Copper heat spreaders for RAM: worthwhile or a waste of money?

Discussion in 'Abit' started by pigdos, Feb 18, 2006.

  1. pigdos

    pigdos Guest

    I've managed to fry some PC3200 GB DIMM's in the past by giving them 2.9V
    (on an NF7s v2.0). I think the NF7s v2.0 overvolts VDIMM (at least according
    to speedfan, Sisoft Sandra and the BIOS). Would adding copper heat spreaders
    make any difference over the stock AL ones? What about adding some sort of
    thermal paste between the individual RAM chips and the heat spreader?

    --
    Doug
    pigdos, Feb 18, 2006
    #1
    1. Advertising

  2. pigdos

    Mark C. Guest

    "pigdos" <> wrote in message news:spLJf.58355$...
    > I've managed to fry some PC3200 GB DIMM's in the past by giving them 2.9V
    > (on an NF7s v2.0). I think the NF7s v2.0 overvolts VDIMM (at least according
    > to speedfan, Sisoft Sandra and the BIOS). Would adding copper heat spreaders
    > make any difference over the stock AL ones? What about adding some sort of
    > thermal paste between the individual RAM chips and the heat spreader?


    The amount of voltage dimms can handle depends on what chips
    are used, e.g. Samsung TCCD's should never be run over 2.8V,
    while BH-5's and BH-6's can easily handle 3.3V or even more.
    This is regardless of whether and what kind of heat spreaders
    are installed.

    If you've got decent airflow over your dimms, heat spreaders
    accomplish nothing except to act as insulators, and make the
    chips run hotter than they otherwise would. They're a marketing
    gimmick and are useful only in rare cases.
    Mark C., Feb 18, 2006
    #2
    1. Advertising

  3. pigdos

    Wes Newell Guest

    On Sat, 18 Feb 2006 20:26:00 +0000, pigdos wrote:

    > I've managed to fry some PC3200 GB DIMM's in the past by giving them 2.9V
    > (on an NF7s v2.0). I think the NF7s v2.0 overvolts VDIMM (at least according
    > to speedfan, Sisoft Sandra and the BIOS). Would adding copper heat spreaders
    > make any difference over the stock AL ones? What about adding some sort of
    > thermal paste between the individual RAM chips and the heat spreader?


    As mark said, the heat spreaders do more harm thana good in most cases. If
    you are dtermined to use them, make sure you use some thermal compound and
    they fit flush on the ram.

    --
    Want the ultimate in free OTA SD/HDTV Recorder? http://mythtv.org
    http://mysettopbox.tv/knoppmyth.html Usenet alt.video.ptv.mythtv
    My server http://wesnewell.no-ip.com/cpu.php
    Wes Newell, Feb 19, 2006
    #3
  4. pigdos

    pigdos Guest

    After what you and Mark said (thanks for the info as well Mark), maybe I
    should just remove the heat spreaders altogether? My real problem is now
    wondering if the VDIMM voltages that are reported are actual. If I set my
    Nf7s v2.0 to 2.7V VDIMM, the sensors indicate 2.79V to 2.8V. If I set my
    VDIMM to 2.8V, the sensors report nearly 2.9V. A patriot tech support rep.
    told me that with their
    PEP1G3200LL (1Gig extreme low-latency PC3200) you should never go over 2.8V.
    Yet the only way I can get this RAM stable at ANYTHING over 200Mhz is to use
    2.8V VDIMM (which the on-board sensors are picking up as ~2.9V).

    --
    Doug
    "Wes Newell" <> wrote in message
    news:fUPJf.2987$6f2.1099@trnddc02...
    > On Sat, 18 Feb 2006 20:26:00 +0000, pigdos wrote:
    >
    >> I've managed to fry some PC3200 GB DIMM's in the past by giving them 2.9V
    >> (on an NF7s v2.0). I think the NF7s v2.0 overvolts VDIMM (at least
    >> according
    >> to speedfan, Sisoft Sandra and the BIOS). Would adding copper heat
    >> spreaders
    >> make any difference over the stock AL ones? What about adding some sort
    >> of
    >> thermal paste between the individual RAM chips and the heat spreader?

