980M Blackscreen on 84C+

Looking at the 680M i'm seeing 4 MOSFETs (correct me if i'm wrong) This is just plain plain stupid NVIDIA.
During Vbios trials, Prema told me that MOSFETS reach their stable supply limit, thus causing a black screen.
Running Firestrike @1420/2905 @1.17v the card would blackscreen almost immediately on the First test.

 

Yup, and higher ASIC really isn't all that like some people think. 980M was a nice upgrade for 780M owners, but NVIDIA seems to like gimping stuff to make things look better on the next upgrade and not have to try as hard to compete with themselves.

 

Mr. Fox I've also read about your failed 980M card which was alive when it had a MOSFET replaced. Next is So if my card is getting hotter by this Prema mod could be happening but Since I hadn't noticed this with the same game back when I ran the card with Prema vBIOS only, So I can add the Tbonesan's point to this maybe the heatsink is gigving out the issues but afaik Alienware heatsinks are good quality, also to say with that slight OC (both the Prema, woodzOC vbios) the card is hitting limits basically regarding the driver I ran 362.00 and then the 369.15 that's all nothing more or less drivers thus Premamodv2 only (I understand that AW17 can't make the 980M work like in the p77xDM/ZM idk what's the case, that's another flip-side of the AW mess)

Further if the 1.065 is too much for my card I think I can't get this high leakage card to OC well I guess If it's the MOSFET fault, Do these MOSFETs die out so quickly ? Then Nvidia is running an Abomination and people who are running stock everything don't even know that their HW is gimped to core and the BGA is just DOA, AMD can't compete at all now I'm afraid this cancer is spreading very fast from the Green greedia corp, Mobile spectrum is really weak with all these factory gimping, Pascal unstable clocks, high temps, DPC latency, Fan failures on Scammers editions, Overpricing, Omg, these guys are worst than that Micro$lot,

Also Am I only one experiencing the blackscreens on 85C mark ? Or every 980M does exhibit this behavior, I don't get if they do how people manage to cool this beast and OC it. I'm not an Overclocker but curious and long path to OCing for me, Excuse me if these are stupid questions.

Gelid GC Extreme paste only, Also to mention before it was 64-68C on Dota2 with same ambient and with AC a 2C less that's all and now It's topping only at a 60C with Dota2 same settings just the voltage is less now. I can remember one thing on the day I had the card with the stock AW17 1.6 CFM fan I was getting an 80-83C with Remember Me, old game immediately I changed the fans from the M17x R4 which is 11.4CFM, dropped to 70+Cs.

 

I see. Test out Evolve Stage 2 and see how it works. Run a 120fps limit. If you're hitting 85c in a match something's wrong, definitely.

 

The last version they released, was for windows 10 AFAIK, and they held back from updating for ages, until finally updating that too. You might need it. Latest I mean.

As for your issue, check your contact to the heatsink, maybe check the entire card, do a full inspection. Try a different card and compare the results. There has got to be something we can put our finger on here.
Like, why are you having these issues and no one else ? (don't get me wrong, many many people have issues, but if i start listing whats the most common cases of issues, it would be like I'm accusing you of amateur mistakes, so I will skip that part)

Anyways, when overclcoking every card is different, but you need to clear the way power wise for your system. Clean power from a VERY good modulated sine wave breaker and powerbar, cool temps, a small portable freezer with the rack and an extra thick rubber extension cord, some experience using CLU too, maybe even a pro- copper shim job, a TEP module under the card or a copper backplate making contact to the chassis with thermal pads on BOTH sides of the plate AND both sides of the heatsink ( to slowly cool about 1C extra into the chassis of the laptop..every bit makes a difference right)

don't forget the cooling on the voltage regs and the mosfets on the motherboard too, those affect your OC BIG time... the more efficient your power coming in, the cleaner it is, the less heat you will produce and the longer your components will last too. The PSU adapters for these laptops are ****. Don;t think they give very good clean power, because they really don't, they give worse a cheap PSU gives you at best no better then a Bronze would give you.

