980M Blackscreen on 84C+

Hello people,

Background - upgraded GFX module around March 2016 with new heatsink and thermal pads, had to use the upgrade kit's low end pads on the back on one vRAM stack, another one from my old heatsink (860M from Oct 2013) onto another vRAM stack on back. Next is I ran quite good AAA titles but sometimes temps went past 80C don't remember exact nos & It was fine.

Issue
Now after a long break since April, this month I started AAA with pending Bo3 before it was Dota2 mostly, after a quite long time temps went upto 85C and first encounter with blackscreen had to power shutdown did a repaste immediately but later didn't put that much of load without AC and had to perfrom an OS wipe didn't test Bo3 but Technomancer maxedout and after quite a time faster the temps reached 80+ and upon 84C couldn't see the 85C number but I'm pretty sure that It reached that point blackscreen agai, Replicated same with Heaven maxedout too, Panicking I repasted 2 times in which the last repaste was a clumsy one which caused scratches on the GPU core / too much alcohol waste seeping into the mobo caused removal of the mobo film and dry it out (yeah, dumb asf), removed that Blacktape since some fine lint got stuck, thought that might be the cause. But now after that messy TIM job and hassle I din't bother to test again past 84C mark, Afraid I might fry the GPU. Side note - waiting for a few components to finish off my U3 mod.

A helpful person Solo wing (M17x R4 with 980M owner) said his card also did sometimes blackout with 73Cs on high load due to MOSFET overheating and no repsate helped until fujpoly pads helped to sit at 67C and no issue since then.

Is Maxwell really that bad or something wrong here ? I'll proceed to order this Fujipoly and some extras for a better TIM job, Posting this so that I may learn something that I didn't knew & might be of much help in fixing this. Any tips / thoughts are welcome & much appreciated.

Thank you very much for your time & Patience.

P.S : To mention yest I noticed the card fans were not on on normal load no game, temps went upto 85C and system didn't put that blackscreen, Immediately I turned the fans on.
Also no OC /OV, Light benching for Driver testing.

 

had the same issue with my 980M mxm, I'd suspect the VRMs overheating or not able to supply enough power to the card

you are at stock 1.075v w/prema right?


@Mr. Fox might have some more insight to the vrm cooling solution

 

1.065 is the default not .075 IIRC. Yeah that only.

 

Double

 

Hmm, my previous AW18 with single 980M had 1.075 though

How thick is the thermal pad required on the vrm? AFAIK the fuji/thermal grizlly pads don't compress very well.

 

I used the same layout as the stock one, I don't have any snaps, But added a small set of non standard pads to some blocks which were exposed & I guess the standard rectangular pads size is 1mm thick.

Edit : Search is not working at all, couldn't pull out links..

 

I'd say if you had to stack any pads, it's making bad contact.

@woodzstack can provide some insight maybe?

 

There are two types of Fujipoly pads, the difference is thermal conductivity.
- 11.0 W/mK are easily compressed, and flexible, it's also cheaper.
- 17. W/mK (the ones I got) , less flexibility harder to compress but had no problems with it at all. 1.0mm fills the contact gap nicely and cutting to size is easy.

Putting quality thermal pads might not affect core temperature directly. It will however reduce ambient heat surrounding the core and the higher thermal conductivity the faster heat is transformed from the card to the heatsink.

 

I don't have the stock vBIOS, I'd have tested with that first. Lost it with Win7 format back then b4 Maxwell. @woodzstack could you please help me with the vBIOS, If it's the voltage then I might request Prema for an Uv BIOS.

 

What exactly do you need ?

 

I backed up the vBIOS that you gave it to me with the card but I lost it, sry.
Could you please pm me the vBIOS for the AW17 980Ms ? I would like to test again with that vBIOS. And If you have any idea about this too, It'd be great.

 

almost every install is a custom job, even a "plug and play" as many of yous started calling them.

Take some strips of paper and judge if things make contact by sticking it between th heatsink and the card to see if the pads make contact in the area's you can not see well enough (like pull/tug the strips to see tension)

Always do a complete visual inspection on various planes. See that the heatsink does not bend the card and or the heatsink is not bent itself.

Remember that thermal paste is there to fill in microscopic air gaps. Air has higher thermal resistance. But no matter what thermal paste you use, copper is twenty times better. So - less is more , here. Just need to fill in the spot that are nearly invisible to the eye. Most pastes should be almost see through if you can spread them thin enough.

The only expert thing is - to make sure the heatsink make perfect contact when it is on the GPU, you would have to thoroughly test this before playing around and keeping it uber thin. I've gotten better results using peanut butter then friends of mine get using ICD7 or MX-4 or such, just cause I knew better what to do. Anyhow , some thermal pastes are special, like diamond ones - do not spread easily and will scratch and produce a lot of volume and alot of resistance physically. Some are super thin and grease like and or liquid metal. You could almost achieve the same results using a thing Bearing Oil. But this never stops the market from all the hype selling snake oils, erhm Snake Paste ?

