•  
      CommentAuthorSpode
    • CommentTimeJan 24th 2011 edited
     

    My good friend Karl's laptop broke a few years back - leaving a garbled mess on the screen. We pin-pointed it down to the GPU itself, but didn't have a clue how to fix it.

    A couple of weeks ago I was reading about reflow soldering, and was putting an order into Farnell. I picked a 1500W heat gun up for around £11 thinking that at some point I'd get to use it. I mentioned this Karl and as it turned out - he still had his laptop, so bought it around.

    In hind sight - I probably should have watched more YouTube videos in advance - but I knew the principle and just went at it alone. I had a spare graphics card that was broken and decided to use this to experiment on - after all - we had no idea how long it would take for the solder to melt, and if there was potential for damage.

    As you can see in the image above - GPUs are surface mounted using a Ball Grid Array (BGA). What we are trying to do, is remelt this solder to hopefully recreate the dry joint that has developed.

    I used my heat gun on setting 1 - so I assume around 750W. I kept it 2-3 inches away from the PCB and applied air in a circular motion. I didn't want to scorch the PCB or surrounding components - it's a hell of a lot of heat.

    Whilst testing our old graphics chip - even applying heat to a focussed area - the whole PCB got too hot to hold.

    As well as heat - I used liquid flux. I picked this up a while back from eBay in the hope that I'd have a use for it some day. It's messy (gets everywhere) but it gets under the chip properly and should help make a better joint.

    We found a memory chip could be heated enough to come off the PCB in around 13 seconds - while a smaller surface mount chip (like an AGP/PCI-E bridge) took more like 30. We decided that around 1 minute would be enough for the Nvidia GPU.

    We had to pull the whole motherboard out of our laptop in order to unscrew the heatsink arrangement protecting the GPU and memory. Once off, we cleaned up all the thermal paste and applied the liquid flux. I heated the GPU for 60 seconds and the memory on each side of the PCB for 20.

    The biggest risk factor we found was accidentally knocking surface mount components when you've been applying heat to an area - so once heated, you need to leave it for a good 30 seconds before you want to go touching the PCB.

    We're really glad we experimented first, as it gave us a much better idea of what we were doing on the real thing - especially as when we put it back together and booted up - it worked!

    Looking at this YouTube video - this chap didn't use any flux, didn't remove it from the chassis and used a very focussed end (perhaps a better idea). But what is interesting - is reading the comments - specifically that it's the use of lead-free solder that is causing this problem, and also that this is only a temporary fix and it will happen again.

    Let's hope that by using Flux and applying the heat we did - our fix is a little more permanent, but I doubt it. I think Karl intends on using it for 2D only use from now on...

  1.  

    The heat gun is used to heat the solder underneath the chips enough to "reflow" them and restore the faulty connections. This is best done using a BGA reflow machine, but not many people have access to one of those, let alone the correct template for a PS3 chip.

    •  
      CommentAuthorSpode
    • CommentTimeJan 25th 2011
     
    Posted By: Trainersroom

    This is best done using a BGA reflow machine, but not many people have access to one of those, let alone the correct template for a PS3 chip.

    Despite being a spammer, you do raise a good point. From what I saw the best fix is to completely remove the chip and reball it using lead solder. From our experiments of simply removing the memory chip - I wouldn't want to try and put one back on!

    This is supposed to be a cheap and quick fix...

    • CommentAuthorSirkent
    • CommentTimeJan 25th 2011
     

    Well, so far everything looks fine. However I have noticed that the GPU heatsink gets too hot to touch when the laptop is running even only for a few minutes. Clearly, the cooling solution for this particular laptop is quite poor.

    I've heard of this sort of solution fixing Xbox 360s permanently, but who knows in this case.

    •  
      CommentAuthorSpode
    • CommentTimeJan 25th 2011
     

    I think the Nvidia drivers do some good power saving things - but it might be worth underclocking it if you're worried?

    • CommentAuthorSirkent
    • CommentTimeJan 29th 2011
     

    Well so far everything seems absolutely fine. The nVidia drivers did help - the temperature is fine now, although it seems odd that the default when there are no drivers installed is for the chip to go crazy and heat up! SpeedFan tells me that the GPU temp is usually no more than 60 degrees, which seems good.

