1. Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

The AMD Phenom II OC'ers Club

Discussion in 'techPowerUp! Club Forum' started by ShadowFold, Dec 22, 2008.

  1. Mussels

    Mussels Moderprator Staff Member

    Joined:
    Oct 6, 2004
    Messages:
    42,225 (11.61/day)
    Thanks Received:
    9,527
    i like finding the sweet spots where i get large gains from minimal voltage boosts. this seems to be in that range.
     
  2. JrRacinFan

    JrRacinFan Served 5k and counting ...

    Joined:
    Mar 17, 2007
    Messages:
    19,414 (7.07/day)
    Thanks Received:
    4,483
    Location:
    Youngstown, OH
    Here's where I've been keeping mine at.
    [​IMG]

    I know almost a totally different architecture. Can't wait for AMD 9 series to hit, I'm getting rather impatient LOL
     
  3. Dogshitjoint

    Dogshitjoint New Member

    Joined:
    Jul 30, 2010
    Messages:
    170 (0.11/day)
    Thanks Received:
    35
    Location:
    @/In a place where you aint ;) :)
  4. Chicken Patty

    Chicken Patty WCG Moderator Staff Member

    Joined:
    Nov 27, 2007
    Messages:
    28,341 (11.38/day)
    Thanks Received:
    12,227
    Location:
    Miami, Florida
    I'm hopefully putting a better cooler on my X6 today. If so I'll do some clocking. :). Have t overclocked in the longest.
     
  5. JrRacinFan

    JrRacinFan Served 5k and counting ...

    Joined:
    Mar 17, 2007
    Messages:
    19,414 (7.07/day)
    Thanks Received:
    4,483
    Location:
    Youngstown, OH
    Hey, Dog! Need your feedback on the 870A-UD3. Can hit me up via pm or post here.

    @CP

    If you need a refresher and some help just throw me a txt.
     
  6. Mussels

    Mussels Moderprator Staff Member

    Joined:
    Oct 6, 2004
    Messages:
    42,225 (11.61/day)
    Thanks Received:
    9,527
    just remember one tip: lower temps let you clock higher at the same volts, so dont go excessive on the voltage. tune it them all in for the sweetspot once its initially stable.
     
  7. Dogshitjoint

    Dogshitjoint New Member

    Joined:
    Jul 30, 2010
    Messages:
    170 (0.11/day)
    Thanks Received:
    35
    Location:
    @/In a place where you aint ;) :)
    Cool Jr see ya in PM m8...

    //Dog
     
    JrRacinFan says thanks.
  8. Chicken Patty

    Chicken Patty WCG Moderator Staff Member

    Joined:
    Nov 27, 2007
    Messages:
    28,341 (11.38/day)
    Thanks Received:
    12,227
    Location:
    Miami, Florida
    Thanks guys, ill certainly be a bit rusty, but we'll see what it can do. :toast:
     
  9. crunchie

    crunchie New Member

    Joined:
    Feb 22, 2006
    Messages:
    110 (0.04/day)
    Thanks Received:
    20
    Location:
    Australia
    Are you talking about the cpu-nb? Mine is at 3Ghz with 1.25v. You just increase the multi in bios.
     
  10. AthlonX2

    AthlonX2 HyperVtX™

    Joined:
    Sep 27, 2006
    Messages:
    7,179 (2.46/day)
    Thanks Received:
    1,663
    now you just need to push your ram to 2000 and you can be cool like the rest of us:cool:
     
  11. crunchie

    crunchie New Member

    Joined:
    Feb 22, 2006
    Messages:
    110 (0.04/day)
    Thanks Received:
    20
    Location:
    Australia
    Yeah, I have 4Gig Flares now running at 1800Mhz 6-8-6-22-1T and will do 2000Mhz 7-9-7-24-1T :)
     
  12. Mussels

    Mussels Moderprator Staff Member

    Joined:
    Oct 6, 2004
    Messages:
    42,225 (11.61/day)
    Thanks Received:
    9,527
    that easy?

    i tried 2.6 with 1.3v and it didnt like it, system was unstable.
    maybe its the case of me using too MUCH voltage again, i'll try 1.25v and see how it goes.