 

Im off to zzz, but send me a PM which is which you need and I will see if I have it.

 

just readin the OP now, something seems off, those are VERY high temps, too high infact, From my experience, something is not being said, or information is missing in context. I've seen issues like this happen and last only a day, while people didn't bother installing the windows updates, that were stopping the fan profiles from running the fans, especially on alienwares. We had such an update just a week ago even, and you know windows they now install updates without telling you it happened, and not all of them are even found in the history.

Also, ambient temps, skill in thermal heatsink, how much of the temps from other components are bleeding into the chassis, are the fans making a jets taking off noise ? Because they usually do emit like 80+dba at like 100% fan speeds at 75C+ Thats almost like yelling.

I have small solutions to help people bleed some thermals into the chassis too, using TEP thermal modules that stick to the motherboardchassis and touch the back side of the GPU/Videocard too. These helps when in long gaming sessions.

Normal operating for a 980M when not more then 10-15% OC should be no higher then 70C when in load. 55C when idle. Stock you can minus more, more like 65C in load/ gaming DX12 max details...and like 45C idle. There is overwhelming number of people reporting thier temps to me in thier setups, so I'm well aware of what should be expected. The only time I see higher temps like your reporting is when soemone uses other vbios from other people. we have a full staff of multiple people hired and paid full time salaries to work on vBios, if you think using one you find on a site without knowing what your getting is better, then you should also seek the provider of that vbios out. We have dozens of vbioses for every card and are CONSTANTLY testing them. We have eliminated ALL throttling on all models we support, and it's been months if not years for most models. Did not need overclocks or extra voltage or more watts. (Not that those things do NOT affect performance, they do, just like not having enough power in the laptop or the CPU throttling can also make the GPU throttle. Throttling CPU means throttling GPU BTW, so keep that in mind too. Tetsing a CPU's max frequency for a test but WITHOUT GPU load it like a Shotput Athlete to throw darts, and see how far he can throw them and how much reserved energy he has)

 

I believe he got the vbios from prema

 

Exactly ... Prema vBios turned my GPU into a furnace as it makes the card run at 1.065V vs the stock vBios where the card peaks at 1.035V ... Maybe it's DELL lacking cooling system ... I would say switch back to stock vBios as apart from benching the difference in performance is minimal or switch to SLV7 vBios as it makes the card peak at 1V @1202 MHZ core clock & with the exemplary cards of @woodzstack 1V is quite enough to run at these clocks and the temps will drop significantly.

 

@Ashtrix, the problem is your heatsink.. For the 980M, you need the 3 pipe heatsink to keep it cool.. The 2 pipe 860M heatsink can't keep the 980M cool lol..

Sent from my LG-H850 using Tapatalk

 

Dude I'm having 3pipe.

 

Ah ok, thought the 860M came with 2 pipe heatsink..

I guess the best you can do is get some good Phobya XT 7W/mK pads, arrange them properly and get some good IC Diamond paste...

I would also reach out to Prema to get an UV 980M vBIOS... If @Prema is ok, @Ethrem could share his one and you can give that a shot...

 

No the 860M was 2pipe 75W, I had got a new triple pipe one for this. Ordered the Fujipoly btw..

 

I personally think Fujipoly is a waste of money.. It's just absurdly priced and the temp difference you get form it for the money is well minimal... Well it's your money

 

Yeap, if i post a picture of how many thermal pads i have stored up , people will have a seizure.
End of the day anything above 7w/mk doesnt really doesn't make a difference, since the heat sink cannot dissipate the heat fast enough to keep up as it is.

 

Mine is customized to the ASIC of my cards in my machine, its unlikely to be stable in another machine.

 

Every degree you can reduce..., 2-3 degrees lower temp in bench can be the difference between success or failure

 

I have a feeling is simply a poor quality heat sink not making good contact. It happens. I had this problem on my CPU heatsink recently. 4.2 ghz would give me temps that approach 100 degrees. Although it's a roll of the dice I purchased a new heatsink and now 4.5 nets me under 80 degrees under load.
If I were you I'd just buy a new heatsink. It's probably cheaper than experimenting with expensive Fijipoly pads which I think is unlikely the cause.

 

Could be that, for sure. Most common TBH

 

Seems more likely to me, Since the heat up is really too much of lately and never had this with running maxed out for long sessions !!

Heatsink and Fujipoly same price, I ordered it, Will test everything out now...thanks for the idea +rep.

I like this machine and a personal reason also makes me not to part with this soon, Hope I get things fixed.

 

I've had so many heatsinks in my SM-A... they all suck...

