9550 - Dedicated Graphics Card fried - How to Fix? Where do I buy which part?

I have the Dell XPS 15 9550 which I bought January, 2016.
version: Intel Core i7-6700HQ, 4K touchscreen, NVIDIA GeForce GTX 960M

modifications: Samsung SSD 960 PRO M.2 1TB (MZ-V6P1T0BW); 2x16GB DDR4 2.666GHz (18-19-19-39) Corsair Vengeance; Killer Wireless-AC 1550 (Intel 9260)

I repasted the CPU and GPU. stacked thermal pads on top of the VRAM and PCH so that it's touching the aluminum body. undervolted using throttlestop

What happened?
I had a second monitor connected via Thunerbolt 3 port of my thunderbolt 3 dock (TB 16) - 4K external monitor (extended). I had 1 cord connected to the laptop itself - the Thunderbolt 3 dock's TB3 cord. About 3 weeks ago I was doing some web browsing/office on the external monitor while playing a video on the laptop monitor (Potplayer, MadVR, fullscreen). The computer was working pretty hard with the Fan spinning hard when all of a sudden everything shutdown with screens going black and computer off. This happens once in a while at full load.

Symptoms:
- There's no more dedicated GPU - NVIDIA GeForce GTX 960M. In device manager or HWiNFO64 the 960M is just gone (no longer listed). It only lists the integrated GPU (Intel HD Graphics 530). When I tried reinstalling the drivers the Intel driver installed fine, but when installing the NVidia driver it says you don't have a NVidia GPU

- When it comes out of sleep the Keyboard doesn't work, the mouse works. External keyboard doesn't work either. I'm not 100% sure if the peripherals work if I don't use the thunderbolt dock (i.e. USB 3.0 hub, and power cord).

- When I boot my computer up from off state the Keyboard doesn't work, but when I restart from the computer login screen the keyboard works and I can login. I have to boot and then restart every time I use my computer and can't let the computer go to sleep to use my computer.

- Sometime when I boot my computer, the computer screen is just black and I have to press the power button again.

Will things get worse?
I've been using the computer for about 3 weeks. I'm wondering if 1) things will get progressively worse and 2) is it doing additional damage to my WiFi card, RAM, and most importantly my hard drive. These are parts I'd like to transfer if I get a replacement 9550 or motherboard.

Solutions to get the sleep/wake function working again?
Is there a solution just to get the keyboard and peripherals working again when it comes out of sleep? It seems like I might be getting more battery life without the dedicated GPU which is a good thing but the computer is somewhat unusable if there's no sleep function. Most of the time, I let the computer go in and out of sleep and only reboot if I need to (e.g. windows update, new drivers). This way, I have everything open the way I like (documents, browsers, etc.) and can continue where I left off without losing my work.

Permanent solution to restore everything to normal? Where do I get the parts?
Is it just a matter of replacing the motherboard? I really like the way I have everything setup, software-wise. It'd be nice if I can just replace the hardware and not have to reinstall everything. It seems I can buy a used, tested and working, in good condition motherboard for $500 on eBay. Some sellers sell it for a lot more. There's a guy on Newegg selling it for $954 who doesn't have the best reviews. Are there sites that are more reliable/trustable than others? Any recommendations on how to go about buying an original motherboard part?

For future reference what caused this?
Normal wear-and-tear? Not enough thermal interface material during the repaste? Padding the PCH or VRAM mad things overheat instead of cooling it better? Using the Thunderbolt 3 port alone (1 cord solution) stressed the MB? Undervolting?

 

Not sure. First step is to back up everything.

GPU failure on these boards is rare.

- If you replaced the VRAM padding with too thick and/or too dense pads, then that could reduce heatsink contact with the GPU. Those are tough to measure and typically only one or two (of four total) of the VRAM pads were suspect from the factory.

- Too little thermal paste on the GPU... mininum amount of paste with adequate coverage is best.

==> In either case, you would see poor thermals in the GPU after the repaste. After repasting you tested thermals and looked at benchmarks around the board here?

