NB47J Self Destruct, or The Incredible Shrinking Pushpins

[beboop]

The short version: my NB47J fell off.

The interminably long version: I was having some odd flakiness in Cubase. I ran Memtest86, after 3 passes it gave an error. I opened the case and swapped out the dimm.

That's when I noticed the NB47J was tilted up off the chipset (nforce3 on an Epox 8KDA3J). One pushpin had worked out, I pushed it back in, the other popped out. Then the one I'd just pushed in popped out too.

Both original pins no longer hold, they pop out of the motherboard mounting holes with little or no pressure other than the normal spring pressure. This is in a rackmount case, the motherboard is horizontal, gravity isn't a factor.

I had some spare, identical pins left over from an NB32 I'd glued in another system. These worked fine, popped right in and held.

What has happened here??!!! There's been no stress on the original pins, other than that of the springs. The NB47J was installed maybe two years ago, and not touched since. The system has never been moved. It seems like the pins have simply shrunk, maybe due to heat?

I reinstalled the original dimm, and am running Memtest86 again, so far so good, eight passes without an error now.

Has this happened to anyone? The system apparently ran at least a few hours, at around 40-50% load, with no chipset heatsink. It seems fine now -- is that just wishful thinking on my part?

[Bluefront]

Well that sure was a good stress test.....just what chipset was that? Another black eye for push-pin attachment methods. Sounds like the original pins were defective somehow.

[Das_Saunamies]

--I noticed the NB47J was tilted up off the chipset (nforce3 on an Epox 8KDA3J)--

Funny thing is, I've had push-pins fail on a graphics card but not on my chipsets. The card was some GeForce 4 series model with a factory-default cooler (PNY Ti4800 perhaps), but what I'm thinking is that horizontal positions are perhaps not as good for the pins as a vertical position. In a vertical position the pin acts as a lever, pushing against the motherboard slightly. In a horizontal position the only forces are the plastic's spread and the spring.

The forces are minute, but so is the amount required... more empirical evidence anyone?

And yes, I would concur with the pins being faulty if the NB32K ones hold.

[stevenkelby]

I use nuts and bolts on those heatsinks. Cheap insurance for a couple bucks. I put a little piece of electrical tape on the back of the mobo, poke a hole in it and push the bolts through, so the bolt heads don't short anything. M3 is the size, any length but I use 25mm

[Tzupy]

If your nForce3 is still 100% working, then you are quite lucky. Just memtest passing doesn't say much, since the memory controller is in the CPU. I suggest to run other diagnostic software, like 3DMark or maybe some IDE / SATA testing, to determine if the whole chipset is OK.

[beboop]

yep, some stress test -- for me and the system! It's been a really great system for its purpose, and it'd be really hard to replace now without replacing basically everything (AGP, DDR, IDE drives, etc).

The odd thing is the original pins held fine for a couple of years. Only thing I can think of, I did swap out the processor not long ago. Removing and reinstalling the XP-120 would have flexed the motherboard a bit, maybe enough to loosen those pins slightly, and time and spring pressure did the rest. Once the pins had pulled through the motherboard holes the barbs may have been compromised so that they no longer hold. Who knows.

Bolts, great idea! -- actually I dunno why these didn't ship with them, couldn't have added that much to the cost, and the consequences of failure are so dire. Frustrating that a motherboard can be destroyed by a cheesy plastic pin failure.

I may be lucky -- hard to know, but it passed BurnInTest at the default settings, and Cubase seems fine.

well, no more cheesy plastic push-pins for me! Hope this lesson doesn't turn out to be too expensive

[Das_Saunamies]

Bolts, great idea! -- actually I dunno why these didn't ship with them--

Cost isn't probably the only issue. Bolts are more dangerous and less handy. There is actual force in play with a mechanically unforgiving system like that, and they are a hazard in (micro)electronics. Pushpins are also easier to install, as there's no need to remove the motherboard.

If I had to choose, I'd still take the pushpins for a chipset heatsink.

[geekbanter]

I've been there too, only I ended up with a dead motherboard

viewtopic.php?p=311502&highlight=#311502

[beboop]

well, nylon bolts then, with the threaded length limited to a safe range of pressure, or a plastic spacer-- but yeah, push-pins are easier The best argument for bolts though is they won't fall off.

I kind of can't believe it, but everything seems fine, no problems with anything so far (what's the proper emoticon for "kind of can't believe it", anyway?). Guess I'll look around for a spare 754 board just in case, but apparently I really lucked out. That's some tough chipset!

geekbanter, aaaggghh, that's a painful story! Especially the part where it seemed OK at first. That's where I am now.

oh well, there are still a few 754/AGP boards on Newegg -- none that would have ever been my first choice, but at least I can salvage the rest of the parts if I have to.

[Das_Saunamies]

Non-metallic bolts would be acceptable for me, but I'd still be too lazy to tear out the motherboard once again. I'm betting they would be more forgiving to force as well... breaking before the chip gets chipped that is.

There was an easy fix to the graphics cooler problem that I just remembered: wedges. I think it was some leftover paperboard that I wedged in the pins. Secured them in place nice and tight. I also just remembered there being these mighty nifty pins where there's a center pin that wedges the ends apart when pushed. Now those would be the ultimate in ease, and I dare say security.

Anyway, fingers crossed for the survivor chipset! And condolences for the loss of the other one. (kind-of-can't-believe-it = ?)

[beboop]

yep, the type with the center-pin expander would be much more reliable. Unfortunately that's not what Zalman gave us with NB47J.

After reading geekbanter's story, and a few others like it, I'm realizing the explanation for what happened is probably nothing too exotic. Apparently the pins that shipped with the NB47 are just crap.

Maybe there should be a sticky on push-pin failures -- it seems to be one of those things you only find out about when it happens to you.

In the interest of fairness, I suggest "Push-pins: THREAT or MENACE??!!!". you know, addressing both sides of the issue...

[Das_Saunamies]

In the interest of fairness, I suggest "Push-pins: THREAT or MENACE??!!!". you know, addressing both sides of the issue...

Seriously though, I'm in favour of gathering established faults like this in a sticky. There's already one for whining components.