1. This forum section is a read-only archive which contains old newsgroup posts. If you wish to post a query, please do so in one of our main forum sections (here). This way you will get a faster, better response from the members on Motherboard Point.

One RAM stick tests fine at 1T cmd but 2 do not ?

Discussion in 'AMD Overclocking' started by MHz Tweaker, Mar 5, 2006.

  1. MHz Tweaker

    MHz Tweaker Guest

    As I understand it, the more memory sticks inserted, the more stress
    that is placed on the memory controller.

    Running Epox 9NDA3+ with a pair of OCZ 1gig Gold XTC PC4000 sticks

    Each module tests fine in Memtest 1.65 in either socket 1 or 2 by
    itself with the 1T command setting. When both sticks are in sockets 1
    & 2 then Memtest test#8 always fails.

    Would this be an issue with ?

    1 - A64 3200+ Venice memory controller
    2 - Epox 9nda3+ motherboard
    3 - Some other memory timing other than the 3-4-4-8 settings

    BTW, VDimm is 2.8 VCore is at 1.55 HTT is at 260

    Are the Opteron 148 or 175 more capable of the 1T setting?


    thanks
     
    MHz Tweaker, Mar 5, 2006
    #1
    1. Advertisements

  2. MHz Tweaker

    Bill Guest

    If I'm reading it right, the HTT speed is likely the problem - you can't
    overclock the HyperTransport link very much. You need to lower the HTT
    multiplier to keep the link speed at or below 2000.

    With my x2 3800+ CPU running at 2500 MHz, the clock speed is 250 MHz.
    That means the HTT can not be at the default x5 setting. At best it must
    be lowered to x4. It works like this - clock speed of 250 x2 (for the
    double link) x4 HTT = 2000.

    And since the HTT is so efficient, dropping it further to x3 for a speed
    of 1500 has no effect on memory bandwidth at all. In fact, dropping it
    down to x2 has little effect as well.

    I'm running my HTT at x3 so my OCZ DDR 400 memory can run at 2-2-2-5 1T
    timings at DDR 416. I set the memory for 333 to get the split, and it
    all works perfectly.
     
    Bill, Mar 5, 2006
    #2
    1. Advertisements

  3. MHz Tweaker

    MHz Tweaker Guest


    thanks for the reply

    Yes you're right about the HTL speed..... I had it set at /4 which
    gave me 1040 x 2 = 2080. It does not seem like much of an overclock
    but it sure affects stability here. I went down to /3 and things are
    stable again. Will try the 1T command this evening after I finish what
    I'm working on now.

    I will try the OCZ at 333 (166 setting) and see if I make it to 225
    with the 1T setting and do some benchmarks to see if I get better mem
    transfer speeds at 2T and 260MHz or 1T and 225MHz.

    I remember now reading about how little difference the HTL divider
    made but somehow it slipped my mind..... thanks for jogging it ;-)

    BTW, what mobo are you using to get the 1T rate? I want to grab an
    Opty 175 in a couple weeks once my coins are in order.


    thanks again
     
    MHz Tweaker, Mar 5, 2006
    #3
  4. MHz Tweaker

    Bill Guest

    Most likely the benchmarks will show 2T at 260 to be the faster setup.
    The command rate doesn't make a dramatic difference, although it helps
    to keep everything running as fast as possible.
    I'm using the Asus A8N-E with two memory modules. It doesn't like the 1T
    rate with four memory modules.
     
    Bill, Mar 6, 2006
    #4
  5. MHz Tweaker

    Ed Light Guest

    Apparently none of them do. A fact of A64.


    --
    Ed Light

    Smiley :-/
    MS Smiley :-\

    Send spam to the FTC at

    Thanks, robots.

    Bring the Troops Home:
    http://bringthemhomenow.org

    Fight Spam:
    http://bluesecurity.com
     
    Ed Light, Mar 6, 2006
    #5
  6. MHz Tweaker

    MHz Tweaker Guest

    I may have some other issues in this machine. I'm using dual power
    supplies. One FSP 530-60GNA 530watt unit to power the Epox 9NDA3+
    motherboard, ATI 9800XT and WD Raptor boot drive. I also have a PC
    Power and Cooling 600watt unit that powers 8 other hard drives and a
    DVD burner, I/O panel, and 6 fans. The FSP is the primary unit and
    turns on the PC Power unit via a ground and the Green power on
    connector wire. I have read about possible feedback on some of the
    rails if the both units power components that are connected to a
    single motherboard. I have read that it would not be an issue say for
    instance if the 2nd power supply just powered fans, lights or a water
    pump. In other words components that are connected to the mobo and
    powered by a single supply is OK, but when two different supply's are
    used, then some type of voltage regulation circuitry might cause them
    to fight each other in attempts to raise or lower voltage on various
    rails. I'm not an electrical engineer, so I haven't a clue if this is
    possible.

    According to my tests, the bandwidth difference between 1T and 2T with
    a single stick is like 2875MB a sec for 2T and over 3300MB a sec for
    1T. This checks out with my Sandra benches in dual mode at 5700+MB a
    second with 2T command rate. So I would guestimate I could gain 800MB
    a second bandwidth by being able to use 1T command rate.

    At any rate, the best I can do is about 260HTT stable at 2T cmd rate.
    The board works fine with a single stick in any slot at 1T at 260HTT,
    but not with two 1gig sticks in dual mode.

    I'm a bit worried about the power supply issue and plan on ordering a
    single PC Power and Cooling 850 or 1000 watt unit as soon as I can
    cough up the roughly $500 for the thing.

    I guess I'll just live with what I have for now but like addictive
    hobby, I want just a little more out of what I have ;-)
     
    MHz Tweaker, Mar 7, 2006
    #6
  7. MHz Tweaker

    Ed Light Guest

    I don't know about the psu issue.

    I think in real life the difference between 1t and 2t may not be noticeable,
    all other things being equal. You could try benchmarks both ways and see.
    --
    Ed Light

    Smiley :-/
    MS Smiley :-\

    Send spam to the FTC at

    Thanks, robots.

    Bring the Troops Home:
    http://bringthemhomenow.org

    Fight Spam:
    http://bluesecurity.com
     
    Ed Light, Mar 7, 2006
    #7
  8. MHz Tweaker

    TonyC Guest

    Apparently the latest DFI 939 board does support 4x DIMMS at 1T, and I don't
    doubt lots of folks are running with these settings on various boards.
     
    TonyC, Mar 8, 2006
    #8
  9. MHz Tweaker

    Ed Light Guest

    Could be. I've seen in posts and articles many times the inability to do it.


    --
    Ed Light

    Smiley :-/
    MS Smiley :-\

    Send spam to the FTC at

    Thanks, robots.

    Bring the Troops Home:
    http://bringthemhomenow.org

    Fight Spam:
    http://bluesecurity.com
     
    Ed Light, Mar 9, 2006
    #9
    1. Advertisements

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.