Sstem idle but CPU usage is high!!

Discussion in 'Computer Information' started by Colin Trunt, Jul 8, 2010.

  1. Colin Trunt

    Colin Trunt Guest

    OK AMD 3800 X2 running XP.

    Just installed an extra gig of ram, I had problems it it got some
    beeps, I reinstalled and booted up OK.

    ThenI notice one core is running 90% CPU when system is idle
    other core should a few % 1-5 rrange.

    In task manager, system idle shows 98% as it should becasue
    nothing is running
    however I am still getting shown CPU usage of about 50% overall ie
    one core running almost flat out and the other idle!!


    I though I had got rid of it by using only 1 gig of ram, and I had it went
    so I tried the other gig by itself and the problem reappeared,
    worst thing is is that it is the old 1 gig stick showing the problem now
    not the new one so I cant even send it back as fault.

    I am going to but the new 1 gig in again and see if the problem goes away
    Colin Trunt, Jul 8, 2010
    1. Advertisements

  2. Colin Trunt

    Colin Trunt Guest

    Colin Trunt, Jul 8, 2010
    1. Advertisements

  3. Colin Trunt

    Paul Guest

    The information shown in Task Manager, should have consistency, so check
    it again, to make sure you aren't mixing up some busy thing, versus some
    idle thing. Look in the Process list, for what is using the cycles.

    For an alternative tool, you can try Process Monitor.

    There are things that can loop in an OS, but I'd expect what you're
    seeing in Task Manager to add up in the "bank balance" sense.

    It is possible to turn the Task Manager display into nonsense, by
    installing your own Idle Task using something like RMClock. But if you were
    doing something like that, you'd remember setting that up.


    And when you buy new memory, you should test it *before* booting
    into Windows. Bad memory can lead to registry corruption,
    as the registry can be read into RAM, corrupted, and then written
    back out later. Use a memory test program and run it for a couple
    complete passes. You could do a run, with just the good
    stick present. And a run with the bad stick present, just
    to confirm your suspicions. Then, one final run, with
    all the sticks present that will be your permanent configuration.
    No errors are acceptable.

    (Scroll half way down...)

    Since System Restore in WinXP allows you to return to a time
    before the bad RAM was installed, you can repair your registry
    that way. But be careful, because System Restore tracks changes
    in areas other than "My Documents", and if you store your downloads
    outside that folder, using System Restore to an older date, could
    result in some downloads being lost. If in doubt, do a backup first,
    then run System Restore to a previous date in time (having the backup,
    you know you can't lose anything). You can "undo" a Restore point,
    if done in normal mode. But doing it in Safe Mode, you may lose
    the undo feature. And don't do that sort of stuff, until your
    memory is fixed :) You need to keep those Restore points in
    good condition, for when your RAM is "known good".

    Paul, Jul 8, 2010
  4. Colin Trunt

    Colin Trunt Guest

    I would expect the system to do a memory test on start-up.
    Infact i think it does, I got beeps at start-up sometime, this
    was becasue I had badly seated memory.
    I am not sure of the numberof beep I powered off after 3 or 4.
    It might have been 1 long beeep repeated.

    Anyway the problem is cured, I woudl say when I had the CPU
    usage problems there were no beeps (IIRC).
    I think carefully inserting the memory properly may have been
    the cure. I also have a graphics card which is badly secured and that
    cause problems sometimes as it can become badly seated, however
    usually results in no boot-up at all just the fan on maximum speed and
    very loud which immediately make you want to poser off (at that stage
    only pulling out the plug will work, the power switch on the front will
    not work).

    Anyway, thanks for the help however it does seem to be cured
    for the time being, that stuff will come in useful in the future I think.

    Also I have to add the last stick of ram, the last 256meg I have
    never got it to boot with the full 1.0 + 1.0 + 0.256 = 2..256 GIG memory.

    I always got beeps with all three stciks in, I will have to carefully count
    beeps next time, I have tendancy to just power off incase damage is
    being done.

    Maybe I was not seating the ram properly, I have great problems
    doing that.

    Also the mobo is dirty and full of fluff and dirt all covered in tobacco
    smoke deposits so that does not help, a bit of dirt will cause an open
    circuit it it gets in a slot.

