ATX connector welded to Tyan Tiger MP S2460; can it happen again?

Discussion in 'Tyan' started by Shane Beasley, Jul 25, 2003.

  1. This was the machine as it was:

    - Tiger MP S2460
    - 2x Athlon MP 1600+
    - 2x PC Power and Cooling CPU Cool AG heatsink/fan
    - 2x Kingston 256MB PC2100 registered SDRAM
    - ATI Radeon 8500 (+ fan)
    - SigmaDesigns Hollywood Plus MPEG decoder
    - SBLive! Value sound card
    - Netgear FA-311 NIC
    - Antec PP-303X 300W PSU (30A on +5v)
    - 2x 80mm fans which drew from motherboard fan power connectors
    - 2x 80mm fans which drew from PSU drive power connectors
    - Raite RDR-108H DVD-ROM
    - IBM Deskstar 60GB 7200 RPM
    - Western Digital Caviar 30GB (may not have been plugged in?)
    - 3.5" floppy drive from before the dawn of time, or something

    The machine (assembled Jan/Feb 2002) worked great until recently.
    That's when it started spazzing on me, crashing more and more
    frequently. These crashes usually came with "random" video artifacts
    (reminiscent of what happens to an Atari 2600 when it loses contact
    with the cartridge). It even crashed once while in the BIOS (text
    artifacts there). I eventually pulled all the hardware but the
    essentials (CPU, RAM, board, video), but it was gone -- it would power
    up, but I'd either get a very short uptime or no POST at all.

    I tried swapping video cards, but no go. I don't carry spare MP
    hardware, so I made an educated guess that the motherboard (and not
    the CPUs or RAM) was the culprit. It looks like I was right.

    I do development on this machine and needed it back ASAP, so I bought
    a Tiger MPX S2462 (AFAIK this is essentially a new revision of the
    board I had). In dismantling the old setup, I found that *three* of
    the +5v leads on the ATX connector had welded themselves to the
    motherboard (I would've hoped that the fuse would blow before this, or
    at least that the connector would also be capable of taking the
    juice). Funny thing is, the PSU itself seems fine, aside from the
    charred 20-pin connector which we had to pry off the board. Still, I
    bought a new 400W PSU.

    (If it's relevant, I originally had a [probably no-name] 400W PSU in
    this machine. It went up in smoke [literally!] a few months after I
    bought it; I just pulled the Antec from another box and went on with
    life, not at all thinking that maybe it was part of something larger.
    Then again, maybe it wasn't.)

    Right now, I've got the new board and PSU plus the old video, sound,
    keyboard and mouse, and it's working great. My concern now is whether
    this might happen again -- if the culprit *was* something beyond the
    other board, it may trash this one as well, but I can't find anything.
    My dad and I went at the board with a multimeter but didn't find
    anything suspicious ourselves, and none of the other components in the
    machine appear blackened or anything which one would normally
    associate with an unusual overdraw of current.

    My cursory research on Google/Usenet hints that maybe one of the VRMs
    on the board failed. I know that one was running much hotter than the
    other, though I've also read that this may be a byproduct of its
    unfortunate location on the board. I also just noticed that the same
    is true of this new board (it's at 54C -- that seems a tad too warm
    [am I supposed to add cooling to that or something?]).

    Also, should I bother trying to RMA the old board? (Not that I could
    do much with the replacement except to have a spare on hand...)

    Any input on the subject would be appreciated; if further details
    would be helpful, feel free to ask and I'll try to provide what I can.


    - Shane
    Shane Beasley, Jul 25, 2003
    1. Advertisements

  2. Interesting... That seems to explain everything. I'll go with it.
    It's been connected for months on end, apparently with no problem. For
    it to lose contact after all that time seems weird, or so I'd think. I
    do know that the connector itself was in all the way (both ends were
    flush) when I found it melted, which made it difficult to get leverage
    to pry the thing off.
    Couldn't be further (Chicago, Illinois, USA, nearly halfway between
    the East and West coasts and closer to Canada than Mexico), nor can I
    imagine anything else in the vicinity that could be responsible (and
    I'm trying).

    Anyway, thanks for the enlightenment about the connection. I'll keep
    it in mind in future ponderings.

    - Shane
    Shane Beasley, Jul 25, 2003
    1. Advertisements

  3. I had the enermax431 and it was surviving for a while, however my
    motherboard now wont boot, Fortunatley my connecters weren't smoked,
    luckily i grabed a 2460 off ebay for cheap and it's working great with my
    new (non enermax) 450 power sup, and i have even COOLER temperatures,i
    can't even believe it they are 43 and 45 CPU1+2 that is ridiculous for summer
    i average in themid 50s, the ebay 2460 did have an extra fan on the
    northbridge, maybe that is doing it, well i don't care it's cool.
    del.puncutation_&_chars.except-this-ch.u&r*, Jul 26, 2003
  4. Shane Beasley

    Tech God Guest

    Tech God, Jul 27, 2003
  5. Shane Beasley

    Paul Murphy Guest

    Not the official fix though.