Did you take apart the laptop recently before the incident? If so could you have dislodged some cables?

There are a few posts here recently of people losing the GPU in the 9550 and/or 9560; do a search and see if that helps.

You might try updating BIOS. Then consider doing a clean install. This could brick your system obviously so do at your own risk.

eBay is a terrible place to buy goods and there is poor buyer protection. Amazon can't be worse but there are not a ton of options at retail. For example, parts-people looks like a legit shop based on detailed inventory (I never purchased from them):

https://www.parts-people.com/index.php?action=category&id=143&subid=560&refine=motherboard

 

It does indeed sound like a mobo failure, but just make sure first before buying, as a mobo will run you at least 400 USD. On the bright side, you might as well buy the 9560 board + battery for an upgrade

 

If you plan on replacing the motherboard, you might just YOLO it first and try baking the motherboard in the oven. It has fixed GPU's in the past (I did with my ancient 8600GT). Take off everything you can, wrap important bits in aluminium foil and bake it for 10 minutes at 200C in the oven. It helps with reflowing solderpoints that sometimes break due to constant heating and cooling down of the GPU. You can find some guides on youtube.

 

Have people upgraded the motherboard to the 9560 successfully? Is it just a matter of swapping out the motherboard? Do I necessarily need to buy a new battery? the 97Wh battery did not fit well with the 9550 when I tried.

I'll try a search about this, hopefully there is a thread on it.

EDIT: it seems like a complicated process with a lot of misalignments issues to address:
http://forum.notebookreview.com/threads/my-9550-mb-upgrade-to-9560.810342/

indeed I will have to get a new battery it seems.

 

I didn't mess with the hardware before the incident. It's been a few months. I didn't look at the thermals too carefully but the GPU temperature temperature seemed reasonable when I did check after the repaste.

Unfortunately, it seems like the only place I can get the 9550 motherboard at a reasonable price is on eBay. Parts People are out of stock. Although PP has the 9560 in stock for $649.95. But I can find it on eBay for $459.99

 

I love this idea! Unfortunately the computer is still usuable as is and I can't risk having it completely unusable which I assume is a possibility if I put it in the oven. Which means I can't try this until I replace the Mobo and have the computer in working order. Maybe then, I'll try it

 

The oven trick is ghetto but works. I recommend isolating the heat with a heatgun over the GPU. When I used to do it all the time I would cover the rest of the motherboard with aluminum foil (shiny side up) so that I could concentrate the heat over the GPU. Worked most of the time.

 

Maybe a 9560 on Dell outlet for roughly $1140 makes sense (7700HQ, 4k, 512GB nvme, 16GB ram, W10 home)

Then sell parts of your old laptop (screen, wifi card, ram, SSD, case, etc)

 

Woah. I would buy this if I can just swap out my hard drive and RAM and everything will work like it's supposed to. Will it work? If I have to reinstall everything and go through that whole hassle, I'm thinking of just get the HP Spectre from Best Buy:

I really, really want longer battery life so I don't have to plug in laptop everytime I go to the library. I'd also like a graphics card upgrade so I can play video on the 4K laptop screen and do office work / web browsing on my external 4K monitor. The 9550 with its 960m card doesn't handle this too well and it's <4hr battery life is not cutting it.

Will the 9560 meet these demands? After doing some research it seems, the Spectre will give me good battery life but I'm always skeptical of reviews (obviously a lot of variables). I'm not sure about the NVIDIA GeForce MX150, I don't think it'll be that much of an upgrade from the 960m whereas the GTX 1050 4GB on the 9560, I'd expect a noticeable performance increase. Thoughts?

 

don't have a heat gun, but I do have an oven. I'm currently backing everything up as I contemplate this option. In your experience has putting Mobo in oven ever do more damage to it?

 

I mean it can but if you take everything off that will easily come off it will help. Just remember, plastic melts, lol. I would still cover the rest of the board with aluminum foil to help reflect some of the heat. I'm sure there are youtube videos on it. I'd go for it since you don't have much to lose (provided you want to use your dGPU).