    But I am OK at the momoent, I have both cores working properlly
    and 2 gig of ram, the last 256gig is not a major concern really, a marginal
    ammount extra.
    Colin Trunt, Jul 8, 2010
  5. Colin Trunt

    Paul Guest

    I don't know the details of your motherboard (make and model),
    but I would recommend sticking with the 2x1GB configuration, and just
    leave the 256MB in your junk pile for now. On a dual channel capable
    motherboard, the 2x1GB can be plugged in, in a dual channel configuration.
    And that will do a better job on memory bandwidth. With AMD, there is
    a performance penalty due to bus loading, so some of the four stick
    configurations won't be quite as fast as a two stick configuration.

    It sounds to me, like you may have an S939 system, in which case
    I'd stick with the 2x1GB. If we go back far enough in time, there
    are revisions of S939 processor, that will ignore that third stick.
    As far as I know, Revision E or later, on S939, should be fully
    flexible (i.e. allow three sticks installed of four available), but
    there would still be a performance penalty associated with that.

    As an "inquiring mind", you should test your system in single channel
    and dual channel modes, to see what the impact is. Plug in a single
    1GB stick. Run SuperPI benchmark, which has elements of CPU performance
    as well as memory performance. The 1 million digit benchmark has a
    8MB memory footprint. The 32 million digit benchmark has a 256MB
    memory footprint. You want to run a test case, where the memory footprint
    is significantly larger than the processor cache (so any memory accesses
    while calculating PI, hit main memory).

    Now, plug in 2x1GB in dual channel, and repeat the same benchmark. How
    much difference did it make ? If the results are the same as with
    a single stick, you haven't put the two sticks in dual channel slots.

    Then, you could install the 256MB, and see what the third stick
    costs you in practical performance.

    You can get SuperPI benchmark here.

    super_pi_mod.exe is 104,960 bytes and has an MD5SUM of
    ca977a1cd46b170ddea7adc4e19bd6ba . A scan of the
    file should take you here.

    To understand that benchmark, you have to consider how it was built.
    It was written maybe 15 years ago, by a guy in Japan. The source
    code may have been lost. The original program, might not have
    given very accurate timings, due to the software method used
    to do timing.

    Someone modified the executable code file with a hex editor.
    They painstakingly added code, to make the program more robust
    (harder to create fake results), as well as increasing the digits
    of accuracy. And part of those efforts, may have resulted in
    the triggering of heuristic AV detection. I've been using that
    particular copy for some time.

    The benchmark used to be staged here, after it was hand modified.
    But it isn't there any more.

    There may be a copy on , which snapshots the file when
    it was on Extremesystems.*/

    The MD5SUM of that one, seems to be the same.

    You really should use memtest86+, and test your memory :)

    Have fun,
    Paul, Jul 9, 2010
  6. Colin Trunt

    Colin Trunt Guest

    * Four 184-pin DDR DIMM sockets
    * Supports dual channel, eight memory banks DDR 333/400 DDR DIMMs
    * Supports 2.5v DDR SDRAM DIMMs
    * Maximum HP/Compaq approved memory is 4 GB* without ECC

    It is a 939.

    I was running with a 256 meg stick in slot one an a 1 gig stick in slot 2
    for a long time, i wondered if it would be better with one stick in there
    but well its 25% more memory and less paging.
    Now it will only be 12.5% more memory so not such a big increase.

    However I will try those memery benchmarks as I would be interested to
    see the results.
    The only problem is I have so much trouble getting the sticks into
    the slots, I think that may have been the cause of a lot of my problems,
    there are a lod of cables over the slots so it hard to see and hard to put
    pressure on the sticks to put them in properly, sometimes I think I might
    Also I only have one really old back-up computer, howver so far I have
    not broken either yet (touch wood).

    It does seem to be running quite nicely at the moment though, very
    little paging, fan rarely comes on, nice and quiet.
    Still it has now gobbled up 90% of the memory but I think that
    is normal, it will use as much as it can.

    Will try those tests later.
    Colin Trunt, Jul 9, 2010
  7. Colin Trunt

    Colin Trunt Guest

    So it seems mine is set up as a matched pair at the moment,
    although I do not know it what matched means, does it mean
    the same size and type, or is there more to it than that.
    They should be both 1 gig unbbuffered non-ecc.