    >
    > As mark said, the heat spreaders do more harm thana good in most cases. If
    > you are dtermined to use them, make sure you use some thermal compound and
    > they fit flush on the ram.
    >
    > --
    > Want the ultimate in free OTA SD/HDTV Recorder? http://mythtv.org
    > http://mysettopbox.tv/knoppmyth.html Usenet alt.video.ptv.mythtv
    > My server http://wesnewell.no-ip.com/cpu.php
    >
    pigdos, Feb 19, 2006
    #4
  5. pigdos

    Mark C. Guest

    "pigdos" <> wrote in message news:x5TJf.58461$...
    > After what you and Mark said (thanks for the info as well Mark), maybe I
    > should just remove the heat spreaders altogether? My real problem is now
    > wondering if the VDIMM voltages that are reported are actual. If I set my
    > Nf7s v2.0 to 2.7V VDIMM, the sensors indicate 2.79V to 2.8V. If I set my
    > VDIMM to 2.8V, the sensors report nearly 2.9V. A patriot tech support rep.
    > told me that with their
    > PEP1G3200LL (1Gig extreme low-latency PC3200) you should never go over 2.8V.
    > Yet the only way I can get this RAM stable at ANYTHING over 200Mhz is to use
    > 2.8V VDIMM (which the on-board sensors are picking up as ~2.9V).


    First step is to see if Patriot can tell you what kind of chips are used
    in those dimms. Then do some research and find out what the upper
    voltage limits are for those chips.

    If Patriot can't or won't tell you this info, it's a judgment call whether
    you want to void your warranty and remove the spreaders.

    As for the difference between set vs. sensor readings, it's safer to
    assume the sensors are correct.
    Mark C., Feb 19, 2006
    #5
  6. pigdos

    - HAL9000 Guest

    If one removes the memory heat spreaders, and stop air flow around
    memory chips (that's a stretch), one would find that some memory chips
    consistently get warmer than the others. As I recall the chips
    handling the least significant bits get warmer than the MSB. I may
    have this switched backwards though. The primary reason is because of
    the bus default level (what it floats to) when reading and writing 8
    or 16 bit data instead of 32 bit data (full bus width). That is, a "0"
    and a "1" don't generate the same amount of heat in a memory chip.

    Also, the average value of all bits is much closer to zero than it is
    one - in both the address space and in the data space.

    Notice that they are not called "heat sinks". They are called "heat
    spreaders". This is appropriate because some chips are getting warmer
    than others. Heat spreaders reduce localized heating.

    Forrest

    Motherboard Help By HAL web site:
    http://home.comcast.net/~mobo.help/


    On Sat, 18 Feb 2006 20:26:00 GMT, "pigdos" <> wrote:

    >I've managed to fry some PC3200 GB DIMM's in the past by giving them 2.9V
    >(on an NF7s v2.0). I think the NF7s v2.0 overvolts VDIMM (at least according
    >to speedfan, Sisoft Sandra and the BIOS). Would adding copper heat spreaders
    >make any difference over the stock AL ones? What about adding some sort of
    >thermal paste between the individual RAM chips and the heat spreader?
    - HAL9000, Feb 20, 2006
    #6
  7. pigdos

    pigdos Guest

    LSb's getting warmer than MSb's (b is for bit, big B is for Byte)? LOL, this
    is the most ridiculous crap I've ever heard. Do you have any proof of this
    statement? There's all kinds of data stored in memory that are NOT integer
    data. There's floating point data, bit masks, compressed audio/video data,
    lookup tables, code etc. and none of these data types would tend to have
    their least significant bits on more than their MSb's. Even for integer data
    I don't even think it's that easy. What if I use chars or shorts to store
    small integers instead of longs or ulongs?

    Your idea about: "the average value of all bits is much closer to zero than
    it is one - in both the address space and in the data space" is interesting
    though. Are you saying that since most of the memory space of modern systems
    is unused most of the time, that the average of all bits in the system would
    be closer to zero than one?

    From what I've read here and other places it seems like clip-on, RAM
    heatspreaders are more for looks than anything else.


