Ever have hardware resurrect itself?

[Rob Williams]

As I mentioned in our review of Sapphire's HD Radeon 5750 card, the particular sample I had "died" on me before I could finish testing. I'm not sure what caused an issue, but it happened. I tried to "revive" it, but was unsuccessful. I installed other graphics cards to see if it was an issue with the motherboard, but nope. After I tested out those other cards, I plugged the HD 5750 back in, and still dead.

So earlier today, I was straightening up a bit and noticed I hadn't chucked the card like I meant to. I am not much of a pack rat, and I hate things lingering around that are useless. But, I'm glad I didn't toss it, because it turns out that it revived itself. I plugged it into the PC, and it booted up just fine. This after it refused to work, even after fiddling around with it for an hour or two.

I should stress that this is the exact same PC, and nothing changed component-wise except the graphics cards I test. This also isn't the first time this has happened to me, but I can't quite pinpoint the other example, as it was a few years ago. I just remember that it happened (chances are it was also a graphics card).

Have any of you guys had this kind of thing happen, or know what could possibly cause it? I should note that absolutely no overclocking was involved. It "died" before I had the chance to get to that.

 

[Psi*]

The other part of the test that you might have done or just not mentioned is to put the card in another computer to see if the problem followed. Without that it is difficult to tell with certainty (in my opinion) what is the culprit.

 

[2Tired2Tango]

Have any of you guys had this kind of thing happen, or know what could possibly cause it? I should note that absolutely no overclocking was involved. It "died" before I had the chance to get to that.

I've seen this a few times. Usually it's dirt in connectors that causes it. Although it may not look dirty it only takes a hair or a fleck of dirt in the right place to prevent a contact from making on most edge type connectors.

Try cleaning the edges of the boards with one of those pink pencil erasers... and rub pretty hard, you want to polish the gold.

Spray the card slots on the motherboard with contact cleaner or rubbing alcohol then pass a soft tooth brush up and down their length a couple of times... or spray then insert and remove a board several times to displace the dirt.

When you think about how these connectors work you realize that even the most tiny bit of gunk can prevent the spring connectors from mating with the card edge.

Another possibility is bad solder on the card itself. Grasp the opposite ends of the card while it's inserted and working and twist very slightly, if the card dies you probably have a bad solder point on one of the chips.

 

[Doomsday]

i think it just needed some alone time! time to rethink about its life, its purpose! and once it did, it revived it self to do thy bidding!

 

[Tharic-Nar]

I would have to second the bad solder option. If it was dust or some such, the device wouldn't have worked to begin with. Just take inspiration from the console market, all those consoles failing because of the graphics unit (360 and PS3), due to poor cooling and bad soldering. People have been stripping their consoles down, then using a heatgun over the main chips to remelt the solder. For the most part, it seems to work. (if only temporary)

I make a habit out of re-inserting things multiple times and/or cleaning them with alcohol before using. I had bought some new memory a few years back for a new system, put everything together, turned it on and the system beeped violently at me. So there was me in a disgruntled mood wondering which of the 7 components could be DoA (anyone who's had to deal with this will know the feeling... ). After some quick browsing of the net, someone suggests re-seating the memory a couple times. Low and behold, it worked.

 

[2Tired2Tango]

I would have to second the bad solder option. If it was dust or some such, the device wouldn't have worked to begin with.

Yes, it is likely that a dirty connector will fail right away... but I've seen time play some weird tricks. Fan vibrations, physical movement, oxidation, etc slowly jiggle things around and connections do break.

Card edge connectors, for all the development they've had remain one of the most failure prone parts of any electronic circuit. They're very convenient but not a very good connection.

 

[Tharic-Nar]

Card edge connectors, for all the development they've had remain one of the most failure prone parts of any electronic circuit. They're very convenient but not a very good connection.

Which begs the question, what suitable alternative is there and/or how would you improve the existing method? For the most part, they are a necessary evil, they provide an interface for a very broad range of adapters and connections. Contact density will be a problem down the line, just look at memory modules, they have 240 points of contact to make and if one of them doesn't, it could damage the module or at least prevent it from working. Retention systems for LGA's and BGA's still suffer the same point of contact problem. LGA and BGA systems could be put into motherboards for expansion, but then you're subject to the motherboards discretion on placement.

Developing a new slot would also break compatability. Creating a slot for a single device, like a graphics card is simple enough, which they did with AGP, but then you severely limit your expansion choice, then you end up with the good old days of motherboards with AGP, PCI and ISA slots. Going with a more universal slot type allows for a broader range of devices but there may be a performance impact of more specialized devices due to the generic nature of the slot architecture.

The only alternative is to build complete systems with everything soldered into place (laptops), but that leads back to the problem of upgrades.

If the slot contacts were multi part, like a brush, then it reduces the chance of debris preventing a connection. However, if one of these brush filaments were to break off ( as they do), then it could bridge between multiple contacts and short out. Also, if an insufficient number of filaments within the brush make contact, then those that do could be subject to over-current. Dust guards only work if both the slot and the device use them.