 

The truth is the voltage on the 980M is so obscenely high....without Prema's help, the card would have surely fried itself for now. I'm running undervolted same as Ethrem, though his may be a slightly stronger undervolt. I'm running stock boost cocks, but at 1.000V cap instead of that ridiculous 1.0675V it originally ran. Since then my black screens and temperatures have been phenomenal in the M18x R2, topping out around 73C in my warm room or even under 70 C in a cool room. 980M should not be volted so high, its just too much in such a confined space. Boost clocks on the standard prema vBIOS we are talking like 140W power consumption, no wonder nobody is able to keep it cool

 

I'm not that much further under you. Prema stopped at 975mV. I'm pretty sure these cards could go lower too. I still can't game though... Summer here in Denver is one of the hottest on record and our A/C just can't keep up... it regularly goes over 80F in my room now. My 4940MX is idling @ 50-55C...

 

I feel you man, the ambient temp in my room is 82+ because the vents do not work in my room. I'm ready for winter!

 

Yeah I love winter lol. I think if I got up and covered my skylight it would drop the temps in my room but then I'd never be forced out of bed instead of sleeping all day (sun hits me around 2PM) haha.

But yeah heat + electronics... we all know how that goes. My desktop has even seen a spike and its liquid cooled. Although the 780 Ti handles its heat well enough, the CPU has gone up 4~6C

 

Update -

Got the vBIOS from @woodzstack And as said by @Raidriar , the Prema's 1.067 stock voltage is high for the 980M. At-least for High ASIC (High leakage) cards I guess (77.1% here), Ran the Unigine Heaven BM with max focus / size / distance with ultra preset and ran in free form mode for 10 mins and this time no AC only Fans full blast. No bench nothing ran directly Heaven and avg fps was 25 with these settings maxedout iirc.

Comparison
Prema vBIOS (no OC) & GPU Z no util throttle
Temps - peaks at 84C and upon 85C card blackscreens.
Clocks - According to HWinfo the Clockspeeds are GPU Clock - 1126.4MHz / Memory Clock - 1252.8MHz / GPU video clock - 1.036.8MHz
Power - At 1.067v and 114W peak.

Woodz vBIOS (No OC) & GPU Z no util throttle
Temps - peaks at 79C no BSOD.
Clocks - GPU Clock - 1202.3MHz / Memory Clock - 1252.8MHz / GPU Video clock - 1106MHz
Power - At 1.000v and 104.x W peak 10mins-15mins approx

Any thoughts on this ?

P.S - For the maximum load I always traversed on that rocky path.

 

I'm not seeing a question here...? The second vBIOS is probably the better one to use at this point. Keep your temperatures as low as possible, always

 

What is the ambient temperature?
79c is still high for a 980M @ 1.000v

 

weird, not usual. Don't know what it could be try different drivers or check what was changed/updated since.

 

I see, I didn't check them dang !
But I expect it to be around 76-78F Max.

Weird in which sense ? I couldn't get you there... I used the 365.19 driver nothing else, I keep my Windows free of updates since the days of 7. Temp peaked that much due to consistent load with overlong and sharp rocky path, I felt it a bit smoother (tiny amount, sorry I didn't note down the numbers, Will try to run again with this vBIOS not with that extra volts ever again) maybe my eyes, GPU might be stressed with all that extra voltage..

My battery is not going to desktop mode, That's a change to mention which is another sad story.
More here, If interested.

 

I think your MISSING updates. Windows will prepare itself for updates by removing things that need to be updated that it has access too, even if you use a registry hack to prevent it from downloading and installing the updates, it can uninstall software and other things automatically, leaving you with poor performance.

My concern with your issue, is it was working initially correct ? and then started getting high temps, maybe, if you like, you can send me your laptop, and I can fix it, after finding 100% what the issue is, and return it. But the amount of trouble shooting, without more info is going to be difficult, most our our first conclusions are going to deduct its user error in installing the videocard and or missing software updates or bad drivers. Like 365 was never a great driver to begin with, and i's not old but its not new either.
Are your results the same with other drivers ? What about after updating windows fully and updating the AWCC and OSD and such too. Are you 100% sure BOTH GPU and CPU fans are running full, and not just the CPU fan, is there blockage, is the fan dirty or what is the ambient temps. I do realise it's summer outside and some people are having heatwaves, and your getting hotter temps.. that CAN be it too. ALSO - I noticed your 10-20% overclocked (which should mean higher temps too)

Also I would like to point out, that our stock driver did better then Prema's in what you reported. Higher clocks AND less temps with LESS voltage, so maybe you should stop using the other vbios's. Maybe keep the laptop in a well vented area, and have something cool blowing air on it from underneath.

 

79C is NOT hot for a 980M running Heaven. If you've maxed it out, Heaven will beat the crap out of any card. It is a worst case scenario for gaming (furmark doesn't count because it doesn't simulate gaming loads). Usually whatever you get in Heaven at max is what optimized titles like Witcher 3 and ROTT will max out at as well. Dropping AA settings will massively effect temperatures too.