    I guess running those tests should provide some answers.
    Colin Trunt, Jul 9, 2010
  8. Colin Trunt

    Paul Guest

    Gobbling up 90% of memory isn't normal. I have 2GB of RAM on
    WinXP here, and a few minute ago, with Thunderbird and Firefox
    open, I was using 462MB. With your 1GB of RAM, that would be
    about half of it.

    I still think something isn't right with your system, and
    that you should be looking more carefully in Task Manager,
    to see where the memory is going, and whether there is a
    memory leak.

    WinXP doesn't work the same way, as some of the later OSes.
    Some of those, make more usage of RAM, and show it that
    way in displays of status. But WinXP and Win2K don't do
    that here for me. Memory usage should be proportional to
    the usage the programs are making of it. The system
    component shouldn't be ballooning.


    When you don't have documentation, on which slots hold the
    matched set, you can always test. And SuperPI can help you
    there. Try a DIMM configuration, and see how it benchmarks.

    With respect to the force required to seat modules, I have
    the most trouble with "low profile" Kingston modules. They
    have a high insertion force, that hurts my fingers when the
    modules finally seat. The lock latches on the ends, should both
    rotate into the upright position, if the thing is seated. And
    you might hear a bit of a click when it seats.

    Placing 1GB in one channel, and 256MB in the other channel,
    doesn't match them. So they'll operate in virtual single
    channel mode, at least, as long as the processor is Rev.E or

    If you were to place the 1GB and 256MB on the same channel
    (perhaps slot 1 and slot 3, not sure), the result would be
    even slower than just having 1GB in slot 1. When the bus
    has two loads on it, the memory clock rate is set a bit lower
    on S939 (could drop from DDR400 to DDR333). A user can tweak that,
    if the motherboard has adjustments. Whether it can be set more
    aggressively than the BIOS configures it, depends on the
    individual processor and the voltages involved.

    Another tool for detecting matching, might be CPUZ. It
    should be able to display "dual channel" or similar
    terminology, when showing your hardware details. Now
    that you don't have a matched set of RAM, then there
    isn't any chance of it saying "dual channel" or 128 bit

    (Select one of the no-install versions)

    The memory dimensions are rows, columns, banks, and ranks.
    The first three, occupy a single memory chip. The memory
    chip might have four banks, selected by BA bank bits. The
    rows, columns, and banks, account for the number of bits
    inside the memory chip. They put enough chips side by
    side, to build a 64 bit wide array, and that is a rank.
    A double sided module has room for two ranks.

    If you had a single sided, 8 chip, 1GB module, and a
    double sided, 16 chip, 1GB module, those modules would not
    match. An S939 processor would have to be set up, to run
    those sticks in virtual single channel mode. That's why,
    when buying the memory DIMMs, it pays to buy a pair in a

    Companies like Kingston have got into a bit
    of trouble with their customers, by forgetting about
    the matching thing, and just changing the module chip
    composition without telling anyone (as a function of
    which chip type is cheapest today). In fact, on some
    motherboards, where this practice resulted in customer
    returns on memory, they stopped recommending memory
    entirely for those motherboards, so it couldn't happen
    again. If you, as a customer, insist on buying DIMMs
    one at a time, then you have to put extra effort into
    it, to make sure your rows, columns, banks and ranks match.
    Roughly speaking, if two modules had 16 chips, and they
    were 1GB total capacity, then chances are the dimensions
    are the same. With some RAM companies, you can't tell in
    advance, whether you're getting an 8 or a 16 chip module.
    Kingston used to be good about that sort of thing, and
    Kingston was one of the few companies that had datasheets
    for their retail products (complete with drawing of the module
    and info stating 64Mx8 for the chip). You used to be able to
    rely on the datasheet, to predict the module composition
    in advance. But there have been some products shipped, where
    they couldn't be bothered to update their documentation,
    which is a shame. If they want to play that game, it would
    have been simple enough to give the alternative products
    their own SKUs. They could price them independently,
    such that customers could decide whether they wanted the
    cheaper one, or wanted the other one because of its

    Paul, Jul 9, 2010
  9. Colin Trunt

    - Bobb - Guest

    Test results ??
    It's getting tough following this: title shows CPU yet issue is memory ?
    How about a brief summary of the situation as of now - and what didn't work.
    Also, right from your HP link I see:

    Video graphics are "Integrated" (on-board) so that's using some of your
    memory too.
    Someone didn't add a picture background to your display ???
    If so, make the background NONE - just a color.
    - Bobb -, Jul 9, 2010
  10. Colin Trunt

    Colin Trunt Guest

    The CPU problem seemed to be related to badly installed memeory (probably)
    I have a grahics card with 1/2 a gg of memory.
    Colin Trunt, Jul 9, 2010
  11. Colin Trunt

    Colin Trunt Guest

    If you open a lot of windows it will use more, I think it just
    leaves things in memory in case it need them again,
    Try running a video youtube and see what that does to meemory.