 

Nope

I am lol​

 

If you can get a heatgun somewhere, it is a slightly safer method. Obviously putting electronics into ovens and heating them up isn't exactly a risk-free procedure. When I took my previous laptop out of the oven, the green stuff around the GPU chip was completely full of bubbles. I honestly thought it wouldn't work. But after letting it cool down for a couple of hours, the laptop booted and I had a working GPU again.

That said, if the laptop still works, I wouldn't risk it (though again, with a heatgun, the risks are lower as you can concentrate the heat only on the GPU chip, while protecting the rest with aluminium foil). I would only do it if the entire thing is dead, or if you plan on recycling the motherboard. I mean, you can always use it as a secondary laptop for non-GPU workloads, or you can sell it as-is. Hell, you can even still completely use it with an eGPU.

 

hmmm... decisions. Definitely will invest in a heat gun if I decide to reflow the GPU. Will report back if I go through with it. Thanks!

 

No, reflowing/reballing a BGA GPU doesn't work:

 

Your best bet of fixing this laptop is finding someone like Louis Rossmann (his company only fixes apple products) who does component level repairs on XPS 15 motherboards.

 

I don't have time to watch that but I assume it's going to talk about the limited time that it works for. I will tell you out of the ~100 reflows I did, I only had a handful come back within the 6 months I told them they could come back where I would do it again for free. I also took the time to explain the process as well as the risks and benefits.

 

I think Louis Rossman's point is that GPUs often fail for other reasons.

With the 9550 I haven't seen many GPU failures so no idea what would be the cause and the fix. . .

 

I'm liking the HP Spectre that I'm testing out. So I'll try out the reflow with heatgun once I move all my crap over to my new PC (which might take a while). Will report back.

I've been looking through the HP forum and I don't like it as much as the XPS forum. It's fragmented, especially with regards to the many Spectre versions out there. There's a lot more instructive guides/threads regarding clean installs, dealing with thermal throttling, troubleshooting known issues, etc with the XPS models.

 

That's because the XPS is so much cooler than the Spectre!!!

Actually I have no idea why it's so much more active here than the HP side. My guess is that the Dell design attracted a lot of attention (the 9550) and many enthusiasts, myself included, bought it.

I think they are great systems and wouldn't mind the one with the Vega since it's reasonably priced compared to the 9575, but the battery life isn't as good as I need.

 

I was thinking about this the other day also.

At release, the 9550 was rather innovative with little competition. It also had a lot of QC issues.

Eventually, the 9550 issues were largely sorted. Simple refreshes in the form of 9560 & 9570 were released. And a lot of competition entered the space. The XPS is still a nice laptop but times have changed. . .

 

He says heating up chips to reball doesn't work, but then he says it does work because heating up the chip itself fixes the issues. So...it does work, but just not on the part we expected it to work on?

I know for a fact that it can and does fix GPUs. No, it isn't a perfect solution, and no, it isn't even a permanent solution. But it works. And for a free DIY trick, it is a hell of a lot cheaper than investing in a completely new GPU.

 

Lol @RampantGorilla is literally the king of posting links to youtube videos as responses to questions. I think that guy has seen everything there is to see on youtube.

 

The lead free solder balls that connect the GPU package to the board melt at over 200 C. If you heat the chip up to 120 C, the solder that connects the GPU die to the package melts and reflows, but the solder balls connecting the package to the motherboard does not. This will fix the problem, but only temporarily before whatever caused the GPU to fail in the first place (usually overheating) happens again. In this process, the solder under the die is exposed to the oxygen in the atmosphere, leading to corrosion which may lead to chip failing again.

https://en.wikipedia.org/wiki/Flip_chip

These series of articles by Charlie Demerjian appear to explain the issue very well (links to the three previous articles are linked at the top):
https://www.theinquirer.net/inquirer/news/1050052/nvidia-chips-underfill

 

Right but it doesn't change the fact that it works, and can work for a very long time. Plus in many cases it's free to try.