 

I reached the 85 mark before but never had blackscreens. As for the temps, summer here in Germany is very hot and AC is nonexistent so with any modified vBios whether it is SVL7 or Prema the card is running in the 80s for seriously heavy games like battlefield and witcher 3 and 50s-60s for low demanding games like diablo 3 and LoL. SVL7 vBios is stable for me on all driver releases, not sure about Prema's stability with drivers as I found that the temperature hit from that vBios to be very huge compared to the lower voltaged vBios of SVL7 so I ditched Prema's very fast.

Running @woodzstack vBios changes the above situation completely as the card never reaches the 80c mark at all due to some sort of dynamic clock modulation or if you would like to call it intermittent throttling . Personally I didn't feel the impact of such throttling while playing (Didn't have stuttering or lag ...) So eventually I decided to use the stock vBios till one day the time comes when the 980m is completely obsolete and really needs to be pushed with a modded vBios to get these extra frames but atm. I don't believe you need to.

I tried this card with every non metal thermal paste imaginable and diff. types of thermal pads ... Thermal delta between all of the cases is minimal to almost nonexistent ... I would recommend that you flash the stock vBios and have a long gaming session and see what happens.

 

I've reached 92C and never black screened, it's definitely not a good sign.

 

I changed it, 92, not 94. 94 was the 880M.

 

Ran Evolve Stage 2 just now, 2 bot games one as monster and one as hunter with maxed out (normal SMAA - monster then TXAA - hunter, FOV - 90 on both) and started at 70s and ramped up to 80s and It blackscreened again at some post 83-84mark FPS was fine at some 70-80's..

With 1.000v vBIOS, Is the card failing ? I tried 2 repaste jobs as said it in first post and the heatsink was tight to remove as well due to the thermal paste, no chance of low pressure because the Clevo Xbrace needs tight pressure.

Coupled with the battery issue, I'm in a huge mess right now, very unfortunate !!

P.S : Ambient 71F

EDIT : Idles at 45-50 (In Fans Auto since AW jacked up the EC, they will only start pushing more when 50C is reached dials back to 45 and repeat such a stupid botched job Dell programmers did, if manual at 2000RPM (no options either 0,2k,max this EC) it stays at 35-37 depending on the ambient, cannot use this because It doesn't help much with a constant 2k RPM as CPU needs more sometimes..Hotwell. Unfotunately another EC is not possible even with more granular RPM with single slider for both CPU and GPU because the GPU fan wont spin on both manual and auto with this Maxwell card)

 

Ok, so yes. You're officially getting as hot as I am in a MUCH cooler, much less humid scenario with a generally much cooler card, running the SAME voltage (1v) that I am.

Something is wrong with the cooling on the 980M. I suggest re-checking the heatsink contacts and the thermal pads on them. There is no way that my currently-unmodified (and thus worse) cooling, in my hotter, more humid environment, with a hotter card, is MAXING at the same temps you are getting shutdowns with, and ONLY in one game.

 

I will do it, Ill get some prep pads, will do that work day after tomorrow with the existing heatsink and TIM and couple of leftover OEM pads. Ordered another heatsink too, I guess I should contact Eurocom for the bottom copper sheet they offer with thermal pads ( Here at 1:41).

Any signs of failure of the card from my given experience so far ? Others seem to have it past 85C without any issues...And prema's mod has 92C throttle limit on Maxwell too.

@Solo wing , did your card blackscreened at 73Cs ? Can I know what's your ASIC ?
Thanks.

 

Those are pretty high for dota 2 , my 980 with the OC in my SIG tops out at 60C when playing dota 2 all maxed out at 1440p.

 

Before I changed my heatsink (my old one was just inferior) and propped up my laptop, I used to get 85c somewhat often, depending on what I played. My cards are about two months short of being 3 years old now, and both have no issues. I also spent entire days in GTA V at 88c when waiting in the Heist lobbies because GTA would shove itself at 180fps and my cards would get hot as hell, and then when I was out of the lobbies I alt-tabbed and added an overclock and played in the mid 80s in actual gameplay.

85c is not the temp for danger to any current mobile nVidia card that I know of. My 280M also spent MUCH time above 100c (yes you read correctly) without any damage to it after already being 2-3 years old because Borderlands made it so hot. As far as I know the card still works even, and the board was the reason my D900F died.