    Tyan have developed a Y splitter cable which they are providing with some
    (all?) RMA'd S2460s. They have also replaced the originally specified ATX
    connector on the motherboard with a newer higher spec design ATX connector
    (which looks the same as the old one). These new connectors are capable of 9
    A per pin and with the assitance of the Y cable, those PSUs which only have
    ATX connectors capable of 6 A per pin, can use one HDD connector in
    conjunction with the motherboard connector to provide all the current
    required without a meltdown. I found this out through a phone converstaion
    with a Tyan Tech. The other "fix" is just to use a PSU with excellent
    quality connectors in the first place.

    My concern with doing a "home" fix is it will void your warranty for any
    future problems.

    Paul Murphy, Jul 27, 2003
  6. Not in the United States it won't. If you have to make modifications to
    get the device to operate as intended, they can't void the warranty unless
    they can show that your modifications contributed to the failure.

    Otherwise, car companies could put "do not tamper or your warranty is
    void" seals on the hood of your car and force you to go to the manufacturer
    for every oil change and wiper fluid refill.

    The law is called the Magnusson-Moss act. If your motherboard needs a
    repair to fix a problem or workaround a problem, you can do it any way you
    want, the manufacturer cannot require you to use their parts of service. If
    you want to fix it yourself, you can fix it yourself.

    Check your written warranty. If there are exceptions, they are required
    to be noted.


    David Schwartz, Jul 27, 2003
  7. It is definitely the case that if you don't know what you're doing, you
    could fry your board in the process of attempting this mod. In that case,
    the manufacturer isn't likely to be willing to give you a free replacement.
    You could just say that you had modified it and then modified it back.
    Since there is known manufacturing defect in the board, I don't think
    they're going to play hardball.
    Definitely. But if you're competent to make the mod, I wouldn't
    unreasonably worry that you'll lose warranty coverage. However, unless you
    have some really good reason, you should definitely let the manufacturer fix
    it since they're willing to. (Will they repair/modify any board shipped to
    them, inside or outside of warranty?)
    Changing the ATX connector is not an easy job either. Anyone who
    couldn't do the other mods couldn't do that either, at least, not safely.
    But I definitely agree that's the best fix, other than redesigning the board
    to run the VRM's off the 12V.

    David Schwartz, Jul 28, 2003
  8. Shane Beasley

    Arno Wagner Guest

    But the problem could be transferred to the 12V lines...

    Actually it is not quite that simple. Since the VRMs for the
    CPUs are switching mode, they basically do not draw a current
    but a power (in a certain voltage window). That means at 12V
    they draw roughly 42% of the current they draw at 5V input.

    However there is only one +12V in the ATX connector, while there
    are four +5V lines! That means the single 12V line gets to
    carry 166% of the current, and consequentially gets 275% of
    the power drop that was enough to melt the 5V lines! (Power
    goes with the square of the current in a resistor.)

    That is why e.g. the Tiger MPX absolutely requires a conventional
    plug with no other load to be connected to the mainboard. This still
    gives 80% current, equal to 64% of the "melting load"! Better, and
    im my opinion the only good solution, is the additional +12V
    connector that has two additional +12V lines, finally bringing the
    current per line down to 55% and the power drop to 30% of
    the melting load. Of course this is a very rough estimation.

    They should have left the 3.3V lines completely out of the ATX
    connector and instead have put in three additional +12V lines....

    I hope that sheds a little light on the background of things.

    Arno Wagner, Jul 28, 2003
  9. Shane Beasley

    Paul Murphy Guest

    Thanks for that excellent explanation of what's going on there. Is there any
    reason why the 3.3 Volt Bus couldn't be utilised to power the CPUs as well
    as the +5 Volt rail (shared somehow)?

    Love the sig - I have an improvement to it - wonder if it applies "The more
    corrupt the state, the more numerous the lawyers" - Murphy ;-)

    Paul Murphy, Jul 28, 2003
  10. Shane Beasley

    Arno Wagner Guest

    With 3.3V only you have even higher current and filtering the input
    becomes difficult. In addition many PSU deliver less or the same curent
    on 3.3V as they do on 5V, i.e. 3.3V has only 66% of the power 5V
    can deliver.

    As to shared, that would be possible. However it would be expensive,
    because it would basically require a 3-phase or 4-phase swithcher
    that could deal with 2 different input voltages. While feasible,
    this is a significantly harder design. One big problem with
    switchers is stability. In this case you would have two different
    power sources with different dynamic capabilities, making things
    massively more unstable. And in addition, the input filters for
    3.3V are even moer expensive than those for 5V, since they need
    to deal with a higher current. Prpbably even more so for a mixed-
    input design. That is one other reason why using 12V and an
    additional 12V connector makes very much sense. The higher the
    voltage, the cheaper the input filters.
    That is probably the modern version. ;-)

    Arno Wagner, Jul 28, 2003
  11. Shane Beasley

    Arno Wagner Guest

    One option I did not think of would be to power one CPU from 5V and
    one from 3.3V or 12V. This would solve the connector problem.
    However it would make the design "asymmetric" and cause twice
    the effort to work out the VRM bugs.

    Arno Wagner, Jul 30, 2003
    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.