My M17x R4 blew up - literally - or the 980M did

Hello all,
So yesterday I was playing Mass Effect Andromeda, running svl7 johnksss vbios @ 1262/1399 1.037v. Temps were fine, never exceeded 67c with a recent paste job and Fujipoly pads.
When all of a sudden I started hearing this crackling sound followed by the computer shutting down and the light in the PSU going out. Then a loud BANG and smoke coming of the computer. Disconnected the computer , did a power drain and opened the under cover to see WTH is going on. The CPU and the surrounding area looked fine, checked the die side of the gpu and nothing seemed off.
So when I removed the Graphics card I found this:


At first I thought the chip on the MB blew up/shorted. But I checked the under side of the GPU and here's what I found:

It looks like a cap or something blew on the card and the black smoke stained the MB.
With the card removed I powered up the computer and was greeted by the 6-beep video card failure. which led me to believe the system is still working and all I need to do is replace the GPU.

@woodzstack mate I'll PM you regarding my next GPU

 

Hi, I'm sorry to hear that. I have a 980m newly laid and I'm happy with it. I have a 680m for sale in case you might be interested.


Enviado desde mi iPhone utilizando Tapatalk

 

Holy friggin hell !

Never seen anything like it , thats crazy !

is this one of my older GPU's ? I might be able to do a warranty exchange for you. Would love to look at this a bit better. It appears as if the component caught fire and fried the motherboard components as well.

That's just disasterous. You know for sure, this is not a common place thing with these cards... but giv eme a shout , yeah, i will see what I can offer/do to help.

Is that a AW17 ?

I have a friend here selling a 120Hz screen and cable, you might be interested in, and going with a 1060N or if your crazy into modding a 1070.

just mentioning him, cause the price he is asking for, for the cable and screen is like giving it away, think he took it to the NBR market place.

(of course, if this is the laptop you have in the signature, then you have a 120Hz already, would be more economical for me to help you get a 1060 going.)

 

if you can hook up an HDMI or DP display you might be able to still picture without a videocard in there if you have the 120hz screen. I know for sure for the DP display using the miniDP port, I forget now if the HDMI will work.. wait no, cause its part of the HD intel..so just the MINIDP... this will tell you if your laptop is good or not... I can't see the extent of damage on the mothertboard but it might actually be fine. here's top hoping !

 

Holy crap....that sucks to say the least, sorry man

Hopefully Woodz can help you out he is awesome. Hope you get it sorted out soon. Let us know if the motherboard was salvagable.

 

Thank you, but I'm more interested in upgrading rather than downgrading. GLWS

 

Cheers mate
Yup, that's my second purchase from woodz.
I bought a 1060 , I'll keep this thread updated.

 

Just going to chime in, this is not a clevo specific failure. I had a Dell 980M fail in the same spot in the same way, was running a higher voltage vBIOS at the time. Never again.

 

But his GPU failed at 1.037 V which is stock for the 980m and that is freaking me out.

 

I ran Prema vBIOS for months before I switched to the Svl7, Maybe the MOSFET issue which made it blew up, 980M has missing VRMs on the PCB, Hardmodded ones have much higher OC potential and stability over the non modded ones, 980Ms from what I read from a post MSI ones seem to be better...rare tho.

Glad that OP sorted it out, Mobo is good things look bright !!

 

Oh yeah ?

Wish I knew this sooner, i would be trying things in the form of prevention. I want to know exactly where on the chip the component is, maybe its common enough that a small prevention can stop it from happening.

as a seller, you know, it's a nightmare to have this happen to you, one of your clients having some crazy thing explode or catch fire, that's bad press or whatnot, so I want to make sure I can brag it never happened to me !

(which sounds more like prevention then luck IMO)

you know, about vbioses, no one every compliles which vbios did what... we should get that info, if either of you know which ones failed you in the past, I would make a record of it, and maybe gradually figure soemthing out to help people in the future. Imagine we find out its vbioses that were stock for a specific laptop that always failed... like if it was a 84.04.22.00.0A from MSI or like a 84.04.22.00.10/11, or even prema's EC versions or such...

whatever it was, that info could also prove to be valuable. I'm saying it's a missing variable.

 

I am actually running the card undervolted on stock clocks just to be on the safe side, and IMHO 980m is still, even at stock clocks, an amazing GPU.

 

did your vram vrm blow up?

 

It was in the same general region, not sure if that exact chip or not.

 

I knew the 980m blowing up was a matter of when not if. Overclocking is something I do for fun not just to gain a few fps. And the card blowing at stock voltage shows how shi*y nvidia''s/OEMs design & QC is - 3/6 MOSFETS-
I had a 680M that had all components on board and overclocked like a beast.
The last few months my card wouldn't boot up with Prema''s vbio's (custom & v2)
At the same time the memory clocks were stable up to 3005 mhz, then it went down to 2600 mhz with anything higher and it causes weird artifacts and lockup.
The cards sold by Woodz are probably the best cards in terms of quality and ASIC scores. The card is made in Canada after all but you can't do much if the design from the beginning is bad.

