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.

90nm Athlon 64s - anyway to unlock them?

Discussion in 'AMD Overclocking' started by Asfand Yar Qazi, Apr 13, 2005.

  1. Hi,

    Is there a pencil trick or a copper-wire-round-the-pins trick to
    multiplier unlock the Athlon 64s? The overclocking potential of a new
    90nm Athlon64 3000+ is amazing, but to get the most out of it, it'd be
    nice to unlock it.

    I'm talking about the results of this article, among others:

    Asfand Yar
    Asfand Yar Qazi, Apr 13, 2005
    1. Advertisements

  2. No. And you don't need to either if you've got a board that can handle a
    high HTT speed.

    Michael Brown, Apr 13, 2005
    1. Advertisements

  3. Asfand Yar Qazi

    Wes Newell Guest

    There's really no need to unlock the K8 cpus' as long as your MB allows
    you to manually set the ram speed, adjust the HT link multipler, set
    core voltage, and adjust the FSB speed. On my old Jetway S755MAX board
    it's a snap to take my 3000+ from the default 2000MHz to 2330MHz by
    lowering the base ram bus to 133MHz, lowering the HT link multiplier to
    3x, and raisng the FSB clock to 233MHz (unfortunately the max setting for
    this board). It did take 1.65v to get it there though. Newer cores should
    do better. A MB with a PCI lock will probably be needed to get the FSB
    clock much over 233MHz, or at least a 7 divider to keep the PCI bus at a
    reasonable speed. Most newer boards with newer chipsets do have a PCI lock.
    Wes Newell, Apr 13, 2005
  4. You lose me here I'm afraid Wes. I've overclocked my A64 3500+ by
    dropping my multiplier to 10, and running my FSB at 240MHz. And a small
    Vcore increase too. That's on an Asus A8V Deluxe.

    Should I be doing something with the HyperTransport settings too?


    John Whitworth, Apr 13, 2005
  5. Asfand Yar Qazi

    Wes Newell Guest

    Not if it works at that speed. Or maybe your board sets it automatically.
    or defaults to something less than a 5x multiplier for it. 5x 240=1200MHz.
    Default is 1000MHz (5x200). If you get IO errors the high speed may be the
    reason. With the ram on a seperate bus now, the HT link FSB is real
    overkill. You wouldn't notice much difference if it was only running
    240MHz (1x).
    Wes Newell, Apr 13, 2005
  6. Thanks Wes.

    John Whitworth, Apr 13, 2005
  7. Asfand Yar Qazi

    gerry Guest

    wow that's a lot of work for just 330mhz cpu speed. Neat trick
    though. What kind of improvement are you seeing ? I'm wondering if
    you don't rob peter to pay paul here.
    gerry, Apr 13, 2005
  8. Yeah ... but it's all about fun and achievement. :)

    A 330MHz increase would likely be a couple of hundred pounds if you
    bought the chip rated at that speed as stock!

    John Whitworth, Apr 13, 2005
  9. Asfand Yar Qazi

    Wes Newell Guest

    I don't know what you mean by a lot of work. It only took a couple of
    minutes to do, including reboot time. I didn't measure the improvement. I
    didn't even leave it at that speed. I only did it because all the
    reviewers and everybody was saying you couldn't overclock more than about
    215Mhz without a board with a PCI lock. I don't need that speed. Most of
    the time I'm running the cpu at 800MHz with powernow.
    Wes Newell, Apr 14, 2005
  10. Asfand Yar Qazi

    cynic Guest

    Cranking down one side of a system to pump up another suggests the
    real gain may not be anything more than a number on paper. Benchmarks
    before and after tell you if there's any real improvement in computer

    You can get cpu/chipsets up to boot up at pretty high speeds, but most
    people judge whether it's reasonable by stability tests.

    No tests no glory.
    cynic, Apr 17, 2005
  11. Asfand Yar Qazi

    Wes Newell Guest

    There's only 2 sides to the cpu speed, the multiplier and the fsb clock. I
    didn't crank down either of these. The miltiplier was left at 10x and the
    fsb was cranked up to 233MHz from the default 200. The cpu nows runs at
    2330Mhz, which is approximately 17% faster any way you look at it. The
    only thing that went down was the HT link speed, from 800MHz to 699MHz,
    and even if it had dropped to 200MHz, it wouldn't have mattered. The ram
    bus speed stayed the same, but that wouldn't have mattered much either if
    it had dropped, so I don't know what you are ranting about. I don't need
    benchmarks to tell me the different in power between 2000MHz and 2330MHz
    and the same data bandwidth on the ram bus. It's there, and it's real.
    Wes Newell, Apr 17, 2005
  12. Asfand Yar Qazi

    cynic Guest

    uh.. yeah, sure.... (moving away slowly) riiiight.
    cynic, Apr 18, 2005
  13. Jumping around the thread a bit, sorry ...

