GPU Died. Venting at Eurocom Support

So today my x7e2 GPU just died after a month trying to convince @Eurocom Support @EurocomTechspert to help with the high GPU temps. Games just won't run more than 5 minutes and artifacts will just appear. I've tried 5 different drivers and all produce the same results.
It doesn't help that trying to contact eurocom has been a pain in bullocks. They hardly answer their phones. I once called them 20 times in a day and all i get is voice mail.

 

@Papusan @Phoenix @Mr. Fox @bloodhawk


@GreaseMonkey90 sorry to hear that happen man. I really hope you can get a full refund or a replacement machine.

 

Sorry to hear about this. Did they not even issue you an RMA, when you informed them about the bad temperatures?

Hope so too.

 

They had me RMA it once and they said they found no issues and temps were on par with their expectations.

CPU were up to 100c and GPU were up to 90c and above. I had @Phoenix helped me undervolt it just to get it around 70-80ish Celsius.

 

@Phoenix - you got the wrong guy bruv :v

 

Fixed

 

Sounds like it might be a heatsink issue. Repasting and re-attaching the heatsink could fix that if it's a pasting or contact pressure problem.
On the other end of the spectrum, it could also be failed heatpipes. If they get pierced

GPU sounds like it's in a bad way though. With all the safety built into Pascal I'm not sure if high temps on their own could kill the GPU unless it ran like that for extended periods. I would've thought it'd throttle down super low before any damage was occurring. 90C on the core is bad, but mostly because it would suggest cooling failure and thus your VRMs are probably up at 120C which is no good for them.

Either way, they should definitely be replacing it now.

 

I had it RMA back in February. They returned it to me saying nothing was wrong and the temps were satisfactory level. I tried to convince them about the high cpu and gpu temps and they told me its just the monitoring tools that I used. Basically they are saying HWInfo, HWMonitor, MSI Afterburner is lying.

I just took out the GPU and it smell like burnt plastic/silicon. That is very bad.

Spoiler

 

Temps are a tough one because "technically" speaking anything below TCase can be considered "acceptable" (that's not to say that it's normal). However, if the GPU is at 90c it's WELL over thermal throttle (which for Nvidia is 79C or 81C) so you can tell them to shove that excuse.

From that picture, it looks to me like there's nowhere near enough mounting pressure. It would also be good to have a better lit photo as well as of the heatsink itself. The heatsink pic will tell us if the thermal pads are correctly contacting the VRM.

Even when you use a lot of paste (normal for ODMs because the heatsinks come with pre-applied dry-paste), most of it should bleed to the edge and leave a relatively thin layer upon removal. You have what appears to be the opposite problem in very little bleed but HUGE chunks on the core (and on the heatsink I assume?).

Now that you've opened it, I would re-paste it with something relatively cheap for now. Either Gelid GC Extreme, Noctua NT-H1 or ideally, Thermal Grizzly Hydronaut. This is just to see if you can get the temps down by repasting and re-attaching the heatsink properly.

I suggest these as you may have to RMA again if the GPU has failed, so you should not waste any long life (usually conductive) TIM. Furthermore, if you used ICD or liquid metal, Eurocom can use that as further ammunition to void your warranty as it's evidence you modified the machine yourself. All 3 pastes are standard grey and should be indistinguishable from the standard stuff (other than in it's application) so it shouldn't raise any suspicion.

 

I'll wait for them to give a go ahead. I suspect it might be the thermal pads itself.

 

Quick update with some picture

http://imgur.com/a/XZf3p

 

Those artifacts are usually due to memory issues. Try downclocking your ram by 200mhz or so and see if it goes away.

What are you temps now that you've re-pasted?

Furthermore, a basic test of if your heatpipes are functioning is if the air being blown out is hot or cold. If it's hot then all is working normally, if it's cold then heat is not being transferred properly.

 

Still 90C at full load.

Never OC'd the ram before. Not sure if there's the option to do it.

 

I should probably clarify, but what is "full load"? Full load is very different for something like Furmark, compared to say 3dMark

 

Full load as in when I'm playing games ie, BF1, Fallout 4, WoT, PUBG, Division. These games make the GPU run at 99%. For CPU even though it is not 99% it still reaches 100C if I don't use throttlestop to undervolt it.

Besides I can't even run the games more than 5 mins before these OX artefacts starts popping out freezing my entire system.