    Doug
    "- HAL9000" <> wrote in message
    news:...
    > If one removes the memory heat spreaders, and stop air flow around
    > memory chips (that's a stretch), one would find that some memory chips
    > consistently get warmer than the others. As I recall the chips
    > handling the least significant bits get warmer than the MSB. I may
    > have this switched backwards though. The primary reason is because of
    > the bus default level (what it floats to) when reading and writing 8
    > or 16 bit data instead of 32 bit data (full bus width). That is, a "0"
    > and a "1" don't generate the same amount of heat in a memory chip.
    >
    > Also, the average value of all bits is much closer to zero than it is
    > one - in both the address space and in the data space.
    >
    > Notice that they are not called "heat sinks". They are called "heat
    > spreaders". This is appropriate because some chips are getting warmer
    > than others. Heat spreaders reduce localized heating.
    >
    > Forrest
    >
    > Motherboard Help By HAL web site:
    > http://home.comcast.net/~mobo.help/
    >
    >
    > On Sat, 18 Feb 2006 20:26:00 GMT, "pigdos" <> wrote:
    >
    >>I've managed to fry some PC3200 GB DIMM's in the past by giving them 2.9V
    >>(on an NF7s v2.0). I think the NF7s v2.0 overvolts VDIMM (at least
    >>according
    >>to speedfan, Sisoft Sandra and the BIOS). Would adding copper heat
    >>spreaders
    >>make any difference over the stock AL ones? What about adding some sort of
    >>thermal paste between the individual RAM chips and the heat spreader?

    >
    >
    pigdos, Feb 21, 2006
    #7
  8. pigdos

    Wes Newell Guest

    On Tue, 21 Feb 2006 00:54:55 +0000, pigdos wrote:

    > From what I've read here and other places it seems like clip-on, RAM
    > heatspreaders are more for looks than anything else.
    >

    The manufacurer of the pc3700 ram I use discourages the use of heat
    spreaders saying most do more harm than good. There are however some good
    ones, but they will have fins on them and they must be installed properly
    to function properly. And the means good contact and thermal compound. it
    usually also means you'll be limited to one stick of ram per every 2 slots.:)

    --
    Want the ultimate in free OTA SD/HDTV Recorder? http://mythtv.org
    http://mysettopbox.tv/knoppmyth.html Usenet alt.video.ptv.mythtv
    My server http://wesnewell.no-ip.com/cpu.php
    Wes Newell, Feb 21, 2006
    #8
  9. pigdos

    - HAL9000 Guest

    Sure, heat spreaders.

    Forrest

    Motherboard Help By HAL web site:
    http://home.comcast.net/~mobo.help/


    On Tue, 21 Feb 2006 00:54:55 GMT, "pigdos" <> wrote:

    >LSb's getting warmer than MSb's (b is for bit, big B is for Byte)? LOL, this
    >is the most ridiculous crap I've ever heard. Do you have any proof of this
    >statement?

    < snip >
    - HAL9000, Feb 21, 2006
    #9
  10. pigdos

    John Lewis Guest

    On Sun, 19 Feb 2006 05:10:53 GMT, "pigdos" <> wrote:

    >After what you and Mark said (thanks for the info as well Mark), maybe I
    >should just remove the heat spreaders altogether? My real problem is now
    >wondering if the VDIMM voltages that are reported are actual. If I set my
    >Nf7s v2.0 to 2.7V VDIMM, the sensors indicate 2.79V to 2.8V. If I set my
    >VDIMM to 2.8V, the sensors report nearly 2.9V. A patriot tech support rep.
    >told me that with their
    >PEP1G3200LL (1Gig extreme low-latency PC3200) you should never go over 2.8V.
    >Yet the only way I can get this RAM stable at ANYTHING over 200Mhz is to use
    >2.8V VDIMM (which the on-board sensors are picking up as ~2.9V).
    >


    PC3200 does mean 200MHz memory clock rate, so what did you
    expect... a miracle? Unless the manufacturer's specifications include
    recommended alternate timing settings for the clock rates at which you
    wish to run and you then manually set these timings in BIOS. No such
    specification alternatives - no guarantee at any clock rate in excess
    of 200MHz. Lacking any such higher clock-rate specs, your only hope
    is to set CL to a higher value and/or set the command rate to 2T
    instead of 1T and hope for the best while running memtest86 for at
    least 8 hours. Trying to meet YOUR clock-specs just with a higher
    voltage and no timing changes is likely to just fry the memory
    devices.

    How about using memory sticks genuinely spec'd for the maximum clock
    rate you are ever likely to use, with latency specs (at that speed) to
    your taste and then running them at the manufacturer's spec'd voltage?

    Grrr... some people buy a Pinto, then expect Corvette performance.

    John Lewis
    John Lewis, Feb 22, 2006
    #10
  11. pigdos

    pigdos Guest

    Get real. All I really wanted was another 5 Mhz, 2.5% out of spec. Of course
    I changed the timing (all the way to 3-3-3-11) but that didn't help, the RAM
    still wouldn't pass memtest.

