Back to Article

  • boozed - Tuesday, March 19, 2013 - link

    I've repaired two Dell XPS M1730s to working condition by "re-baking" the GPUs, though I don't know how they're going now; both owners replaced them not long afterwards.

    The common component here is a higher end mobile Geforce 8000 series GPU. You'll remember they had some horribly dodgy batches a few years back. Surprisingly though the ones I fixed were well out of warranty by the time the problem manifested.

    Depending on the circumstances the heat either reflows solder balls that have cracked or become separated underneath the GPU package, or for solder that has an excessively high Tin content, melts the whiskers that have grown between the solder balls and caused shorts.
  • Galcobar - Tuesday, March 19, 2013 - link

    200C, not 200F. Significant difference between half-melted solder at 200C and not-quite-boiling water at 200F.

    I resorted to baking my desktop GPU a few times -- an 8800 GTS 640MB, as it happens.

    It generally is a short-term fix. At 200C, the temperatures are generally too low to completely liquify the lead-free solder now used in electronics, so it's unlikely to resolder any joins which have cracked due to heat cycling. However, the solder does soften enough to close those cracks back up. You end up with a mechanical joint rather than a chemical one, which means more heat cycling can open those cracks back up.

    I kept that card alive for two years with periodic rebaking. Probably would have been required less often if I could have found proper replacements for the stock thermal pads. The failure was invariably in the card's memory, which the heatsink doesn't actually touch -- the thermal pads have to span a good millimetre or two.

    The point of doing it in an oven, rather than with a heat gun, is to get a slow and even rise in temperature. This way you can avoid burning any component. That, and you'd have to heat most of the card anyway since you don't know where exactly the broken connection is hiding.
  • dragosmp - Tuesday, March 19, 2013 - link

    Agreed it's a temporary fix. On the laptop board it was so temporary that the laptop had to stay on after baking/remounting (still warm). Once you turned it off and it cooled it wouldn't start - I suppose the inner solder joints would crack again due to the thermal stress. It had a 7xxx Nvidia chipset, from the same soldering lineage as the 88xx.
    ...and it was 200C (though I used 180C).

    Since it had to be done a few times on that laptop board I found an easier way: unplug the fan from the motherboard and leave it cook. The CPU turns itself off at over 100°C and the chipset continues to cook itself for a few seconds before the board shut down. This made the whole thing a 15mn fix.
  • marc1000 - Tuesday, March 19, 2013 - link

    that's one brave cooker! Reply
  • Wolfpup - Tuesday, March 19, 2013 - link

    Hmm, I hope newer GPUs don't fail like this. I have a mobile Geforce 9650GT that hits temps of 105c and crashes probably once a week if not more on average, but then it's I think 55nm and basically the same as the 8600GT m and the like.

    I've got a newer desktop Geforce GT 430 though that's literally been running Folding @ Home (which pushes it to the max) 24/7 for 2 years, sooooo not sure if newer ones are just more reliable, or of notebooks just have much much worse cooling.
  • JarredWalton - Tuesday, March 19, 2013 - link

    I think newer GPUs have learned from past mistakes (e.g. the 8000M series from NVIDIA), but that 9650GT probably needs to have new thermal material applied because anything over 95C is far too hot for normal use. Reply
  • epoon2 - Tuesday, March 19, 2013 - link

    yeah I read about this somewhere else, on Dan's Data. It's related to the soldering.
  • shriganesh - Tuesday, March 19, 2013 - link

    Wow! Nice one! I would be really interested to know about the HDD transplant method! Reply
  • marc1000 - Tuesday, March 19, 2013 - link

    yeah me too! GLHF on that hdd transplant, Jarred! Reply
  • Orbs - Wednesday, March 27, 2013 - link

    If that works, I'd love to see a video of that too :). Good luck! Reply
  • Freakie - Tuesday, March 19, 2013 - link

    Hey Jarred, I've done the platter transplant (successfully) and here's a few tips: powder free gloves, non-cloth surface and environment (wood, plastic, metal desk/table/counter ect), facemask, synthetic t-shirt that covers as much as possible, freshly cleaned hair and maybe even head covering, and a can of compressed air.

    Careful with the compressed air, don't want the chemicals coming out and depositing, also make sure to be extremely careful with the needles, don't go playing with them just because they are there! You'd be surprised at how easily you can screw up the needles as well as the platters! And make sure to keep track of all of the filters in the drives. There are usually two white filters on the end opposite to the drive heads that come out VERY easily so keep an eye on them and don't let them go flying everywhere and collecting dust and particles.