 

Did you test the air blowing out the vents out the rear? Is it cold or hot air? Basically stick your hands behind the vents while testing and it should feel noticeably hot.

There's no way you're actually generating that much heat. Either the heatsinks are not attaching properly (could be warping and may need a shim, but that's unlikely it'd be off on both the CPU and GPU) or the heatpipes have failed or leaked resulting in a failure to transfer the heat away.

If the air is cold, get Eurocom to replace the heatsink assembly with a new one.

 

could also check if the heatpipe went puffy or have a small hole

 

I'd check whether fans are blowing hard enough, heatsink fit is ok they're more likely than a punctured heatpipe.

Run an encoding benchmark (Realbench, something x264) send them proof it thermal throttles at stock speed and don't let them tell you that's "acceptable"

Does jacking up the back an inch help?

 

Heatsinks looks fine with no cracks. I might ask them to change a lot of things if they refuse a refund. Starting with the GPU, CPU, Heatsink and mobo

Just checked. Looks fine.

I can't run anything. I think the dying GPU is affecting everything in the system. I get BSOD if I run any benchmarking stuff. I also forced the fans to run overclock mode in the CCC.

 

Well that's a home run then.

More fool them for not listening to the customer, now they have to fix it anyway and cop an expensive replacement part on top

 

Something weird like that happened to one of my 1080 GPUs in the DM3 and the awesome folks at HIDevolution ( @Ted@HIDevolution @Donald@HIDevolution @thattechgirl_viv and @Zoltan@HIDevolution) got me sorted and I'm back in business with my SLI monsterbook. When you clean off the old paste on the die look very carefully to see if it is visibly fractured anywhere. Mine had a shiny melted area on the black plastic shroud like yours as well. I trust that @Eurocom Support can get you fixed up with a replacement part or rapid repair option since you have had nothing but troubles with that GPU all along.

It probably needed to be replaced before, so it's not going to really cost them any more now than if it had been replaced before. It's just inconvenient and bad for the customer experience to have the machine down for repairs twice.

 

I'd be interested to know if it was the heatsink or if maybe the card itself had a faulty part pumping too much voltage in.

 

So Eurocom just emailed me and asked:

 

The human testicle can survive up to TJmax of 41c, but do you really want it to be that hot?

 

Update?

Sent from my OnePlus 1 using a coconut

 

Or maybe Mod Bios

Gesendet von meinem SM-G935F mit Tapatalk

 

Modded bios can't do this.
Heck there is no vBIOS for mxm GPUs that has voltages unlocked.

 

I think everything is possible. Today we have this issue, we put hdd into slot, machine no boot, when hdd is removed, than boot. With another the same model work without any problems. Anyway, check sli cable, which one is installed? P/N number? And possible is just heatsink faulty, they must change for new one.

Gesendet von meinem SM-G935F mit Tapatalk

 

That would be a comparability thing. If something else dependant on the system bios failed, then maybe. That also is super remote.
But a GPU? Nope.Simoly because it's not possible to mess with the GPU voltage without a hardware mod or a modded vBIOS, which doesn't exist. Yet.

 

And... customer have a mod bios. We flash with stock one and everything works. The same situation we have with p870dm3g with gpu's - terrible performance and temperatures. We change second card for new and we change for another one - sli cable and we flash bios for stock from Mod. Problem is resolved sometimes depend from many things.

Gesendet von meinem SM-G935F mit Tapatalk

 

If the customer doesn't know what he is doing with a tool, Then the problem lies with the customer not with the tools.
Then again that would be a CPU heating up problem and not GPU voltage, which is what we are talking about.
Most people try to overclock without knowing heads or tails of what they are getting into.

Obviously the customer also doesn't know that he just stepped into throttle hell with the stock BIOS.

 

Have eurocom even posted in this thread?

 

Nope

Don't think they ever do.

 

Would you happen to have support case #? All of our technicians are available via phone too. Pls email to [email protected]

 

the request number is 42625. It is on route to your facility. I find phoning is very difficult as it keeps ringing and no one answers. Email was good but not great in terms of the response I get.

 

Just did at the bottom. lol

 

Part of why I say the modded BIOS is not for everyone

 

I personally haven't seen anyone from frying a system with the Prema BIOS. Now if one makes a half ass mod and messes things up, that is a different story.
Another problem is that a lot of technicians either don't have the know how of what's going on with a system or don't have the time to actually understand what's up.