    --
    Doug
    "John Lewis" <> wrote in message
    news:...
    > On Sun, 19 Feb 2006 05:10:53 GMT, "pigdos" <> wrote:
    >
    >
    > PC3200 does mean 200MHz memory clock rate, so what did you
    > expect... a miracle? Unless the manufacturer's specifications include
    > recommended alternate timing settings for the clock rates at which you
    > wish to run and you then manually set these timings in BIOS. No such
    > specification alternatives - no guarantee at any clock rate in excess
    > of 200MHz. Lacking any such higher clock-rate specs, your only hope
    > is to set CL to a higher value and/or set the command rate to 2T
    > instead of 1T and hope for the best while running memtest86 for at
    > least 8 hours. Trying to meet YOUR clock-specs just with a higher
    > voltage and no timing changes is likely to just fry the memory
    > devices.
    >
    > How about using memory sticks genuinely spec'd for the maximum clock
    > rate you are ever likely to use, with latency specs (at that speed) to
    > your taste and then running them at the manufacturer's spec'd voltage?
    >
    > Grrr... some people buy a Pinto, then expect Corvette performance.
    >
    > John Lewis
    pigdos, Feb 23, 2006
    #11
  12. pigdos

    pigdos Guest

    Just as I thought, none... Additionally, since all chips on a stick of RAM
    hold bytes of data and are not bit-addressable how could any individual one
    of them be guaranteed to hold the LSb's? In short, they can't. In short,
    you're completely wrong...

    --
    Doug
    "- HAL9000" <> wrote in message
    news:p...
    > Sure, heat spreaders.
    >
    > Forrest
    >
    > Motherboard Help By HAL web site:
    > http://home.comcast.net/~mobo.help/
    >
    >
    > On Tue, 21 Feb 2006 00:54:55 GMT, "pigdos" <> wrote:
    >
    >>LSb's getting warmer than MSb's (b is for bit, big B is for Byte)? LOL,
    >>this
    >>is the most ridiculous crap I've ever heard. Do you have any proof of this
    >>statement?

    > < snip >
    >
    >
    pigdos, Mar 10, 2006
    #12
  13. pigdos

    - HAL9000 Guest

    *IF* you do some research on the subject you will find out that not
    all memory is as cool as the stick of ram in your computer is today.
    When memory got hot, and memory overheating was important, the heat
    spreader was invented. It's just been carried over, rightly or
    wrongly, to the present.

    What does bit addressable have to do with guaranteeing a certain bit
    is the least significant bit?

    Are you aware that (32 bit) compilers make a habit of aligning data on
    32 bit boundaries thus assuring almost any access, 8, 16, or 32 bit
    always has one particular bit as the least significant bit?

    We seemed to got off in the "data space" world. As I mentioned
    before, the additional heating occurs in the "address space" as well.

    Forrest

    Motherboard Help By HAL web site:
    http://home.comcast.net/~mobo.help/


    On Fri, 10 Mar 2006 03:47:49 GMT, "pigdos" <> wrote:

    >Just as I thought, none... Additionally, since all chips on a stick of RAM
    >hold bytes of data and are not bit-addressable how could any individual one
    >of them be guaranteed to hold the LSb's? In short, they can't. In short,
    >you're completely wrong...
    - HAL9000, Mar 10, 2006
    #13
    1. Advertising

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

It takes just 2 minutes to sign up (and it's free!). Just click the sign up button to choose a username and then you can ask your own questions on the forum.
Similar Threads
  1. Kenneth A Kauffman

    Memory and Heat Spreaders

    Kenneth A Kauffman, Jul 7, 2003, in forum: Abit
    Replies:
    3
    Views:
    639
    G Ras
    Jul 8, 2003
  2. Wblane
    Replies:
    1
    Views:
    318
    Morgan
    Aug 22, 2004
  3. The ATM

    Money Money Money

    The ATM, Mar 11, 2005, in forum: Chaintech
    Replies:
    0
    Views:
    773
    The ATM
    Mar 11, 2005
  4. The ATM

    Money Money Money

    The ATM, Mar 11, 2005, in forum: Asus
    Replies:
    0
    Views:
    339
    The ATM
    Mar 11, 2005
  5. The ATM

    Money Money Money

    The ATM, Mar 11, 2005, in forum: ECS
    Replies:
    0
    Views:
    604
    The ATM
    Mar 11, 2005
Loading...

Share This Page