XPS 9550 - GeForce 960M - BSODs and unstable system

Hi, I have issues like BSODs, system freezes, super snail performance and nvidia graphic's driver crashes. It's even hard to get to login screen unless I disable nvidia chip in Device Manager. Then everything is working normally. When I try to run anything a game or other 3D intensive application using nvidia chip I get system freeze and then BSOD. What's weird to me is that I don't see any yellow triangles or red circles in Device Manager. It takes long to install drivers for nvidia chip but they also do install.
I've replaced RAM chips, tried different SSD, reinstalled Windows, reset BIOS, flashed older BIOS... nothing works except disabling nvidia.

My problems started after my xps suddenly shutdown itself running Valley Benchmark and Prime95 in the same time. Led in front was blinking white and orange for some time and then laptop shutdown completely.

Could it be that I had some worse quality nvidia chip and it didn't handle the temps too well? As I said already system is very stable now. It wasn't that stable even when nvidia chip worked. I'm waiting now till my warranty transfer will be finished and I will contact Dell as I don't have any other ideas. Maybe some of you have suggestions I could try in the meantime?

 

I think those LED blinks can be used to help troubleshoot.

I think you can get into a troubleshooting mode by staring with "fn+power". That might cause the LED on the caps lock or front panel to blink. Record the sequence and search the internet for potential meaning. I didn't find diagnostics for the 9550 but the process helped me diagnosis a loose LCD cable (caps lock light blinks 3+3+2+1).

Apparently there are a few startup keystrokes that will brick your laptop so do some research beforehand.

 

Sounds like a bad vid card. I'd get an RMA started.

 

Blinking LED happened only once when this situation stared so I don't remember the exact order of blinks orange/white. Troubleshooting mode doesn't give me any answer either as it shows that everything works properly.

 

That's what I think too. Can you even start RMA when you have warranty transfer in progress? I'm in Poland and I would have to send my XPS to US on my own cost. Dell US only told me to contact polish branch but they don't want to talk with me unless my warranty transfer is finished (polish agent said something about that it's different configuration than they have in Europe and it all needs to be in papers, because they will probably have to send it to US if some parts are missing in their department, etc.). I foresee nothing but trouble with this laptop. I should have stayed with ASUS (I had Zenbook PRO UX501VW prior to XPS, but my version was with glossy screen and I didn't want that so I've returned it). That one gave me no trouble at all.

 

Try repasting the CPU/GPU and checking all 4 thermal pads make contact.
The blinking will be a hardware shutdown due to overtemp.

 

I forgot to mention that I did repasting and changed thermal pads for 1,5mm. It was before I was running the Valley Benchmark and Prime95. I also undervolted CPU, CACHE both by -150mV and Intel GPU -105mV.
Just now I've upgraded Windows 10 to Anniversary Edition and also had problem on initial booting with nvidia chip enabled, but when I have disabled it everything is back to normal.

That's what I thought too.

 

Considering I had trouble seating the heatsink on the GPU with 1mm pads that were quite 12W/mK rated you may have bugger all contact on the GPU now you are using 1.5mm pads?

Take the heatsink off and take a picture of the GPU and heatsink as is so we can see how the thermal paste has spread.

 

So what do you suggest? Buy the new pads 0.5mm? Instruction on my pads says that they can be flatten up to 0.5mm so after pressing with screws the pad would be 1mm thick right? Besides even if chip would get to much heat safety features would simply shut down system and nothing else should happen, especially that my testing didn't take longer than 15 minutes (most of this time only for prime95, Valley Benchmark was working for like 3-5 minutes tops).




I know that the nvidia chip doesn't look super clean, but this paste isn't conductive (well it is but not for electric current ) and I couldn't reach to all this tiny spots with my cotton swabs.

 

Hi dontcha -

I am no expert but did a lot of research on thermals and asked a lot of questions here. My preliminary thoughts you might investigate:

1. It looks like the thermal paste did not cover the entire chip. The paste tends to soften up with heat so maybe it would have migrated over the entire chip over a several days of heat cycling and performed better after some time. Maybe not. Regardless, a large percentage of your GPU was not connected to the heatsink.

2. It looks like the thermal pads do not cover the VRAM chips entirely (although I don't think that is so important)

3. Were your VRAM pads really fluffy and soft? I used 1.5mm clay-like pads for the two defective VRAM contacts and 1.0mm for the other two VRAM contacts with no issues; those apparently performed the same as 1.0mm and 0.5mm, respectively, on my system but every defect is different.

I linked a few primers here that are worth a quick check as a start. Then spend some more time researching for more definitive answers:

http://forum.notebookreview.com/thr...ndervolt-repaste.785963/page-34#post-10269283

 

I suggest getting a 1mm and maybe some 0.5mm pads as the heatsink isn't making contact. The 1mm are fine on mine but you can use the thinner .5mm to fill out any part that isn't making contact. The you should look like the cpu when removed as the cpu had made property contact.

Sent from my SM-G920F

 

1. It was my second attempt. This time I was using "grain of rice" method so that's why the chip isn't covered entirely. First time (before testing with benchmarks and undervolting) I was covering the whole chip using some old shopping card. But it looked like part of the paste was pushed in one direction (maybe I didn't screw heatsink with proper order). So after the sudden shutdown I've tried with "grain of rice" hoping that this would help somehow.

2. Well after putting the heatsink on VRAMs they are covered completely, so I don't think that it matters so much. Even the ones from factory weren't entirely covering them.

3. Yes they are very soft and I think that even softer when high temperature is applied.

 

I will try with three 0.5mm from Gelid and maybe one thicker for under the pipes. For now I still have to wait for Dell as there is no point buying stuff for damaged/not working chip. I'm thinking that I also buy some esd antistatic brushes that they use in electronic workshops to properly clean the chip. Cotton swabs leave to much fibers around the heat spreader.

 

In addition to your internet research, study this thread:

http://forum.notebookreview.com/thr...rature-observations-undervolt-repaste.785963/

Some people post that one or two of the VRAM chips generally need thicker pads than the other ones.

Cotton swabs leave a lot of lint fibers. There are lots of alternatives you can find online; I used a lot of camera lens paper, never reusing a side or pressing too hard so not to scratch surfaces.

For cleaning liquids, some people use extra pure isopro alcohol say 99.95% pure or better or use or something specialized like ArtiClean. Make sure you have excellent ventilation and note these products can react badly with plastics, rubbers, etc...

 

I have to say that the high thermal transfer pads are trouble, the originals did the job and considering ddr4 and above don't kick out much heat and probably get hotter from the heatsink under load I will be fitting the old softer ones again

Sent from my SM-G920F

 

Were you able to solve the problem? Did the RMA go through after warranty transfer ?