    Only time will tell if it is set to last, but if I don't over-tax the graphics then I don't see why it should break again.

    •  
      CommentAuthorSpode
    • CommentTimeJan 29th 2011
     

    Without the drivers, the chip defaults to maximum clock speed/fan speed. When the drivers kick in - it starts doing intelligent shizzle, like different clock speeds for 2D and 3D.

    The clock speeds are set in the BIOS for the card - so it makes sense that is what they would default to unless told otherwise!

    Maybe that's what happened before? Did you have drivers installed on your Ubuntu install?

    • CommentAuthorSirkent
    • CommentTimeJan 30th 2011
     

    I did indeed have the drivers installed. I think that the GPU cooling is simply inefficient, or as you pointed out, the solder is of poor quality.

    •  
      CommentAuthorcoyote
    • CommentTimeFeb 6th 2011 edited
     

    Mmm very interesting, although I totally agree with the RoHS ruling it's going to cause many problems such as this. I assume it's to do with cost, but aluminium based solder has got to be the worst metal to use as a substitute to lead as it's susceptible to fracturing, it's so brittle when worked by bending or heating and cooling.

    I bought a reel of the stuff for general use repairing and making electronic equipment and found it to be absolutely awful to use I had to be very careful to get a decent joint and also had to up the temperature of my soldering iron, not good.

    I saw on special offer some pure tin solder with 5% silver, so I thought I would try it. Yes, even on offer it was around £35 for half a Kilo reel, but what a difference. Good joints were very easy and at a much lower temperature. This stuff still isn't a flexible as lead based solder, but much better than aluminium. Only time will tell how good it is, although at at it's normal price circa £100 a Kilo it should be great, but probably to expensive for the manufacture of commercial PCBs

    •  
      CommentAuthorSpode
    • CommentTimeFeb 7th 2011
     

    I was talking to my friend from Nvidia the other night - and apparently they replaced a whole bunch of these 8 series units for FREE for the same problem. Perhaps you got some poor information from Zepto Karlson?

    • CommentAuthorSirkent
    • CommentTimeFeb 8th 2011
     

    Well, it was just out of warranty and as we saw, would have involved replacing the entire motherboard. Zepto also went out of business a couple of years ago, so maybe they weren't doing so well?

    •  
      CommentAuthorSpode
    • CommentTimeFeb 10th 2011
     
    Posted By: Sirkent

    Well, it was just out of warranty and as we saw, would have involved replacing the entire motherboard. Zepto also went out of business a couple of years ago, so maybe they weren't doing so well?

    I think this was the point my friend was making - warranty was irrelevant because it was an Nvidia flaw. Anyways - could be wrong, we were both a bit tipsy at the time :P

    • CommentAuthorSirkent
    • CommentTimeMar 5th 2011
     

    Oh well - it was good while it lasted but the laptop packed up today. The screen won't come on at all and the laptop isn't booting. It wasn't doing anything particularly strenuous before death and hadn't even been on for very long. I'd been very careful with it - hadn't played any games and actually used some temperature monitoring software. The GPU never went over 80 degrees AFAIK.

    This obviously lends a lot of proof to the idea that there is a fundamental flaw with the nVidia chips and that reflowing the solder is only a temporary solution.

    •  
      CommentAuthorSpode
    • CommentTimeMar 5th 2011
     

    Hang on - if it's not turning on at all... ?

    • CommentAuthorSirkent
    • CommentTimeMar 5th 2011
     

    The laptop is turning on, but the laptop isn't booting and the screen is staying off. Perhaps it's not receiving a signal from the GPU at all now, so it's not turning on?

    •  
      CommentAuthorSpode
    • CommentTimeMar 7th 2011
     

    That seems like quite a step back in one jump. I'd try a CMOS reset and unplugging anything that might be causing issues...

    • CommentAuthorSirkent
    • CommentTimeMar 7th 2011
     

    I tried a CMOS reset and tried removing a few items on Saturday. No change. I also went and bought a new laptop :-P.

 
Copyright Andrew Miller (Spode), 2008