 

Cost savings, As we know, One electronic component in a large yield lot indeed adds more to the BOD's pockets, Look at the BGA HQ CPUs total crapshoot (Pair that with Tripod mess and boom !), High Voltages and worst bins, trash yield from desktops come to these soldered garbage machines, Same applies for the GTX 980M and NB chipsets, iirc the Desktop ones which are not good are moved to the laptops, Ofc the handpicked chipsets are the best that money can buy

No one makes noise in the Notebook market on gimped quality or standards, Best Example is the MXM standards /BGA fiasco and the new QHD AUO (TN) 120Hz panel which has lines due to NVSR + Hw-GSync & the b2w or b2g (don't remember exactly) of the AUO FHD AHVA (TN) 120Hz is sick, the response time is so low causing ghosting or motion blur at high refresh rates + high FPS, even the 60Hz 4K IPS GSync is not enough for 60FPS smooth response times also the 6Bit + dithering to 8Bit - quality mess, Very very awful state of mobile market, We need people numbers to win this atleast for us who are steered towards quality and perf and respect user choice over bending over to the corporates...

Sorry for the rant but yeah sometimes these corporates trigger me

Anyways hope the 1060 makes you happier and runs cooler and faster, Only one nitpick is 6GB of vRAM

 

Nicely said friend.
If the 1070 didn't need any modding I might've bought it. I thought about the 6gb vram and the 192bit bandwidth but it won't make a noticeable difference. Only ROTR pushed my vram to the limit ( one game of hundreds )
The 1060 is 20-30% over the 980m and runs cooler and consumes less than the 970m. I hope all of this will allow me to OC the hell out of it.

 

Not to thread jack, but let me know if you need a 980m. I have one for sale in the forums here and it's a 4 month old woodzstack clevo model. Only powered up twice and never overclocked.
To actually contribute to the conversation... I'm surprised to see this extreme of a failure. I'm a firm believer that cooling on the mosfet section of the card is the root cause of the failures in our older laptops. If you look at my 1070 thread you can see I added a heatpipe to cool the vrm section of my 1070 now. I had a 980m and a 1070 fail on me and with this upgrade I have gained far more stability than any other mod I have completed. If you look at all the heatsinks for laptops now that have the 10xx series they have much much more focus on cooling this section of the card.

Sent from my SM-G930W8 using Tapatalk

 

Going back to the OP - did your big wad of thermal pad there in the first pic not cover that chip?

 

It covered half the chip on the MB and the back of the graphics card completely. From what I can see it's there for the upper vram and an insulator between the PCH and the GPU.
The burning/spark wasn't only pointed ↓ I found staining close to the CPU and it burned through the edge of the thermal pad.

 

I think we're on to something.

Look here :



And here :



and here:




This card still works. I was inspecting it, and it looks like it might face the same disaster in the future.
So I've decided to have it sent in for warranty before it blows up and causes a mess.
If anyone is unsure, take a look at your card and see 2-3 months in the condition it is in, early warning signs could save disaster!

The card is also slightly bent at the tip now, and surely doesn't make proper contact with the heatsinks because it's not levelled , in that corner anymore.

If I had to guess, it never made proper contact, maybe close but not close enough and after a time, it started to drift/melt/bend/slope and eventually would have caught fire and exploded/died etc..

SO - I think we should come up with some sort of technique to prolong the life of components in higher end cards like this. Maybe a larger saturation of a cheaper by volume silicon paste that can wash off easily, and use it to create a foam or fill in area's that do not make perfect contact with the cooling pads and heatsinks.

Or maybe some smaller/tiny copper shims and pads . Idea's and suggestions welcome. I think we could easily test these out by throwing a sensor under neath the component and seeing the results in various situations and such.

Got to improve technique's somehow, otherwise this is the stuff that helps convince people BGA is better... when it's not.

 

what the heck did you do to those resistaors ? They're all in parallel or something now ? Soldered together... might as well short them out.

 

I had a Clevo card that matched exactly in that first pic of yours, discoloration/burn in that area. The solution would be to use better quality MOSFETs as well as a good quality thermal pad that can make contact with the heatsink.

For those with cheap mosfet, better to not run a higher voltage vBIOS and would probably be better to stick to stock or a UV BIOS if you can find one. Rudementary EE, probably more complicated than this but Power = Voltage * current so lets not up the voltage for components that might not be able to handle them.

 

You have one broken like this ? One of mine ? Send it to me, lets get it fixed, and try and better our service with the new findings.

You know, every company's cards always have a shelf life, from what I am seeing these Maxwel cards have 2 -3 years of hard work on average. Which is still the highest to date from any of the cards over the years, but I knew the time would come when eventually people would start having dead cards. I would rather get them repaired or replaced then have them saturate up the black market with scammers trying to pawn off dead cards and stuff in the end (Not that I am suggesting anyone here is a scammer, but dead cards lying around cause grief..)