 

Like how the DM3 and KM1 has a crippled stock bios which has 480w power limit?

Just saying

@D2 Ultima

 

Yes, unlimited power budgets are probably left for more advanced users too.

You would be surprised being on the line where people actually come to you for help and are not admitting things publicly on a forum.....

 

Oh nono I totally get it. No one will own up to their mistakes. Specially when it comes to expensive systems..
Last night I borked my 1080Ti , just flash the bios inline. At that point I was like well no other way than to de solder. Did that and recovered , added a few resistors around the power ic , no power limit. Time for Trimpots now.

But my main reservation is with calling out modded bios as a "vague" culprit for most faults. Can't it mess up a system? Sure it can. Can the one we are using? Very very very unlikely.

 

Unlimited? Certainly.

Unable to achieve stock performance? That's a rather different story. DM3/KM1 1080N SLI cannot achieve full stock performance.

If the power limit was something like 600W (well under the rated 660W from the dual-brick configuration) where 190W/95W/system drain was properly achieve-able (hell even 550W) then it'd be fine. I mean the Clevo app even allows overclocking by default... what's the point if the power limits prevent stock? Overclocking isn't a realistic concern if stock doesn't happen.

I've never wanted a "Prema BIOS" as Clevo stock. What I *have* wanted is:
- Problem-free stock BIOS
- Proper functions for built in tools, like how undervolting initially did not work at all in the BIOS, or XMP
- Enough power to run the system at stock, maybe with a little bit of headroom to allow for spikes.

That's all. Prema mod is indeed for enthusiasts and people who want system control, but it's what I consider necessary in this case since the machine does not actually clear stock.

 

I would never buy a Clevo (or any other "high performance" notebook) if living with the stock BIOS was the only option available to me. I have never owned a laptop that was capable of meeting my expectations with a stock BIOS. @Prema is the only reason I own one now. The P870DM3 is worthless to me with a stock BIOS. So were the P570WM, P750ZM and P870DM-G. Likewise, the Tornado F5 is also worthless to me without a @Prema or Svet BIOS mod. Before these, the Alienware machines were worthless to me with a stock BIOS. I have never seen an Alienware damaged by an svl7 BIOS or a Clevo damaged by a Prema BIOS.

Guns don't kill people. People kill people.

 

Well yes. You're an overclocking enthusiast. My P370SM3 from Mythlogic has a Prema BIOS, but it is cut down. I cannot say... adjust bclk. I cannot do everything that a full unlock would do. But it had enough useful options and was stable enough that I kept it.

That's all that's really needed for most, and that's my point. If a real enthusiast needs stuff they'll want the mod regardless of stock.

Of course I can't live without one now. That BIOS is a dream. But I'm an enthusiast. Not an overclocking one really, but a tuner and broad spectrum one.

Sent from my OnePlus 1 using a coconut

 

So it got sent back with new GPU and unlocked BIOS. I saw the notes on their status update and 2 of them were these:

* Ran specific game/BattleField-1 @ max res (CPU 99-100C)
- Recommend limited long time use of high end/resource demanding games/restricting resolution (i.e. BF1 etc.)

I just don't get it. I bought this laptop to play games and I am not suppose to play use it for a long time.....

 

Translating that into plain English, Clevo's poorly fitting unified heat sink abortion is incapable of providing proper cooling to a system with powerful components like yours. They flubbed their dub and that's how our cookie crumbles. Better luck next time in the Clevo heat sink lottery. Note to self: avoid anything with a unified heat sink.

 

What our technician meant was to watch temps when you play for extended period of time as some of NV drivers may be unstable and may cause temps rise. Nvidia is updating their drivers all the time and very often they are unstable.

 

I think some thermal pad tweaking might be in order on that one, I would look at the thermal pads, is one side showing greater pressure marks than the other?

 

I was thinking the same thing as I reflected on this earlier today. This might be a good one to test K5 Pro instead of thermal pads. It's kind of messy, but it would be worth a shot. That way there would be nothing to interfere with core to heat sink contact. I also still wonder if it has the original heat sink if there is something wrong with it. Either deformed or bad heat pipe(s). As much as I do not like the unified heat sink concept, there has to be a simple reason this one is abnormally hot compared to others of the same model, so it has to be something like thermal pads too thick or defective heat sink.

 

Thanks for the info. I might get it and try it out. Oh, and how messy will it be? 1-10.