Clevo p775dm3-g - graphics card dead, what now?

Dear All,

recently I god my hands on a Clevo p775dm3-g (XMG U717) in a seemingly good deal. It should have replaced my Lenovo y510p.

The specs are:

CPU, i7 7700
GPU, mxm gtx 1080 (dead)
RAM, 32 GB
Storage, 256 GB SSD and 1TB HDD

Unfortunately the graphics card died on me just after a few days and reballing had been tried without success.

I'm am a little bit lost and don't know what to do with the machine now. I hope you can give me some advice on what kind of options I might have in this situation.

Maybe something that has to be considered, as a student I don’t have too much resources left to invest at the moment.

Thank you for your advice!

 

Reballing 10 series and newer cards is very difficult. I would think most BGA repair people can't pull it off.

If you want to just get is running, you can pick up a 1060 on ebay for low $200s. 1070's and 1080s are much more expensive.

The system can run 20 series cards, but that requires ordering an updated heatsink.

 

To add some info - There are 1060s pretty cheap, 1070s and 1080s : Folks on the forums here might have one cheaper / quicker than Ebay.

To go to a 2070 / 2080, you'd need to source the card first (much harder than 10xx series) and pick up a matching heatsink - Best source for those is Aliexpress, pricewise.

My recommendation would be to wait for somebody (here or otherwise) to part out a dead unit.

How certain are you the GPU is dead?

 

If you want it running it needs a GPU.

 

Thank you for your answers!

Also, i saw a youtube video, that explained, that reballing does not make sense usually. What is your opinion on this?



So I guess a 1060 might be the best bang for the buck? The thing is, I have a little bit of a trust issue to order a used card on ebay after the disaster with the 1080.
So I am not sure, if i should risk it. I would fear that I get a reballed card, that will die on me again after some heavy load.

Iam pretty sure, but I am not an expert, so here are some pics that show the graphical errors during reboot:
I might see the parts of bootmenu when booting from stick, but most of the time the graphical errors are severe and i am not be able to notice, if the system boots into the OS.

Would it, maybe, also be a reasonable choice for me, to part out my unit and invest into another machine? What price would be reasonable for my p775dm3 without graphics card?

 

He is a quackjob conspiracy theorist. Saying the EC is running a timer that makes it start acting like it needs a reball just after the warranty expires is... nuts.

The truth is around 50% of the time you can get a complete repair by just applying flux and reflowing the problem part. The main issue with this is if it starts working again, you don't know if you got a good connection or not for the previously broken contact, and it may die again in a week. When inexperienced people try this they almost never go hot enough (> 200C numbers are too scary) or don't apply flux, so for them maybe they straighten the board a bit to temporarily bring something broken back into contact. Must exceed 217C. Usually need flux.

For another 25% a full reball will fix it. If it works after a reball you probably soldered every connection well. With < 0.5mm solder ball size processors the past 5 years though are far more difficult to solder than older 0.5mm ball and larger.

For the last 25% something is physically damaged on the processor, such as solder connections from the core's pcb to the core itself is damaged, and you're never gonna fix it.


Your image links are broken.

 

I brought the Clevo to a specialist the other day and he heated the graphics chip with some optical heat gun for some minutes up to 450°C, iirc. To be honest, it seemed a little high for my taste.
Testing afterwards showed, that the problem got actually worse. Even more graphical errors appeared and in addition to beeping noise from the Clevo. He offered me to do the full reballing process for 180€. Fortunatelly, I only have to pay, if it works. I guess its worth a try, although 180€ seems like alot for the job.

PS: I fixed the images.

 

The air temperature was probably 450C, which is common. The part won't get nearly that hot.

Those artifacts are 100% a broken connection issue. It should be fixable.

The beeping is the the motherboard can't read the GPU core temperature. Residue flux and alcohol to clean it commonly cause this for a few hours up to a day until it dries out. I would think you still have a garbled image on the screen. If you have no image at all that is very concerning.

If the job goes smoothly it takes around an hour. If not, which is common with newer parts, then it can take several.

My main concern is with it getting worse that the system wasn't run enough to bake out moisture and that the GPU core's pcb literally exploded. If this happened you will see lumps in it.

 

@jimmi101 That video is pointing out 'reballing' as a problem to material design problems with particular Nvidia / ATI chips on Apple (multiple generations, but here I believe he's talking about older Nvidia GPUs), a bit different than a simple production error one-off problem like on your GPU (as it seems).

@Khenglish - Conspiracy theory, sure. He's often obnoxious, but I like much of Rossman's content, and he does run a repair business well enough!

 

Wow. Almost like clockwork, two hours after yours, mine threw itself off a cliff yesterday with the same symptoms.

I will be parting mine out I think. Just doing one last clean and reassemble to make sure it's not something dumb like dust.