Wireless expansion is a possibility, but has significant hurdles to overcome. Bandwidth needs to increase about 20 fold for them to be deemed a suitable replacement for high demand devices, assuming current generation bandwidth requirements don't increase. Security is the next, making sure the device is only used by the designated system. And last, but probably most important, stability, can a device maintain 100% connectivity without being subject to interference. For a graphics card, this seems unlikely to be suitable, at least as a primary interface. As a secondary unit, then it could prove interesting, especially for those who are not technically savvy enough to dare open the big black box called a computer. But of course, as soon as wireless, graphics and streaming are said in the same sentence, the media companies will be jumping all over it, implementing 101 ways to 'secure' the device, even though it could only work up to 6" from the system.

I really can't think of anything else that could be suitable. Sorry, seemed a bit of a random post to make, that and i'm making a mountain out of a pimple....

 

[2Tired2Tango]

I really can't think of anything else that could be suitable. Sorry, seemed a bit of a random post to make, that and i'm making a mountain out of a pimple....

And this right after I watched "Evolution".... Ewww....

For pure durability purposes, pin connectors (like on IDE drives) are far more reliable than board edge. There are board mount versions of both male and female sides. They can easily be enclosed to minimize dust incursion. Pins can also carry more current than card edge connectors.... But as you point out the industry isn't about to change anytime soon.

 

[Psi*]

Technically speaking, connector design is about normal force or the force between the springs (the contact) and in this case the PCB pad. Low normal force & you don't have good wipe which is the cleaning action as the contact slides over the pad. Low normal force will allow foreign matter in the point of contact area and ... bad contact eventually. Vibrations are not necessarily a bad thing as that causes a sort of micro-wiping action.

Just as important is the plating. There are issues such as porosity ... imperfect plating of the gold so that there are pits exposing the under metal. Thickness of the gold as well as hardness (there is soft gold & hard gold). These contribute to durability or how many mating cycles a system could be rated for. I am not a fan of the erasure cleaning method, for what it is worth. Hydrocarbon based contact cleaners are preferred ... I haven't needed one for quite a while in spite of my recently posted pix ... I would not use alcohol.

With those 2 (there are more) brief characteristics in mind, we then have OEM purchasing departments. They are all about lowest cost widgets. Just thinking about those 2 characteristics above, does a connector sound like just a "widget"? The widget buyers are the bain of OEM warranty periods. Dell had a real problem a few years ago when marketing arbitrarily extended warranty to 3 years but were still using cr@p connectors (aka cheapest). They spent a lot of money before they got a connector spec to buy to.

BUT, Rob indicated that he pulled the card out, installed others to test the m/b, then re-installed the card that failed and it was still failing.

So after all that I said above, I don't think it the card edge connector. It would have been a really good thing to have tried the card in different PC or even a different slot? Can we do that with these PCIe systems???? Can you use a video card in PCIe x16 in the 2nd slot without the 1st being used????

And, did someone say something about contact lubricant? If you need it, use KY.

 

[Tharic-Nar]

For pure durability purposes, pin connectors (like on IDE drives) are far more reliable than board edge. There are board mount versions of both male and female sides. They can easily be enclosed to minimize dust incursion. Pins can also carry more current than card edge connectors.... But as you point out the industry isn't about to change anytime soon.

Don't let rob read that . While i would agree it provides a more secure physical and electrical connection, it poses a significant problem with regard to alignment and force required for the connection to be made. An IDE cable required 40 pins to make contact, this sometimes lead to difficulty in pushing the cable in, or indeed, removing it. God forbid that a pin was 0.5mm out, then you could bend the pin and spend the next half hour trying to fix it. Scale this problem up to a PCI-E x16 connector with 82 pins to align, or even memory with 240 pins, the force required to fit a connector would be like that of the push pins on the LGA775 Heatsink.

So i'm taking it to a bit of an extreme, but you see the problem. As Psi pointed out as well, there is the force of contact to consider, make it tight with good contact - but significantly increase the force required to insert the card, or make the contacts loose- increasing electrical resistance and the chance for dirt to build up but make it easier to insert/remove.

 

[Rob Williams]

The other part of the test that you might have done or just not mentioned is to put the card in another computer to see if the problem followed. Without that it is difficult to tell with certainty (in my opinion) what is the culprit.

I should have done this... I didn't even consider it.

I've seen this a few times. Usually it's dirt in connectors that causes it. Although it may not look dirty it only takes a hair or a fleck of dirt in the right place to prevent a contact from making on most edge type connectors.

I'm not sure if this was the case, because this is something I considered also, and I rubbed the contacts completely clean, and looked them over to make sure that was the case. The same goes for the slots themselves. I used canned air for those. If dirt was an issue, then it still makes little sense, because the card wasn't touched since it was last taken out, and other cards worked just fine. If it was dirt in the slot itself, you'd imagine that the other cards I installed would clear it out, since they worked fine. But no, after the fact, that HD 5750 still wouldn't work.

As for the bad solder, I'm not sure. Not going to push my luck with it, since I might need the card for future testing.

Just take inspiration from the console market, all those consoles failing because of the graphics unit (360 and PS3), due to poor cooling and bad soldering.

That's a good point. Still a tad strange on a brand-new card, but I wouldn't throw that issue past it.

I won't comment on everything here, but you guys all bring up some great points and ideas. One I'll toss into the "connector" mix is LGA or perhaps even BGA. The problem with these is that the connectors would all have to have flat bottoms, and that just might not work well with the PCB and circuitry. You'd imagine it'd be more reliable than card edge connectors or the pin-design.

Don't let rob read that .

Haha, well it is true. It still doesn't make me want to use IDE, and you mention the exact reasons why.