    I currently have 770 mega avilalble but it I ran a lot of programs
    it woudl eventualy gobble that up, especially it I was runnin diagnostic
    programs and searchnig for a solution on the internet.

    The things using memeory are Firefox and Chrome, but I
    have 8 chrome and 2 firefox tabs open.

    I also have something called jqs.exe, java quick starter, I don't
    think I need this, I think I only have it because it has 'quick' in the
    I have killed it off and changed it's setting to not cache files, it had
    been set to cache up to 1 giga byte.
    I will see if that makes a diference to anything.
    Yes they hurt my fingers too, at one point I was using the back of
    a screw driver to apply the pressure, not too sensible really.
    I think I have two 16 chips modules, old 256meg module is
    one sided.
    Thanks, I am fairly sure mine should be matched, all dimensions
    that are listed seem to be the same.
    They are als in the same coloured slots which is necesary I belive.

    I have not got round to running the memeory tests stuff yet
    but I will have a go later.
    Colin Trunt, Jul 9, 2010
  12. Colin Trunt

    - Bobb - Guest

    "using the back of a screw driver to apply pressure" !!

    I have had several PC's with DDR400 and none of them have a problem.
    You line up the memory stick in the slot, then push down slightly.
    IF it should bind at all ( make sure no obstruction under the levers) lift
    up on the lever edges as you push down and should ease right in. Maybe
    another pair of hands to help ?
    - Bobb -, Jul 9, 2010
  13. Colin Trunt

    Paul Guest

    If you look at your Task Manager graphs, when there isn't any heavy user
    stuff running, are you still seeing one core running at 100%. Or is the
    system now, really "idle" ?

    Also, your statement about memory usage. You seem to be saying it follows
    program usage, which is good. If Chrome and Firefox aren't running, and
    the system is effectively idle, are you seeing 90% memory usage at
    that point in time ?

    Paul, Jul 9, 2010
  14. Colin Trunt

    Colin Trunt Guest

    Ot is OK now however before nothing was running, ie system idle 97%
    but one core was at 95%, however that problem now appear to be cured.
    Possibly related to badly seated memory or some other memery problem.
    OK I kust closed chrome and firefox and memory available increased
    from 700 to 1400 meg
    Colin Trunt, Jul 9, 2010
  15. Colin Trunt

    Colin Trunt Guest

    O just closed another program (poker client) and Outlook Express an it
    increased to 1500 meg available.

    I am not sure memory was freed so much before, it might be because
    I killed this jqs.exe process which caches java for 'quick start', that may
    have kept a lot of stuff in memeory stopping it being freed.

    I guess I could switch it back on and see the difference, it was set up to
    caches up to 1 giga byte which is lot of java!!
    It's not something I really need though, I think it is better not running.
    Colin Trunt, Jul 9, 2010
  16. Colin Trunt

    - Bobb - Guest

    Have you tried booting into Safe Mode and look at memory then ?
    Press F8 while booting ( Google it for details).
    If a "memory seating issue" ,should still have a problem.
    If not a seating issue, then should be minimum loaded into memory / CPU
    usage minimal.
    - Bobb -, Jul 10, 2010
  17. Colin Trunt

    Colin Trunt Guest

    I did do this and it showed the ram was there.
    IIRC it showed the ram was there even when I had a CPU usage issue.
    The CPU issue has fortunately disapeared now.
    Colin Trunt, Jul 10, 2010
  18. Colin Trunt

    - Bobb - Guest

    So problem is solved ?
    - Bobb -, Jul 10, 2010
  19. Colin Trunt

    Colin Trunt Guest

    Yes, wel it has gone away for now.
    As I am not too sure what the cause is it could come back but
    I am pretty sure it was poorly inserted memory.
    Colin Trunt, Jul 12, 2010
    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.