CPU mhz vs DDR mhz

Discussion in 'Overclocking, Cooling & Modding' started by No.7, Apr 26, 2006.

  1. No.7

    No.7 Newbie

    Hello.

    I have a question regarding overclocking in regards to Cpu mhz and DDR mhz.

    I have Asus A8n32sli nforce 4 mobo and an athlon 64 3500+ (2.2ghz) cpu together with 2*1gb corsair 3200 cas2 memory.

    Now, when overclocking CPU i can reach stable 2.75ghz with e.g 250*11. (Max cpu fsb that I can run stable is 330mhz.)

    Doing this with DDR mem set to auto puts the DDR memory to run 100mhz instead of 200mhz as they are SPD'd to.

    Setting CPU to auto and overclocking memory I can easily reach a stable 233mhz with cas3 and also boots to windows in 250mhz but not stable. (Asus mobo set mhz in intervals, ...200, 216, 233, 250)

    So, in theory I would like to run cpu at 2.75ghz and memory at 233mhz.

    But this fails, completely.

    When overclocking both parts, I can never go above 222mhz with the cpu while keeping the memory at 216 or 233.

    Why can I not go higher with CPU fsb?

    What is the limiting factor and what can I do to try and tweak myself higher?

    My max multiplier is 11 on the CPU and I would like to run at 2.75ghz or as close to it as possible while keeping the memory at at least 200mhz.

    The best compromise I have right now is running the cpu at 222mhz*11 and memory at 233mhz.

    Br,
    7
     
  2. Papercut

    Papercut Newbie

    I would suggest...lower your multiplier to 10x, set the HTT to 275, and RAM frequency to 166MHz (DDR333). That way you'll end up with a CPU clock of 2.75GHz and your RAM at 229MHz (as calculated by a little program called A64memfreq)

    Edit: Hope I read your situation correctly :p
     
    Last edited: Apr 26, 2006
  3. ChampionLLY

    ChampionLLY News Writer

    read a lot of articles before, most of them seems to point out that a 1:1 FSB:RAM ratio yields the overall best performance...
     
  4. Chai

    Chai Administrator Staff Member

    Doens't really apply to A64.
     
  5. Papercut

    Papercut Newbie

    Not necessarily, especially if you are compromising on latencies to achieve 1:1. Give you an example...Samsung UCCC vs Winbond BH5 (RAM capacity aside). You may be able to get the UCCC to run 1:1 with your A64 at maybe 280-290HTT, but with 3-4-4-8 timings. Take the BH5, put in a lower divider to get it running at 230-240MHz 2-2-2-5, I guarantee you the BH5 settings will give you better performance in almost any real-world setting.

    And as Chai mentioned, A64 does not suffer a performance hit when running asynchronously with the RAM unlike Athlon XP, etc.

    I also have read evidence to show that the difference in performance between high frequency RAM and standard PC3200 isn't that great :wicked:
    http://www.madshrimps.be/?action=getarticle&number=1&artpage=1313&articID=325
    http://www.legionhardware.com/document.php?id=486
    http://www.techspot.com/reviews/hardware/corsair_xms_vs_value_select/index.shtml

    Sure, Sandra shows bigger bandwidth numbers but in the end it's only like 2-5% extra frame rates in games :shifty:

    That's why I could never justify spending too much on RAM (Value VX *cough* )and I don't particularly recommend it to others either. And when helping people with their overclocks I always tell them to drop the RAM frequency if they haven't already...there's no point keeping the RAM speed as high as possible if it's stopping you from getting 500MHz more out of your CPU. :nuts:

    Damn, that turned out to be a long post.
     
  6. werty316

    werty316 Newbie

    Yep that is true not having a 1:1 OC in a A64 rig makes no difference. Don't worry about it just keep your memory at 200MHz (400MHz effective) and over and you will be fine. You could also trying increasing the CPU voltage if you haven't done so.
     
  7. PsYkHoTiK

    PsYkHoTiK Admin nerd

    Yup. My TCCDs at 300MHz 1:1 2.5-3-3-7 is slightly slower than my UTTs at 260MHz 2-2-2-6... :mgreen: * both on 1:1 divider (1s on superpi 1mb and 1m on 32m)

    But I am bridging that gap. One thing that I do see is that theres more bandwith.:mrgreen: Slightly higher latencies (timings duh :p ) but better read and write in everest (note: its a synthetic test)
     
  8. ZuePhok

    ZuePhok Just Started

    ah, just forget about OC-ing the ram No.7

    you get negligible performance increase.

    stick with DDR400 + the tighest timing possible, and spend the rest on other components (sound card, graphics, storage capacity, booze..ahhh whatever)
     
  9. ChampionLLY

    ChampionLLY News Writer

    hmmm, thanks for the info..
    never went into such details...
    (i don't own a desktop PC)
     
  10. ZuePhok

    ZuePhok Just Started

    you mean you own an apple? or what?
     
  11. No.7

    No.7 Newbie

    Ah, great info. Thanks!

    I found the amd calc tool and also did the settings that Papercut suggested but for some reason it didn't boot with the ram at 227 or something mhz.

    It was a bit difficult to find a good combo of settings that I liked, my aim was to get the ram at around 200mhz. The second best I found was setting the ram to 287 and I rejected that.

    Instead I am now running cpu at 310mhz *8.5 which sets the ram to 202 which seems to be working ok. It gives me a cpu running at 2.6ghz and that is an OK OC for me. It is not 100% stable in for example 3dmark05. It doesnt pass my 'run all night' stability test but it runs for at least and hour and also my real world application game test is doing fine (playing BF2 online for a few hours straight). It can also handle superPI so it looks good.

    The bios for my mobo is known to have several issues, especially with memory handling so I am waiting for an update from asus before not accepting the 99% stability issue.

    Thanks again!
    7
     
  12. ChampionLLY

    ChampionLLY News Writer

    as in I only have an Acer laptop...
     
  13. Isn't raw CPU power more important than any RAM settings? :think:
     
  14. hrobinson57

    hrobinson57 Newbie

    CPU vs. RAM

    I have only built two AMD 64bit rigs so far ( a 3200+, and a 4400+X2), but I have to agree with those who say get the most out of the CPU, and just set the RAM so you run stable. I found after days of tweaking and testing that the speed of the CPU is much more important.

    Hal
     
  15. No.7

    No.7 Newbie

    So about raw cpu power, should one aim to get as high HTT(fsb) as possible?

    I have found, in sisoft, that now that I run 310mhz HTT that I get the same RAM bandwidth regardless if I run cas2 or cas3 with all other settings the same. But it is only a testprogram, but still pretty interesting.

    7
     
  16. Chai

    Chai Administrator Staff Member

    I rather run at 15x200 than 10x300... :D
     
  17. ZuePhok

    ZuePhok Just Started

    you don't need high HTT as the memory is no longer on it.
    even at 700MHz you get more than enough bandwidth for the rest of your system components.
     
  18. Papercut

    Papercut Newbie

    I prefer Super Pi over Sandra for testing RAM settings. Sandra just gives you a number, but your RAM settings will affect the calculation time in Super Pi...
     
  19. PsYkHoTiK

    PsYkHoTiK Admin nerd

    Yea, I use superpi and Everest. :mrgreen:
     
  20. No.7

    No.7 Newbie

    Ok, thanks!
     

Share This Page