new system bootup woes

Discussion in 'PC Hardware' started by Adam, Dec 17, 2014.

  1. Adam

    Paul Guest

    Burn in, is for weeding out infant mortality.

    You want a stability test that functions as an
    acceptance test. To determine whether all
    the gear functions well together.

    Memtest86+ from, is a good test for
    stuck-at faults on RAM. Those are bad RAM locations,
    that won't store data properly, and always return the
    same stuck bit value. I've had a DIMM, where an entire
    chip died, and it was pretty comical to see a stream of
    errors printed on the memtest screen.

    The other one is Prime95 (torture test only), from Versions available for Linux
    and Windows. The OS portion of memory cannot be tested.
    The stability aspect of this, is the generation of as
    much electrical noise as possible, and the search for
    transient errors as a result. Prime95 is multi-threaded,
    and you can have a thread per core. You can manually
    assign an amount of RAM for the threads as well. And
    using Linux "top", you can check how much is being used.

    Any where from four to eight hours of that, with none of
    the test threads stopping on an error, is sufficient.
    After that, it's "on with the show".

    Sometimes, I add in a video game demo loop, while
    Prime95 is running. But that seemed to be more
    important during the AGP slot age of video cards.
    I think the only 3D game I have here for Linux, is
    the Linux port of Quake 3 Arena, that uses the map files
    off the Quake CD. And Quake is one of the most light-weight
    resource users, in terms of 3D. It's very economical,
    and makes even low end video cards look heroic. Setting
    up something like that would be optional, and likely a lot
    of work for little benefit. That's always been a weakness
    of my hardware testing, is I've never really been able
    to give other people a simple recipe for video testing.
    Sometimes, I end up fiddling with that stuff for a week
    here, before giving up (on the notion of keeping it simple).

    Paul, Dec 29, 2014
    1. Advertisements

  2. Adam

    Adam Guest

    Thanks (Guru Paul) !!

    Oh yeah, I remember seeing Memtest86+ on the GRUB boot menu.
    Will definitely run Memtest86+.

    Prime95 is new to me. Will have to try it out.

    I don't do much gaming so not sure about
    the video game demo loop (in parallel with Prime95) yet.
    Adam, Dec 29, 2014
    1. Advertisements

  3. Something to be said for the newer packages by Intel with "bumps" not
    pins. No more anguished moan when a bent pin becomes a missing pin on a
    not-too-cheap CPU.
    Jonathan N. Little, Dec 29, 2014
  4. Adam

    Paul Guest

    Both schemes have their pluses and minuses.

    For the pin and ZIF socket case, you can use a ball point pen
    refill (the old kind with the metal tube), and slide that over
    the pin to use as a lever, to straighten it out. You don't
    need to use needle nose pliers to fix one. If a pin gets bent
    over to a 90 degree angle, chances are it's toast when
    straightened up.

    The ZIF sockets have pretty good properties. I've not heard of
    electrical issues with them. They can withstand the odd "pull-out"
    accident. They can be completely destroyed if you put your
    muscles into it (the top will pop off).

    The Land Grid Array concept, the socket is the weak link.
    When I bought my last motherboard from a local retailer,
    I couldn't leave the store with the purchase, until we
    went over to their support desk. And had a "socket check"
    before leaving the store. That's a visual inspection for
    damage to the socket, so later you cannot bring the
    motherboard back and complain the product shipped with
    a damaged socket. But that does indeed happen - products
    do leave the factory with crushed spring contacts in
    the socket. The evidence suggests the motherboard got
    damaged at the factory, just before being put into the box.

    LGA sockets vary in quality. There was an incident recorded on
    Anandtech, where socket springs seemed to be making poor
    contact, which was detected later on when overclocking and
    the other contacts would overheat. I've never heard of
    a ZIF socket quality problem, so perhaps they're a bit
    easier to make. The springs in an LGA are pretty brittle.
    The spring in the socket "bites" into the land pad on
    the other side. You can see a mark on the CPU, once it's
    been inserted into the socket. I don't know how many
    cycles such a scheme could take. I haven't been pulling
    CPUs out of LGA motherboards here all that often,
    to discover what the limit might be.

    Maybe the LGA scheme can be built to higher contact
    counts, than a ZIF can. The very largest LGA has
    such a high contact force (due to the number of
    springs times the force per spring), that it uses
    two levers to close the lid. Whereas the same degree
    of force isn't as evident when closing a ZIF lever.

    If you had a motherboard with a ZIF socket on it,
    it wouldn't need a "socket check" just after you
    bought it at the computer store. It takes deliberate
    tampering to mess one up.

    Paul, Dec 29, 2014
  5. Adam

    Adam Guest

    Lesson learned ... always pull that heatsink "straight" up (with "both"
    steady hands)!!
    Adam, Dec 29, 2014
  6. Adam

    Adam Guest

    And, always look at the copper before putting it down. :)
    Adam, Dec 29, 2014
  7. Adam

    Adam Guest

    For the ball point pen refill, which ones (make, model,
    fine or medium?, etc.) are best? URL?

    In that case, I think I prefer the "pin and ZIF socket case" myself.
    It just takes a little more delicate cautious handling.
    Even a newbie like me can handle CPU reinstallation now.

