.
    Last edited by TheGlasMan; 02-03-2008 at 09:56 AM.


    "The significant problems we face cannot be solved at the same level of thinking we were at when we created them."
    - Albert Einstein (1879-1955)

  6. #6
    Joined
    Jan 2008
    Posts
    5

    Re: Matching base core speeds of DDR2-800 and E6600 for Folding@Home

    Thanks for the replies !

    Will look into the Folding Forums for tips to improve Folding@Home performance on Intel.

    Wanted to indicate how I compared performance between Clocking settings ...
    Ran 2 gNewSense VM guests under control of Parallels on XP. Took tar snapshots of the folding progress of the Linux Folding console client. In that way I could "restart" a folding work unit at the same point and compare output logs between runs after making overclock changes ... (FAH work unit only allowed to complete once).

    Based on suggestions I am retesting clocking setting to see if edging up base CPU core speeds with constant 9x multiplier (and proper memory multiplier changes) can match changing core speed of 400 Mhz with CPU multipliers of 7 or 8.
    So far best performance over stock settings was a 20% increase of Frames/Hour using a CPU base core speed of 400 Mhz, CPU multiplier of 8, and memory multiplier of 2.
    Now retesting with CPU base core speed of 320 Mhz, CPU multiplier of 9, and memory multipler of 2.5.
    Thanks for suggestions and help.

  7. #7
    Joined
    Jan 2008
    Posts
    5

    Re: Matching base core speeds of DDR2-800 and E6600 for Folding@Home

    Followup ... tried TheGlasMan's suggestion that high CPU clock rate was more important than matching Memory Freq to CPU base FSB ...

    TheGlasMan was correct ... found that 9x355 produced as good FAH performance as 8x400 ...
    ... even though at 9x355 I had the memory multiplier at 2 to give a 710 MHz to interface the CPU with the DD2-800 memory .!

    I only have Kingston Value Ram so could not over clock memory ...

    Found that the benefit of the 9x355 approach was that I could leave the CPU voltage at default levels and therefore have CPU run cooler by about 8C !

    So many thanks for the TheGlasMan's help ( and for the others who replied as well ! )

    Keep on Folding !

    HerbVista

  8. #8
    Joined
    Jan 2004
    Posts
    3,463

    Re: Matching base core speeds of DDR2-800 and E6600 for Folding@Home

    1:1 FSB/DRAM ratios don't boost performance on recent Intel chipsets.

  9. #9
    Joined
    Aug 2000
    Location
    Vvardenfell
    Age
    58
    Posts
    10,924

    Re: Matching base core speeds of DDR2-800 and E6600 for Folding@Home

    Quote Originally Posted by oralpain View Post
    1:1 FSB/DRAM ratios don't boost performance on recent Intel chipsets.


    Or indeed any chipset except nForce2.


    M

  10. #10
    Joined
    Jan 2008
    Posts
    5

    Re: Matching base core speeds of DDR2-800 and E6600 for Folding@Home

    Guess from the feedback that 1:1 for FSB to Memory base frequency matching no longer matters ... not just for Folding@Home ... but doesn't matter in general ... with the new Northbridge chipsets with the Core 2 Duo ...

    HerbVista

  11. #11
    Joined
    Sep 2001
    Location
    Toronto, CANADA
    Posts
    1,392

    Re: Matching base core speeds of DDR2-800 and E6600 for Folding@Home

    Doesn't 1:1 matter for system stability?
    Intel E6750 C2D
    400x8 (3.2GHz) 32C idle/58C Load @ stock 1.31V/HSF | Gigabyte DS3R Rev 1 bios F13 | OCZ XTC Rev 2 @4-4-4-15 | Antec Sonata 3 500W | Palit 8800GT driver v.169.21 | WD 500GB | 22" 226BW | Logitech x-540 5.1

    AMD64 (754) 3000+ @2GHz |
    | ASUS K8N-E DLX BIOS 1.11 | 512x 2 OCZ Premier DDR400 2.5-3-3-7 | Antec Sonata 380W | 9600 Pro 128MB | 19" 930B

  12. #12
    Joined
    Aug 2000
    Location
    Vvardenfell
    Age
    58
    Posts
    10,924

    Re: Matching base core speeds of DDR2-800 and E6600 for Folding@Home

    No, except in the sense that it might be required to make the RAM run at "best speed". There is nothing magical in any way about running in sync mode, and hasn't since nForce2.



    M

  13. #13
    Joined
    Aug 2004
    Posts
    10,610

    Re: Matching base core speeds of DDR2-800 and E6600 for Folding@Home

    Actually, p4's appreciated it as well, even the 800 FSB models. The larger caches and faster FSB's of c2d's limit the amount of time that maximum memory bandwidth is necessary.
    I was very surprised to find out how slow the memory runs default on c2d's. Works well even with 4 pipelines and 2 cores.


    "The significant problems we face cannot be solved at the same level of thinking we were at when we created them."
    - Albert Einstein (1879-1955)

Posting Permissions

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