    (so far best indication is S3 sleep mode overnight, if anything at all is wrong with the ram/NB, it wont wake up from it)
     
  13. Mussels

    Mussels Moderprator Staff Member

    Joined:
    Oct 6, 2004
    Messages:
    42,225 (11.61/day)
    Thanks Received:
    9,527
    trying 2.8Ghz NB with 1.25v cpu-nb. everything else that isnt CPU or ram is on auto/'normal' volts.
     
  14. AthlonX2

    AthlonX2 HyperVtX™

    Joined:
    Sep 27, 2006
    Messages:
    7,179 (2.46/day)
    Thanks Received:
    1,663
    what helps mine out is running the cpu VDDA @ 2.55v - 2.6v
     
  15. Mussels

    Mussels Moderprator Staff Member

    Joined:
    Oct 6, 2004
    Messages:
    42,225 (11.61/day)
    Thanks Received:
    9,527
    hmmm, i dont think its called that on mine. i'll take a photo or something next reboot.
     
  16. crunchie

    crunchie New Member

    Joined:
    Feb 22, 2006
    Messages:
    110 (0.04/day)
    Thanks Received:
    20
    Location:
    Australia
    Mussels, try lower the HT back to stock 2000Mhz. Have not bothered raising mine as from what I have read around different sites there is hardly any gain to be had by raising the frequency.
     
  17. Jack Doph

    Jack Doph

    Joined:
    Jan 24, 2011
    Messages:
    508 (0.38/day)
    Thanks Received:
    109
    Location:
    Upright down-under (Brisbane, Australia)
    There is a difference in memory-intensive tasks.
    It also reduces latency.
    How much benefit can be had from going over 2600MHz NB is open to conjecture though..
     
    Velvet Wafer says thanks.
  18. Velvet Wafer

    Velvet Wafer New Member

    Joined:
    Jun 16, 2009
    Messages:
    5,123 (2.66/day)
    Thanks Received:
    990
    Location:
    North of Germany
    yeah, i also had to find out that i can hit the 1600 6-7-6 24 27 1.66v 2700NB, but only stable,
    when i used below 1.3 v on the CPU-NB.. needed 2 whole weekends to tweak that out :laugh:

    Theoretically, the speed should scale indefinetly until it hits the a bottleneck of the CPU clock, or the architecture itself.... if you keep raising the ram clocks, and the NB clocks, at cl7 and below, you will certainly run into instability after a while, but if you could keep the stability up,the memory bandwith, the speed of the IMC,and especially of the L3 cache could be increased by a very extensive amount.... probably what has been done to Bulldozers brand new IMC... i guess thats also the reason for the new socket, as the power requirements probably go beyond dual plane (triple plane maybe?)
    I guess Bulldozer wont win a prize on main clock efficiency alone, eh? ;)
     
    Last edited: Mar 16, 2011
    Jack Doph says thanks.
  19. Mussels

    Mussels Moderprator Staff Member

    Joined:
    Oct 6, 2004
    Messages:
    42,225 (11.61/day)
    Thanks Received:
    9,527
    stock is 2000? i think mine was on auto and it was running at 2600
     
  20. Jack Doph

    Jack Doph

    Joined:
    Jan 24, 2011
    Messages:
    508 (0.38/day)
    Thanks Received:
    109
    Location:
    Upright down-under (Brisbane, Australia)
    By default, all mobos run a 10x multiplier internally for the NB.
    This is to make sure the system will boot with whatever CPU is installed.

    EDIT: well, by default, all mobos *should* be running that ;)
     
  21. Mussels

    Mussels Moderprator Staff Member

    Joined:
    Oct 6, 2004
    Messages:
    42,225 (11.61/day)
    Thanks Received:
    9,527
    who knows, i coulda read it wrong. it think its at 2600NB, 2000HT 1600 CL7 ram and 3600 CPU atm.
     