    When the pins are straight, the CPU naturally "falls" into
    the ZIF socket when properly aligned/positioned.
    Adam, Dec 29, 2014
  8. Adam

    Paul Guest

    You go out to the kitchen, and start taking apart the old
    clickie type ball point pens, and look for the right size
    tube. It's been a long time, since the proper sized metal
    tube refill, has been for sale.

    I'm sure that somewhere, there is a real tool for the
    job. I wouldn't expect the kind of retailers I have
    here, to stock one.

    Paul, Dec 29, 2014
  9. Adam

    Adam Guest

    With the exception of the old PCI Creative SoundBlaster Live! 5.1 sound
    all componenets (video card, drives, etc.) have been installed in the Antec
    case and
    the system acknowledges all components on boot up with Ubuntu Live CD.

    The only issue is ... I am using ...

    Crucial Ballistix Tactical 240-Pin DDR3 SDRAM 8 Dual Channel Kit 1866
    (PC3 15000) BLT2CP4G3D1869DT1TX0

    But, system only acknowledges 8192 MB (DDR3 - 1333 MHz)

    System sees all the RAM but is 1333 MHz RAM the max that this mobo supports?

    Other info shown in BIOS setup ...

    CPU 1.344V
    3.3V 3.360V
    5V 5.035V
    12V 12.252V

    CPU +107.6F/+42.0C
    MB +87.8F/+31.0C

    Fan Speed...
    CPU Fan ~2973 RPM
    Adam, Dec 29, 2014
  10. Adam

    mike Guest

    Hobby Store, brass tubing.
    mike, Dec 29, 2014
  11. Adam

    Adam Guest

    Ballistix 1866MHz only does 1333MHz on SABERTOOTH...
    Adam, Dec 29, 2014
  12. Adam

    Paul Guest

    8GB Kit (4GBx2)
    DDR3-1866 9-9-9-27


    The AMD site is just useless for real-world problems.

    The cpu-world site can sometimes fill in the gaps. II X4 965 - HDX965FBK4DGM.html

    Memory channels: 2
    DIMMs per channel: up to 2

    Supported memory: DDR2-1066 <--- CPU in AM2+ mobo
    DDR3-1333 <--- CPU in AM3 mobo (yours)

    So that's the stock speed (1333). Any more than
    that, would involve some level of overclocking
    of the memory bus.

    Using your 965 part number, and going through the reviews for
    that processor on Newegg, most people seem to stop at DDR3-1600
    as a setting. Which suggests it doesn't have a lot of headroom.
    Now, if I data mined an overclocker site, I'd likely find
    a higher result than that (maybe requiring Vdimm to be bumped up).
    But it doesn't appear people are overdoing it by any stretch.

    Paul, Dec 29, 2014
  13. Adam

    Adam Guest

    Thanks (Guru Paul), sounds like the "most" limiting factor is
    not the mobo but rather the CPU. If so, I'll be keeping my eyes open for
    good CPU deals. What's a good not-too-expensive CPU for my system?

    BTW, memtest86+ v4.20 has been running for 43+ minutes and
    so far so good with no errors.

    Memtest86+ shows...
    Settings: RAM: 666 MHz (DDR1333) / CAS: 11-11-11-28 / DDR3 (64 bits)
    Adam, Dec 30, 2014
  14. Adam

    Adam Guest

    Actually, running the latest Memtest86+ v5.1.0 shows...
    RAM Info: PC3-14900 DDR3 XMP 933 MHz / 9-9-9-27 / Crucial Technology BLT4G3
    Adam, Dec 30, 2014
  15. Adam

    Paul Guest

    You can tighten up your timing parameters a bit.
    Using speed ratios and rounding up gives...

    DDR3-1866 9-9-9 = DDR3-1333 7-7-7

    The easiest setting to change, is to change just CAS.
    For example 7-11-11-28 is better than nothing. The
    utility of the parameters decreases from left to right.
    The left-most one (CAS) is the most important. Since
    the clock speed is dropped, a clock period is wider,
    so we need fewer of them (7) to get first data from
    the RAM.

    Normally, I'd tell you to use CPU-Z in Windows, which can
    dump the SPD table and give you some idea why the BIOS selected
    those particular values.

    In this example, the left-most column would be for
    DDR3-1333. The frequency is 622, and double that is
    close to 1333. So that would be the timing for
    DDR3-1333 clock setting.

    Using those exact numbers, you could enter the BIOS, find
    the custom memory page, and enter the numbers.

    It's also possible, to dump the SPD EEPROM and analyse the
    table by hand. (I have to dig up the appropriate JEDEC
    decoder document, which isn't always that easy.)

    I'm running Gentoo on the test box right now, and
    it doesn't have module "EEPROM" in the kernel.
    Without that, i2c-tools program named "decode-dimms"
    won't work. And the information from decode-dimms
    is not suited to directly dialing into the BIOS.
    It's more a conversion of the raw hex, into English.
    But without the interpretation to make timing tables.
    I can't really test this at the moment, because I can't
    access the eeprom. I do actually have i2c-tools loaded,
    but I'm have to rebuild the kernel. And I'm not in the
    mood for that right now. There's no place to
    sit, where that computer is located :)

    Paul, Dec 30, 2014
  16. Adam

    Rodney Pont Guest

    Rodney Pont, Dec 30, 2014
  17. Adam

    Adam Guest

    Hey, we're getting some really creative suggestions. Like!!

    Adam, Dec 30, 2014
  18. DecadentLinuxUserNumeroUno, Dec 30, 2014
    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.