    At one point in time, cynic wrote:

    And in his first post, Wes Newell wrote:

    I think what he was referring to was the lowered RAM speed. I'm not sure
    what your RAM can handle, but it's only running at 146MHz now. Depending on
    the application and your original RAM speed, this could be hurting things a
    Michael Brown, Apr 18, 2005
  14. Asfand Yar Qazi

    Wes Newell Guest

    Before overclocking, the ram speed was set Byspd at 166Mhz. After
    lowering it to 133 manually and raising the FSB clock to 233Mhz, the
    resulting ram speed is back to 166MHz so there was no difference. Even if
    there was a difference it wouldn't have mattered much at all in terms of
    performance. 90% of data request from the cpu are handled by the cache,
    not the base ram. So with the ram even at 83Mhz, it's only 50% slower 10%
    of the time. Slow your ram speed down and note the difference. Now
    disabled both caches and note the difference and you'll see what I'm
    talking about.:)
    If one is still worried about the ram speed, and wanted to run it at
    200Mhz, I could have left it at 166Mhz and then raised the FSB to 233Mhz
    and then the ram would jump to 200MHz. But I say to lower it for OP
    benefit otherwise they'll try and leave it at 200 Mhz, raise the FSB, and
    end up not booting because of overclocking the ram to much and not know
    how to fix it. This stuff is pretty simplem, but there are lot's of people
    that are less than simple.:)
    Wes Newell, Apr 18, 2005
  15. Actually, both you and I are incorrect :) Previously, you had a memory
    divider of ceil(10/(5/6)) = 12, so your memory was at 166MHz. Afterwards,
    you had a divider of ceil(10/(2/3)) = 15 giving you 155MHz. My program gave
    ceil(10/(2/3)) = 16 due to roundoff error, so it looks like I've got a bug
    to fix! So there's a minor decrease; not as bad as I initially thought.
    Regardless of where you pulled that 90% figure from, it's not that simple.
    Consider a simple bit of code that (using traditional integer x86) just sums
    up an array of 32-bit numbers. For a given cacheline, there will be one
    cache miss and 15 cache hits. The cache hit instructions will be very cheap;
    1 cycle per instruction throughput. The cache miss will be very expensive.
    Say you have a 2GHz machine with 200MHz single-channel RAM. Ignoring latency
    (assuming the prefetcher is doing its job correctly), it takes 4 RAM bus
    cycles to fetch a cache line. This equates to 40 cycles. The cache-hitting
    instructions will be inside this, so your cache miss instruction will take
    25 cycles (give or take a few). Changing to a 100MHz bus speed, you require
    80 cycles per cacheline, with a cacheline miss time of 65 cycles. Overall
    performance will still halve (bandwidth limited), despite having a cache hit
    rate of 15/16 = 94%. That's why you can't extrapolate from the cache hit
    ratio to get the expected impact on performance from slower RAM.

    Actual performance change will depend heavily on cache size (obvious),
    processor speed (higher speed processor => more impact), application. For
    example, something that is simply manipulating around 50k vertices (around
    780KB), will be heavily dependent on RAM speed with 256KB cache, moderately
    dependent with 512KB, but more or less independent of RAM speed with 1MB
    cache (at which point it would all fit inside). The same application would
    be very memory dependent in all cases at 100k vertices, and pretty much
    independent in all cases at 10k vertices.
    I'd try, but none of the boards I have support running memory async. None
    support multipliers well either, so dropping the FSB and pumping up the
    multiplier doesn't work either. You, however, are in an ideal situation to
    try it out and report back the result :)

    In this case it'd be 194MHz ... close enough not to worry about any
    performance difference