 

First the first BSOD, that didn't happen before and the temp raised to 85C and then bam that was the first case, never before that and I don't remember the card temps unfortunately but I suppose it was 72C first time, got this from my post here, but later on temps increased to 78C that I remember. So I think this is a new issue, my guess, Sorry I can't be perfect.

And I will be needing the computer, all my work is saved on this machine only I guess I can't do that unfortunately... :-(, Thank you for asking though.

I will try with both drivers tomorrow (362 and 365, I chose this 365 because it's the last stable driver of that branch release), Fans are 100% maxed, no doubt about that. I keep the machine elevated by 30 degrees, fans are new as well both are from R4 and not even a year has passed, CPU fan is newer than the GPU & less dust no clogging, I can feel the air blow from behind the machine.

EDIT : AWCC is a bit older version (1 step back that's all) but I downgraded long back just a personal preference nothing much, It was fine since this version was being used since long time. OSD is the same since Win7 and the 8.1 too.

Yes, My bad I forgot to mention that before maxing the Heaven I ran it with max AA and the temps were at 75-76 (started from 60s then to 68s later upper 70s, the run was quite long, wanted to make sure the temps don't increase on the voltage drop, no AC or no cooling pads nothing wanted to get a natural run) and then lateron when increasing those size / distance and the other factor the frames dropped and the temps increased.

I was very busy today, still wanted to test this out so didn't noted the numbers and time, other params. Will try to get a more streamlined run.

So, 79C heaven maxed is not good for a 980M ? Any ballpark estimate ?
Will install TW3 today and run that too tomorrow for more clarity I suppose (Upto 21 of July things are busy for me, Will try my best)


EDIT : The summer impact is more than usual, I guess that hit most of us with these high TDP rated machines unless a 24/7 cool AC, Back at homeplace, I * always* ran with an AC while gaming to make sure the life of these machines stay good, here It's not the case.

 

Try firestrike ultra stress test, it'll draw much more power than heaven on 4k + max tess

 

Prema's vbios removes the power limit entirely so lower voltage can still pass the power limit that would normally be there and increase temps.

The summer here in Denver has really made my temps go up. Even my liquid cooled desktop CPU went up 4~6C and my 4940MX is idling at 50-55C so I just use the desktop since my 780 Ti fans can just increase their speed to dissipate their heat. The laptop I just use for stuff involving my phone since my desktop is running W10 insider preview. Once it cools down, my temps will go down again.

 

Are you sure this isn't SVL7 vBios ? cuz as far as I know these are the exact attributes of his vBios and it also raises the temps but not as high as Prema one ... Woodz one usually runs stock clocks not with the 1202 overclock.

 

There were 2 vBIOSes from Woodz one was OC which is this one and there was other one, Will flash that too and report, Thanks for reminding me that, I forgot to put that point in that run.

 

Yeah, that clock and 1.0V sounds like the slv7 vBIOS I've tried not so long ago. IIRC, it was crashing through a 3dMark 11 run, I suppose the voltage wasn't high enough as I think once increased to 1.055 or so it passed.

 

The older SVL7 vBIOS (and older @Prema vBIOS) is unstable with newer drivers. I believe those stability issues started with drivers optimized for the latest Tomb Raider game. Too much voltage can cause black screen problems, but 1.000V is not too much voltage. The black screen issues were the early warning signs of 980M GPUs that had problems for me. Definitely could have a bad or failing MOSFET that is putting too much load on the others. 980M was designed for 6 but for some reason they only use 3 and leave the other 3 solder pads empty. This issue is not caused by overclocking (as a clown recently suggested in another thread) as much as low grade components and cutting corners on the design (like leaving 3 MOSFETs off the PCB). Overclocking can accelerate issues with gimped hardware, but this happens to some folks that never overclock.

 

980Ms were designed to fail. End of story.

 

I think we will see the timer run out on more of them following the full release of Pascal mobile. We're just waiting on the countdown for the GeForce Maxwell cancer drivers to be released.

 

I'm not going to disagree with you. Glad I have my warranty through May 2017 lol

 

That is too hot for 980M with an AW17 R1 as far as I know. By that mark, it's hotter than Kepler. I'm constantly around 88F in my room with really high humidity and the ONLY time I've seen 85c on even one of my cards in the last years or so is during Evolve at 120fps a couple of days ago during a 90F room temperature. 980M *IS* cooler than 780M; this would mean if you put a 780M in that machine with svl7's vBIOS you'd be overheating (92c+) at stock speeds in a BUNCH of titles.

What paste are you using on this again? As far as I know, my PC's cooling is inferior, my room is hotter/more humid, my cards are hotter, AND you're getting higher temps?