PC runs V E R Y S L O W after coming out of retirement

Discussion in 'PC Hardware' started by BW, Dec 16, 2012.

  1. BW

    BW Guest

    I got out an old PC from broom closet.It has Gigabyte 965P mainboard,
    Core 2 2666 and DDR2 667 RAM.
    Now it is amazingly slow. I have a home-theatre PC also with 965
    chipset, but only an 1866 MHz CPU and
    that goes about 10 times as fast.
    Everthing seems slow, from disk access to drawing graphics. Booting
    and shutting down takes aeons.
    It has linux installed. I checked "top" command and there are no
    phantom processes using up all the CPU cycles.
    And there is free RAM, it is not using swap disk.
    I checked BIOS settings, and everything is optimum without
    overclocking. I know if you do something like
    turn off L2 cache, that is worth roughly 10% penalty not 10x.
    I ran memest program and it shows the bandwidth of the caches and RAM
    as being normal (compared to
    the other 965 PC). But still it runs like a Trabant stuck in first
    gear.
    I have used about 100 PCs over the last 15 years, and have not seen a
    fault like this before.
    It used to work OK a few years back, before taking a holiday in the
    broom closet.
    So I will scrap it, but I am really curious: what sort of fault can
    slow a PC like so?
     
    BW, Dec 16, 2012
    #1
    1. Advertisements

  2. BW

    SC Tom Guest

    Are all your fans operating at speed? Maybe the CPU or GPU are overheating
    slightly, and the CPU has throttled back (not sure if yours supports that or
    not).
    Any chance you have a hard drive diagnostic program? I had an old out-dated
    server that was turned off for two or three years, and it took four or five
    power on/off cycles before the HDD would spin up enough for it to boot. Once
    I cloned it to a new drive, it worked fine again. Just a thought. . .
     
    SC Tom, Dec 16, 2012
    #2
    1. Advertisements

  3. BW

    Paul Guest

    If an add-in card generates "run-away interrupts", that can drop
    a computer to its knees. You really need to get tools which display
    hardware counters as much as possible, to detect such things.

    There were some chips, known to cause this problem. In other words,
    certain chip designs actually had a bug in the interrupt signal.
    So I'm not claiming a "card has failed". In some cases, 30% of a
    certain chip exhibit the interrupt bug, and would cause the PC to be
    slow.

    You can try removing any excess cards, like PCI or PCI Express
    cards that aren't being used for anything at the moment.

    In the BIOS, you could disable the Firewire chip or a Promise
    RAID chip or anything else you're not using to boot the computer.
    The idea being, that when the chip is disabled, we hope any
    interrupts can't get through.

    But at least it sounds like you're headed in the right direction.
    You correctly looked at cache state, and whether cache is enabled
    or disabled. On some systems, mis-configured cache makes a *big*
    difference.

    Having a disk controller slip from DMA to PIO mode (about 4MB/sec
    on reads), can have an impact on perceived speed. But then,
    operations that are purely compute-bound, are unaffected by that.
    A slow disk affects the feel of the machine, at least until you
    do something which is compute-bound (uses the CPU, but no disk).

    You can use something like SuperPI, selecting number of digits
    to be larger than cache size, to ensure fair testing. Compare
    the two computers, and see how they do. The test has a footprint,
    where I think 1 million digits uses 8MB of RAM, while the maximum
    number of digits might use 256MB of RAM. And at 256MB, no desktop
    processor can hold the entire thing in L3 cache.

    http://www.techpowerup.com/downloads/366/Super_PI_Mod_v1.5.html

    On my E8400 (3GHz, RAM at DDR2-800) right now, using SuperPI 1.5

    1 million digits = 15.469 seconds (8MB RAM allocated)
    2 million digits = 37.407 seconds (16MB RAM allocated)
    4 million digits = 87.609 seconds (32MB RAM allocated)

    It's possible the 6MB L2 cache on the processor, is aiding
    the 1 million digit case.

    Try scaling the results by frequency, and compare to your two processors.
    For the most part, this should be a compute bound test. Yes, the
    disk light will flash, but I think the clock is stopped during
    disk I/O.

    Paul
     
    Paul, Dec 16, 2012
    #3
  4. BW

    Flasherly Guest

    Start by treating the MB as if right out of the box (I'd leave the
    cpu&heatsink alone) - as near to the MB-only as possible - for video,
    a DVD or flashstick to boot from. Short the BIOS setting pins or
    remove battery power. Also reflash/update the BIOS if it's still
    slow. No fun replacing the CPU, but if you've a spare one within mb
    spec's - that's another option. You'll then have given it every
    chance to come up normally - short of replacing Gigabyte's mb support
    chips with surface-mount wave soldering.
     
    Flasherly, Dec 16, 2012
    #4
  5. BW

    Greegor Guest

    When you first pulled it out of the closet
    and ran it up, did you have to reset the
    BIOS settings? (CMOS settings?)

    Do the SATA drive ports have an IDE mode, Legacy Mode or "Combination"
    mode?

    Set on in BIOS settings?
     
    Greegor, Dec 18, 2012
    #5
  6. BW

    Justin Guest


    Obviously your laptop is a baby-boomer.
     
    Justin, Dec 18, 2012
    #6
    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.