 

I am sorry to hear that Harvey! I know how this must feel.
Did you have some symptops that showed, what was coming? I had some random black screens the days before, untill a sudden shutdown finished the graphics card off for good. Maybe reflux or reballing might fix your card and is worth a try?
At the moment I tend to letting the specialist in my town give reballing a try. If it does not work I will try to land a good deal on a 1060 or 1070. 1080 seems too expensive.

As far as i understand it now, the reason for connection issues on flip chip GPUs might either be a broken BGA (2nd lvl interconnection in picture below) connection, or a broken microBGA connection (1st lvl interconnection in picture below, (the motherboard would be the mxm pcb in case of a Clevo-Laptop)).


Reballing refers to the resodering of just the 2nd lvl as far as I know. So if the damaged connection is there, it might be fixed. But if the 1st lvl is broken, reballing will not directly help, but the applied heat during the process might actualling fix the connections on the first level too, if one gets lucky. But the fix might not last long in this case, according to what I read. Also some peaple claim, like Rossman, that the problem is usually in the first lvl, hence his objection to reballing, I guess.

 

Oh the need for reball is absolutely mostly initial design error. AMD had an idiotic trend of making the memory solder pads smaller than the other pads, but then using the same size solder ball everywhere. This always resulted in weaker RAM connections. This used to be ok because the corners, the areas of the most physical stress would have full size sacrificial GND pads to hold the solder rigid, but they stopped doing that with HD 6000 and newer cards and started getting major reliability issues.

As for why mobile GPUs have a high failure rate, I'd say it's because particularly lately Clevo for example only has 4 screws on the GPU heatsink just for the GPU core. This means any physical stress on the heatsink channels straight to the GPU core solder. This is particularly problematic on the large, heavy unified heatsink designs. Not only is the heatsink heavy, but any misalignment results in greatly increased strain on the GPU, and getting a CPU and GPU of different heights to screw to a unified heatsink always results in slightly off angles, slight misalignments, and major solder strain. This flexes pcbs over time for ever more strain, and eventually things break. For high power systems don't use unified heatsinks, and you need more screws other than screws for the core to get the weight of the heatsink off of it.

While some designs are more prone to failure, that doesn't say that a reball doesn't fix the issue. If you break solder you have to resolder it. There is no conspiracy theory internal clock to fake that the device is dying.

 

The first level does often have issues. You can't get flux on it, so even if you do get a broken connection there to resolder, it's probably a weak resolder.

In general 1080 cards are not prone to failure, but Clevos are due to the issue I mentioned above. In Clevo's case you more likely have broken 2nd level, which is very fixable. If you check your card for bending from the side, you can probably see where the area of the core is pulled upward, which is from the heatsink pulling down on it when it is in its normal upside-down position when running. For the 20 series they switched to separate CPU and GPU heatsinks to prevent this from happening as severely.

 

I see, that does make sense! I am getting my hopes up a little, that a reball might actually do the trick and my Clevo runs happily ever after!
My guy can do the job on wednesday, I will keep you updated on how it turns out!

 

It was switching itself off while I was playing 1 game specifically: X4 Foundations... yes, I know, I'm a massive nerd that likes space. These weren't your normal crashes, it was even causing the power-brick to switch off, necessitating plugging it back on. Then, while in X4 again, the whole system hung, sound went mental, screen showed white frozen static. Could only switch it off by long-pressing the power button. Once I turned it back on again, it started showing the same static as yours.

I'm done with it. While it's served me well around the world, I've never been 100% enamoured with it. If anyone on here wants my P775DM3-G for parts, spares, etc, PM me.

I bought a Gigabyte AORUS 15P-XC today and so far it's a big step up.

 

The move to lead free solder and not getting it quite right scuppered a wave of Nvidia chips too. Again on that first level.

 

I see, I would be happy, if you could let us know, how the parting out goes.

Is it maybe also possible to resoder the first level?

 

Oh the initial move to lead free caused so many issues. On early 2000s ram you can pry off the memory chips from the modules without desoldering the chips and the pads would be undamaged. When delidding a P4 the TIM was caked on so hard that it was stronger than the solder, and the core ripped off.

 

Yeah, the package is not supposed to be a socket still caused issues with the apple Nvidia chips especially thanks to thermals.

 

Not without fab level equipment.

 

Quick update:

The Reballing attempt of my 1080 failed unfortunatelly.

So I guess I have to decide between:

1) A new mxm gtx 1060 for around ~300$
2) A used 1070/1080 for 400$ to 500$

I am also asking myself if cards from the 9th generation are also possible in the p775dm3?
Also is a 1080 generally a good idea in this laptop, or is i prone to fail at some point due to high temperatures?

 

Due to shape, no.