 

Not your card, a generic Clevo 980M (no stickers like yours has). It was failed in the corner by one of the MOSFETs and on the back near the memory VRM.

Your card has no warning signs, it's just behaving funny. No scorch marks, no burn smell, nothing really funky. Just blacks out under load. I looked at Ashtrix's thread and I thought maybe about undervolting and checking pad placement so I will do that first before we proceed. Usually the failure points on dead 980Ms are fairly obvious, and I haven't seen one with a dead core (yet). Core seems very robust. ALL 5 of the dead 980Ms I had have been memory VRM failures or MOSFET failure.

 

Yeah, I warn people now to make sure these components are coolers to perfection in the corners to last the longest. To be fair, those MOSFETs are not really taken care of very well by any of the heatsinks.

 

On the 17R1 I noticed that the pads arrangement in the area that is supposed to touch the MOSFETs would never in a million years actually touch them. So, I used hard 2 mm pads followed by very soft 1.5 mm pads to safely reach them from the right location to be sure they have proper contact.

 

given the distance , or size of the gap I should say rather between the heatsink and the component, you would need to cover it in a silicon paste to increase surface area and then make a greater contact with the heatsink, and possibly cool it from underneath the card too to be triple sure. That is my best guess. Once you hit 2.5mm thickness most thermal pads are nothing but an insulator

heat, like electricity travels path of least resistance, thus it would saturate the area first and then slowly dissipate to the pads, so might as well add mass to the area and then cool it from both sides.

I think I will see if we can modify the TEP modules we have to do this, and I think I would start making suggestions to those with Clevo backplates to add some thermal pads to the back of the corners.

If someone like myself sells these cards and doesn't learn from mistakes and faults on these cards and how to use them, then whats the point in even being your main seller/resource for this, you know what I mean ?

 

The problem with the R3/R4 (don't know about the AW17 R1) is that the PCH is right under the MXM area and it reaches crazy temps. So even with the big thick pad between the card and the MB it still broke. I had Fujipoly 11mk pads with a straight heatsink and no gaps at all & this happened.
To change the subject, I received the 1060 and need to mod the heatsink, will report soon.

 

Nice !

Then I suggest something to Isolate the heat from the PCH

 

Same for 17R1 ... DELL uses a super thick sticky strange kind of thermal pad I would say around 10-15 mm to keep the card well away from the motherboard ... Maybe PCH killing cards in 17XR4 is more common than what is reported on the forums and they fixed that in the 17R1

 

wow thats some bad luck there

 

I never had the 100W GPU here I used to have an 860M so so pad for me, All I could do was use the same pads on the bottom as well. But yeah the PCH is right underneath the MXM slot they both just offset each of their heat and cause massive temp increase often the PCH scales to 80+ not a biggie but it's not a good thing from longevity pov. I don't know how to avoid this, the copper cooling pad that Eurocom lists on their 980M AW17 upgrade video might help due to the total area of distribution, Someone has to test it & that thing costs a lot, It's Eurocom, no surprise there.

And also to update, I was running ME:A and the Comet render in the "Faroang" system made the GPU switch to a turbo W16 power guzzler mode, freaking 146W power pull !! First time I saw over 140W+ with non overvolted vBIOS and the GPU was at 100%, being at full usage isn't a surprise, as the MEA often makes the GPU scream at it's peak, but the temperature & power just spiked to hell in that timeframe, and when you move away it drops.

This is like Furmark lol, Which Nvidia doesn't have control unlike the Microcode detection when Furmark runs..Frigging 10C+ drop man when you move away from this scene wtf ?!, Imagine if you run this Overvolted, Perhaps the densest smoke (remember the smoke in the cryo deck ? It's ramps up the GPU core to max also temps, FB3 at work I guess..) is raping the GPU totally, look at the energy it's giving off...Anyways Thanks to Nvidia for the superb efficiency we are seeing here lol, Wonder how the 1060 fares @Solo wing, Waiting to see it's results vs 980M on Wattage and Clock stability.

I was happy that I was seeing my GPU sit at 84C and not crashing, this was for a good amount of time, So was thinking to post that on 980M thread but just to confirm I fired up the game again and directly forwarded to that render scene and let the machine stay there for a while and soon temps ramped upto 84C, I think this time I made it to run at 84C for atleast 6-10 mins and then bam it happened again, The infamous blackscreen...due to being hit 85C mark, As usual. One thing to note is, the Phobya pads worked good since last time I could barely sit at those temps.

So, I'm guessing (need to have more samples and use cases with maybe further pads & examples to finalize) the FETs are awful, the power load is not the problem here the current increase might be the cause to make these cards trip.

P.S - MEA GFX settings are FXAA, Ambient Occlusion, Motion Blur, Film Grain - Off & Mesh Medium, Lighting low. Rest Ultra.

 

After the reports of dead GTX980m's I wouldn't let mine touch anything above 100w power consumption. I am sporting the factory vBios and it wouldn't let it go above 100w and I am happy about that