    Michael Brown, Apr 19, 2005
  16. Asfand Yar Qazi

    Wes Newell Guest

    I have no clue what you are talking about. The 33MHz added to the FSB
    clock also added 33Mhz to the ram clock and it was running at 166MHz,
    not 150 something. i don't have any memory divider or ratio settings in
    this bios, just base speeds and bySpd (which defaults to 166MHz). If I set
    the base speed manually to 133MHz and then raise the FSB clock from 200 to
    233. The ram clock goes up to 166Mhz again, and that's the speed that is
    reported during post. If I raised the FSB to 212MHz, it then reports the
    ram bus at 145MHz.
    I pulled the 90% from a sudy done on such things. And it's never that
    simple. But it is an overall average.
    I can't remember any board within the last 5 years that you couldn't
    manually set the ram speed either as a flat base speed, or by a ratio.
    What boards do you have where this isn't a bios/jumper option?
    I first tried this on old super7 boards. Just lowering the ram speed
    doesn't slow it down noticably. However, disabling cache slowed it down by
    a factor of about 4. IOW's 40 seconds to boot comapred to 10 with cache
    Well mine shows it at 200MHz on post. I don't know where you're coming up
    with your numbers.
    Wes Newell, Apr 19, 2005
  17. You missed the point of what I was saying (which followed the brief quoted
    comment). A cache miss is more than an order of magnitude slower than a
    cache hit. So in many cases, despite cache misses making up only 10% of the
    instructions, they take up a much larger percentage of the execution time.
    Using the numbers from my previous post, with a cache miss costing 40 cycles
    at 200MHz and 80 cycles at 100MHz, compared to a cache hit's single cycle
    .... an "average case" sequence of operations would take 9*1 + 40 = 49 cycles
    at 200 MHz, and 9 * 1 + 80 = 89 cycles at 100MHz. A decrease in performance
    of 45%. Of course, it's not this bad in most cases (due to hardware
    prefetching, better hit rate, and slower-executing instructions), so the
    slowdown is less.
    Soltek 75DRV5 (has the BIOS setting but won't POST if you have anything but
    1:1 ratio). MSI K7D Master-L (has a BIOS setting called "Auto Detect
    DIMM/PCI CLK" with the options enabled and disabled; doesn't appear to do
    anything). Asus A7M266-D (no option in BIOS). These two are not surprising
    given that the AMD762 doesn't support non-1:1 memory. Err, and a Sun Ultra
    2, but I don't think that counts :)

    Report it as a bug to your board manufacturer. Any half-way decent article
    on how to overclock the K8 has an explanation of memory dividers, and you
    can find a short writeup in the readme in:
    But, err, I need to fix the roundoff error in this program, so check it's
    results by hand before you quote it like I did :)
    Michael Brown, Apr 19, 2005
  18. Asfand Yar Qazi

    Wes Newell Guest

    The manual for the Soltak baord shows a Dram clock setting in the bios,
    but it defaults to BySPD according to th manual and it doesn't mention
    ratio settings. It doesn't show the other options, but being a clock
    setting, I'd think it would let you set the clock speed, rather than
    ratio to FSB. What are the options here (manual doesn't say.)?
    Wes Newell, Apr 19, 2005
  19. Asfand Yar Qazi

    Wes Newell Guest

    I'm not sure it is or isn't at this point.
    Wes Newell, Apr 19, 2005
  20. Well, the board is in a server box so I can't reboot it at the moment, but
    by memory it had a +33MHz setting at 133MHz FSB and a +33Mhz and +66MHz
    setting at 100MHz. The "BySPD" setting, as far as I can recall, makes it run
    at the same speed as the FSB (tried a 133MHz FSB CPU with PC2100 and 2700
    RAM, both ran at 133MHz when set to "BySPD"). The board doesn't POST at a
    166MHz FSB so I'm not sure what options are available at that point. So
    basically, you can select from DRAM speeds of 100/133/166MHz but the DRAM
    speed must be at least as fast as your FSB. Except like I said, if you
    select anything but a 1:1 ratio (ie: anything byt BySPD), then it's CMOS
    reset time. Yeah, I wouldn't really recommend anyone this board for FSB
    overclocking :) though it is very stable when the FSB is under about 150MHz.

    FWIW, I don't know of any board that lets you set a particular DRAM speed.
    All of them must run at some simple ratio to the FSB. Boards that do have
    200MHz, 166MHz, 133MHz, etc (or +33MHz etc) are just hiding the underlying
    ratio of x:6 RAM:FSB.
    Then read some more. Particularily this thread:
    (oscilliscope testing) which comes up if you type "A64 memory dividers" into
    google and hit "I'm feeling lucky".
    Michael Brown, Apr 19, 2005
    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.