    Also, before swapping the platters, try swapping the PCB's on the hard drives if you aren't 100% sure it's a mechanical problem. Sometimes it's just irrecoverable damage on the PCB that kills the drive and it's worth trying to swap them. If the drive is discoverable after doing that, I'm not sure how the data will or will not appear, might have to run a deleted file recovery program like Recuva. But I'm not 100% sure what happens to the data if you just swap PCB so I'd do a little bit of research into it first to make sure not important parts of the platters will be wiped xP
  • Paladin21 - Tuesday, March 19, 2013 - link

    I've done a PCB swap a couple of times, mostly with drives that had obviously blown controller circuits on the board. The most annoying part of the process is finding the *exact* same drive revision and size for a donor board. If it goes well, your data is immediately accessible again, no need for a data recovery solution. If you've got a common drive type, I'd suggest trying the PCB swap out first, as it only takes a couple of minutes to verify that you actually have burned out motor or rotor damage instead of having a dead controller signal. Reply
  • JarredWalton - Tuesday, March 19, 2013 - link

    The drive is a Western Digital WD2500JB -- EIDE if you can believe it. I've ordered a similar drive (not sure about firmware or anything else, but it's the same model) off Amazon that should arrive this week. The symptoms right now are that the drive powers on and spins up, then about eights seconds in there's a "click" and it sounds like the power resets. Immediately after the click, the drive repeats the process, so it never fully spins down but it does start to slow down for a moment. After the first click, the process repeats about every three seconds.

    My guess at this stage is that it's potentially a bad circuit somewhere, though it could be a bad drive head as well. We'll see how it goes later this week, but the recommendations on gloves and clothing and such are welcome!
  • Paladin21 - Tuesday, March 19, 2013 - link

    Since your drive spins up and seeks, I'd definitely try a board swap before doing any "brain sugary" on it. I've had several in the past with similar symptoms that worked (at least long enough to back up data) after doing a PCB transplant. With your symptoms, I'd guess that the issue is in the controller board, not the drive itself. Issues with the internals of the drive usually result in either the platters not spinning (burned out bearings/motor), or massive internal damage (crashed head). If you're spinning, neither of those issues should be your problem. Reply
  • JarredWalton - Tuesday, March 19, 2013 - link

    Yeah, sort of what I'm thinking. It's not the click of death (where the drive head seeks back and forth), and it's possibly a short somewhere in the power circuitry. My donor drive should arrive in the next day or two I think.... Reply
  • JPForums - Tuesday, March 19, 2013 - link

    It generally is a short-term fix. At 200C, the temperatures are generally too low to completely liquify the lead-free solder now used in electronics, so it's unlikely to resolder any joins which have cracked due to heat cycling. However, the solder does soften enough to close those cracks back up.

    You are correct. The solder will not liquify at 200C (and that is actually preferable in this case). Left alone this will not resolder any joints. However, it is possible to close the gaps enough to facilitate resoldering. You need to provide even downward pressure on the chip in question. Any horizontal pressure will push the chip out of alignment. The best way I've found is using sand. You have to make sure the board is level while baking, but you can use sand on the entire board if you don't know which chip is the issue. Keep in mind that baking of any kind won't fix a bad chip, just a bad solder joint.
  • MrSpadge - Tuesday, March 19, 2013 - link

    > Keep in mind that baking of any kind won't fix a bad chip, just a bad solder joint.

    And probably not a broken HDD.
  • tipoo - Tuesday, March 19, 2013 - link

    Same idea as the various Xbox 360 heat tricks. I got a few extra months of use out of an RRoD 360 with a corrupted GPU by escalating the heating each time it failed again, first I just did the towel trick, then let it run with q-tips blocking the fans, then opened it up and let it run without the fans while using a high powered heatsink on the front and back of the GPU side, then eventually threw the motherboard in the oven. Each time got me a bit more use, but now it's perma-dead.

    Still haven't gotten a replacement, come on next gen!
  • Daniel Egger - Tuesday, March 19, 2013 - link

    You've just explained the concept of reflow soldering. There's nothing bad about this approach as this is also the way most components are nowadays banned onto a PCB. The typical problem why some connections don't stick properly is that mechanical shearing happens when two connected materials expand or contract differently when applying a temperature change. With GPUs there are actually three common cases why those connections might get lose:
    1) GPU is used outside the specified environment (i.e. gets too hot, or warms up too quickly)
    2) The GPU vendor screwed up in specifying the reflow process needed
    3) The PCB assembler screwed up implementing the specified reflow process

    Using a hairdryer or oven is a perfectly valid method fixing reflow problems, however it is also quite easy to do permanent damage. If it is done right though it might even hold "forever".

    Heck I know some people who bought broken iBooks on ebay in the hope to being able to fix them just by baking and reselling them. They have quite a decent success rate of about 90%.
  • faster - Tuesday, March 19, 2013 - link

    Excellent video. You have convinced me to never take apart my Gateway FX P-7805u.

    A word on the laptop though. My Gateway FX P-7805u has the C2D P8400 @ 2.66, 4GB DDR3, with a 9800m GTS with 1GB. I bought it in early 2009. The reason that I still use it today is because I was able to easily install the early Intel 80GB SSD in the second open hard drive bay and put Windows 7 pro 64 on it. It is a dual boot system with Vista/7, although I can't tell you the last time I booted Vista. I also use the Vista drive as a data storage drive for the smaller Win7 SSD.