  22. Jack Doph

    Jack Doph

    Joined:
    Jan 24, 2011
    Messages:
    508 (0.38/day)
    Thanks Received:
    109
    Location:
    Upright down-under (Brisbane, Australia)
    Understood.
    Naturally, mileage will vary from setup to setup, but I found 2600NB to be the sweet point - no extra voltage required for stability (thus temps stay much the same), but there's a noticeable latency reduction (in the stuff I use anyway).
    Not huge of course, but noticeable nonetheless.
    What I do know is that the raising of HT results in a net gain of zero.

    Perhaps you may find that raising your NB further *does* make a difference, but I have my doubts tbh ;)
     
  23. Velvet Wafer

    Velvet Wafer New Member

    Joined:
    Jun 16, 2009
    Messages:
    5,123 (2.66/day)
    Thanks Received:
    990
    Location:
    North of Germany
    if you cant keep 2.8ghz nb up, try to lower the memory divider to 1333 (you can clock it up later per HTT, till you reach the headroom of the IMC)... if you suddenly gain stability, then you know for sure that your IMC has not enough/too much volts, or if that isnt the case, and you tried every voltage setting with CPU loadline calibration on and off, its nearing the beginning of its wall, where it cant put out more memclocks or NB, no matter what you do;)
     
    Mussels says thanks.
  24. Jack Doph

    Jack Doph

    Joined:
    Jan 24, 2011
    Messages:
    508 (0.38/day)
    Thanks Received:
    109
    Location:
    Upright down-under (Brisbane, Australia)
    Noted ;)

    Well, in my experience, going beyond 2600NB only netted a result I could measure in benchmarking programs, rather than notice in the apps I use.
    Also, stability became an issue after 2800NB, regardless of compensation on all other settings.
    Beyond 3000NB my system simply won't boot and the difference between 2600NB - 3000NB wasn't worth the other hassles, for little to no discernible return.
    I realise this may well be a limitation on my hardware's part, which is why I stated that one's mileage may vary - this is always true ;)
     
    Velvet Wafer says thanks.
  25. Velvet Wafer

    Velvet Wafer New Member

    Joined:
    Jun 16, 2009
    Messages:
    5,123 (2.66/day)
    Thanks Received:
    990
    Location:
    North of Germany
    I guess i know, why high NB clocks sometimes dont net more recognizable speed... its the same reason, that makes procs feel sluggish when beeing overclocked with too low volts...lack of power supply.
    For example, let me tell you the story of my phenom 955 and me ;):
    At first, i used it in a Biostar 790gx a2+... the board doesnt even had support for it,no dual power plane, but strangely, displayed it correctly, and worked with it. Not good tho :D
    i had hard times reaching 3.8ghz on 2400mhz NB with ddr2 1066 back then, volts more or less high, due to the general instability of the board with the 955.
    After about 3 Months, i shot a cap in the NB area, the only one not beeing solidcap and near the PWM.
    That didnt hurt the board too bad tho (it probably lacked power all over the board anyways), as it ran exactly as before,
    just with bsoding once a day (during 24/7 crunching in the hottest part of summer...)

    When i ran the board + proc in linx, it gave me roughly 35 Gflops when the system was tested fully idle, without me tampering with music or firefox, or anything that could manipulate the maximum amount of flops.

    as it started to annoy me, i got me a Foxconn A7DA-S 3.0 790GX, which improved stability and clocking,had dual power plane, clocked to 2800NB, and on which i also made the 4.2ghz valid on the 955... neithertheless, it was bitchy about ram clocks and timings in general, and only had big voltage steps, it wasnt exactly made for OCing.
    This board gave me between 40-41 Gflops in average, at roughly the same clocks/NB

    Now, just shortly i got me the M4A89GTD PRO USB, with 890GX, that clocks ram far more stable, than the Foxconn... is suddenly stable at 3.9ghz when both boards before were not...and gives 51-52 Gflops, all out of nothing, and far beyond 40gflops even on clocks far lower then on the other board?
    This sounds like the power supply of the CPU-NB, has definetly to do with the real world benefits it gives... i also trust Linx very much in this case, as it even reacts Flopwise, when i play a simple MP3.
    Where we come to the Problem that Bulldozer probably has... pulls too much power from the power plane to be efficient in anything that is designed till today ;)
     
    Jack Doph says thanks.

Currently Active Users Viewing This Thread: 1 (0 members and 1 guest)

Share This Page