Bandwidth Efficiency at 35%?

Discussion in 'IBM' started by ranmalive, Jan 12, 2004.

  1. ranmalive

    ranmalive Guest

    Hi all,

    I just bought a new computer system that has a:
    P4 2.8c Ghz CPU FSB800
    ASUS P4R800-VM Motherboard
    512 Hynix PC3200 DDR SDRAM
    Built in ATI Radeon 9200 Video w/ 128 onboard memory
    Case w/300 Power Supply

    I downloaded SiSoftware SANDRA 2004 to test the system to see how is
    it and it seems pretty good except for the memory benchmark.

    It gave the memory Bandwidth Effieciency as 34%-35%.

    This is what SANDRA'S website had to say about it.

    "Q: Why do some systems show close to 95% bandwidth efficiency and
    others less than 80%?
    A: While the code for all CPUs is heavily optimised, the effciency
    depends on chipset performance and memory settings. Only the most
    aggressive settings may yield > 80% efficiency, thus anything higher
    is a bonus."

    The reason why I'm asking this is because I'm wondering whether I have
    a faulty ram and should get it replaced? 35% effiency seems awfully

    However, SiSoftware SANDRA recommended this solution.

    "Warning W5400 - Low memory bandwidth efficiency (advanced test).
    Generally this means a setting is incorrect or set too "safe"
    resulting in too low performance.
    Fix: Check memory timings (reduce them) or check other chipset or
    memory speed settings."

    The problem is, is that I don't see anywhere in the bios or anywhere
    else in the computer that can adjust the memory speed/timings.

    The only thing I can find in the bios is "DRAM CAS Select: Slow/Fast".
    It doesn't matter which one I choose, the same benchmark is shown.

    Here is the whole report in case you wanted to read it.

    SiSoftware Sandra

    Int Buffered iSSE2 (Integer STREAM) Results Breakdown
    Assignment : 2190MB/s
    Scaling : 2196MB/s
    Addition : 2211MB/s
    Triad : 2208MB/s
    Data Item Size : 16-bytes
    Buffering Used : Yes
    Offset Displacement Used : Yes
    Bandwidth Efficiency : 34% (estimated)

    Float Buffered iSSE2 (Float STREAM) Results Breakdown
    Assignment : 2200MB/s
    Scaling : 2205MB/s
    Addition : 2217MB/s
    Triad : 2215MB/s
    Data Item Size : 16-bytes
    Buffering Used : Yes
    Offset Displacement Used : Yes
    Bandwidth Efficiency : 35% (estimated)

    Test Status
    Memory Used by Test : 191MB
    NUMA Support : No
    SMP Test : No
    Total Test Threads : 1
    SMT Test : Yes
    Number of Test SMT (per CPU) : 2
    Dynamic MP/MT Load Balance : No
    Processor Affinity : No

    Chipset 1
    Model : ASUSTeK Computer Inc ??? (5833)
    Front Side Bus Speed : 4x 200MHz (800MHz data rate)
    Width : 64-bit
    Maximum Bus Bandwidth : 6400MB/s (estimated)

    Logical/Chipset 1 Memory Banks

    MMX Technology : Yes
    SSE Technology : Yes
    SSE2 Technology : Yes
    SSE3 Technology : No
    HTT - Hyper-Threading Technology : Yes


    Please let me know if you need any more info. If anyone has any ideas
    as to how I can increase my memory settings/timings, that'd be greatly
    appreciated. If not, I'm assuming I should have the Ram replaced.

    Hope to hear from anyone soon,


    p.s. Maybe I should use a different Memory Benchmark program?
    ranmalive, Jan 12, 2004
    1. Advertisements

  2. ranmalive

    fixhunter Guest

    I have the same issue and I really want to know what's wrong.
    I have the same configuration of the hardware (P4 2.8c Ghz CPU FSB800,
    ASUS P4R800-VM Motherboard, 1x512MB DDR PC3200), so results of benchmarks
    must be the same, but arent really - my memory bandwidth is about 2077
    MB/s. I think that the only component that can make this difference is
    RAM...or something around it(maybe the shared part of RAM can also affect
    - so that can be the reason of this issue...I'll do some tests to find
    this out).
    When we are talking about bandwidth efficiency note that the peak
    bandtwidth with one channel (1x256,1x512,...) is max 3200 MB/s and with
    two channels (2x256,2x512,...) is max 6400 MB/s. So i think that sisoft
    sandra results are 35% of 6400 MB/s ...this carry your 2190MB/s
    (and mine 2077MB/s), in reality this mean that your bandwidth efficiency
    is 68% (mine 64%). But numbers stil doesn't match - more than 30% is lost.
    If we assume that in our best case there will be 95% efficency, there are
    still around 25% lost. In normal case there must be a bandtwidth from
    2700MB/s to 3100MB/s.

    Where are the others 25% ??????????
    Anyone know that????????? I go to check if the shared graphics memory affect that...i will
    post there what I'll find out...

    Anyone who know something about that, help us. Please!
    fixhunter, Oct 7, 2004
    1. Advertisements

  3. ranmalive

    Bob Willard Guest

    I stopped using SiSoft's Sandra years ago, so I won't comment on that
    package. I currently use Everest (see for such stuff.
    But Everest and Sandra share a common constraint, in that they both
    attempt to measure bandwidth from within the OS -- and the results are
    sometimes heavily influenced by the OS *and* by the other apps with
    which the benchmark app shares the OS's resources. So, for best
    results, before running any such benchmark under an OS, give the system
    a fresh boot and run as close to standalone as possible.

    As an example, I just ran the Everest memory read benchmark twice: with
    a couple of low-priority CPU-eating apps running, Everest reported
    4029 MB/s; with those CPU-eaters stopped, Everest reported 5205 MB/s.
    Bob Willard, Oct 7, 2004
    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.