    My job entails hauling this thing all over Manhattan. Its a heavy beast and I have broken quite a few shoulder straps over the years and sent this thing tumbling to the ground. My chassis is cracked with small chunks missing in four seperate places and the two of the main support screws on the left are broken through. I can seperate the chassis vertically over an inch on the left front side by the DVD drive by lightly lifting up on with my finger. That being said, I it still works perfectly and I use it nearly every day.

    The SSD and Win7 make it a snappy performer. I scroll huge PDF's smoothly in presentation and I can play Battlefield 3 at medium settings. This machine has also logged many hours of BF2 and BF2142 without any problems. Not too shabby for a 4+ year old laptop. The downside is the battery life. If I get 3 hours on light usage, I'm lucky.

    My original purchase price was $1500. I received good value for that purchase and this would be a great laptop for someone on budget to pick up used even for modern gaming.
  • PCMerlin - Tuesday, March 19, 2013 - link

    I have a very similar model to this laptop, the P-7811 FX with the 1900x1200 screen, which is one reason why I still use this over any "new" offerings with lower resolution displays.

    I went one step further to help in making this a more permanent fix, rather than just the temporary fix that most here are mentioning. If you take a very small copper shim and insert it on top of the video chip, with a layer of thermal compound (I used arctic silver at the time) on BOTH sides of the shim to ensure 1) a better thermal connection to the heatsink and 2) a little extra pressure on the chip itself to keep it in place. I also removed and replaced the thermal compound on the CPU, and added thermal compoud to the Intel chipset (that gets hot, too!) and replacing the thermal pad that goes on top with another copper shim similar to what was used on the graphics chip, but thicker to make up for the loss of the pad.

    Another thing not mentioned here is the absolute need to do a throrough cleaning while you have the laptop open. Get any dust built up blown out of the way, and especially the fans - you might not be able to see it, but quite a bit of dust can build up in the vents behind the fans.
    After doing this, I have noticed that the laptop runs much cooler than I can recall it ever running before, which I hope will at least keep it going until more laptops come out with retina-level display resolutions.
  • JarredWalton - Tuesday, March 19, 2013 - link

    Oh, the dust bunnies living inside this thing were legendary! I cleaned it out with a blower first, which made a huge puff of dust, but that was outside. Even after doing that, however, when I got to the actual heatsinks on the CPU and GPU I found there were still large chunks of dust bunny to remove ("dust bunny giblets"!) So hopefully the combination of removed dust buildup, new thermal grease, and hair dryer reflow soldering will do the trick for at least a few months, preferably a year or more. :-) Reply
  • Iketh - Tuesday, March 19, 2013 - link

    Please do an article on your platter transplants, successful or not! Reply
  • wicko - Tuesday, March 19, 2013 - link

    This is how I revived my launch PS3 temporarily - long enough for me to back up all my data. Reply
  • rivethead23 - Wednesday, March 20, 2013 - link

    I still have this exact same notebook. Was a great machine and will still play some worthwhile titles. About 2 years ago I turned it on and nothing.. No lights no fan spin nothing. Figured it was cooked and began shopping for a replacement. Decided to call Gateway to see about buying a replacement motherboard (as the power supply was outputting the correct voltage). They offered a flat rate repair of $199. All said and done they replaced everything but the screen keyboard and housing. Still running today. I suspect this GPU issue was likely at fault.. If it dies again I might have to try it. Reply
  • plcn - Wednesday, March 27, 2013 - link

    funny, i've seen a ps3 YLOD "fixed" the same way... Reply
  • compcons - Wednesday, March 27, 2013 - link

    I had an HP pavilion dv4 die due to failed GPU. I checked around and found SRR manufacturing ( They basically disassemble the laptop, remove the GPU, reball/resolder, check with some xray equipment, and put it all back together. Price? About $110. The huge upside is the guy is 15 minutes from my front door. Well worth the cash, low risk, and very effective. I spoke with Dale for a while about the hairdyer and oven tricks. He showed me the results of the DIY "fixes". Sorry to sound like a commercial, but I was thrilled with the result. Something to consider if you want to really save a PC.

  • theohtechguru - Friday, March 29, 2013 - link

    I've used the oven trick (12min @ 445F ) with lead free no clean flux on multiple devices. Video cards, Laptops (Always the northbridge) , Xbox 360's and PS3's. Every time was a success, I've always took great care in insulating the cap's and plastic components though. Reply
  • 7Enigma - Monday, April 01, 2013 - link

    I've fixed about $3000 of equipment resoldering bad joints on everything from a neighbor's car (clock and heater panel), to my sister's LCD TV, my car (body control module), and other electronics. So many failures in the last 10 years due to (my hunch) lead free solder (more rigid and likely to crack under temp swings) and cold-soldering (faster heating/cooling). Most of these defects can be easily seen with good lighting and a magnifying glass. Then it's as simple as reheating a bad joint with a good amount of flux and letting it cool back down. Reply

Log in

Don't have an account? Sign up now