OF COURSE, lower temperatures is better. Never mistake this. But 85c is most definitely not danger zone, and 87c is the default thermal throttle on stock vBIOS in the first place (93 or 94c is the danger point for Maxwell AFAIK, which is why Prema's vBIOS sets it to 92c), so you aren't even close to TJMax at 85c.

To be fair though, your cooling system beyond trumps his. But yes, it's still hot for DoTA 2.

 

Ah true, i thought he had a clevo. My bad.

 

Too much thermal paste can act too much as an insulator too. You want to use the least amount you can while having it spread perfectly thin and EVENLY enough to almost be see through and yet make PERFECT contact with the entire surface of the heatsink. To give an example of how much paste to use, it should be about 5-10% of what's inside a 1.5g tube of thermal paste. From my experience, thats about 5 x less then what people usually use, and a major reason also - why they're 5C-10C-15C hotter, too.

On another note, you mentioned you have the heatsink on very tight, or have led me to believe that. The amount of pressure you want on that heatsink, is only enough that it wouldn't fall off if you held it upside down. Enough to hold it's own weight. Anymore then that and you very likely will crack the GPU die itself on the corners or edges eventually and then random issues will start until it dies.

Another thing to check, is if the voltage regulators and mosfetts - if any of the transistors are not being cooled. while the reported temps might indicate high temps, the GPU sometimes has to pull more power and double its work load if the other components are not working properly, also they would most likely cause black screens and crashes if they got too hot or were misaligned or shorting or damaged. When the GPU gets too hot, especially if gradually, it would produce artifacts that you could visibly see on the screen. I hope that helps determine what might be the cause. However - in cases of complete shut off, the GPU can cause that too if it gets too hot, so maybe we misdiagnosed it, or not, it's hard to say. Do a thorough inspection, show us a picture of the card both sides in high resolution, cleaned, so we can see it to be sure.

 

I can assure you that TIM job, heatsink pressure are fine, I doubt the rear chips, I will take pictures of the card day after tomorrow and post it & no artifacts, only just blackscreen and audio plays on in a loop or just stuck that point of music, like a part of music that plays on in case a game freezes or pauses, but no chipped audio, have to shutdown manually. Hope my GPU is fine..

 

I've had my 980Ms (briefly) at 86-87 mark, and they never blackscreened. Sounds like your card MAY be on its way out, unfortunately. All you can do is lower voltage/clocks and hope it runs till its kicks the bucket.

 

How can we know if a GPU dies of a specific cause - Core / vRAM / MOSFET or someother case ?

EDIT : Very sad this is happening, hoping for an Overheat instead of a failing card...In talks with Eurocom for that pad.

 

x-ray will show core or vram defects, MOSFET I think can be checked with multimeter

 

what pad are you looking for ?

 

This

 

ASIC 76.4%
My problem was whenever the core reached >73c I get a black screen, computer locks-up and I do a hard shutdown. I don't know if the cause is MOSFETs getting hot or reaching their current limit. (guys, am I correct in thinking that resistance rises with increase in temps?? ) I also believe that the root cause was a crappy design on Nvidia's part. Overclocked Kepler and a Desktop Fermi like crazy and never had issues.
I then got Gelid extreme TIM + Fujipoly thermal pads, running several games 1262/2905 @ 1.06v for at least 30-60 min straight no more issues. Skyrim @ 4k & DAI @WQHD so far so good.
Was it a bad paste job or thermal pads I can't say for sure, It might not be directly related but it did the trick for me.
SVL7 vbios ran cooler but overclocking was a headache, the card had a mind of it's own throttled like hell. Prema's proved to be the better one I believe.
Max temps 70c @100%, ambient 16c-21c.

 

I've tightened my heatsinks pretty tight on my GPU and never had any cracked GPU dies or anything, even with repasting a few times/testing different pastes, etc. I wouldn't be too afraid unless you're like using a power drill or using so much pressure that you literally hear it crack.

 

Before we started telling people to be careful, we used to get a lot of broken GPU die's back with the 700M series. So, I would say since I've only seen a single case in the last year with a 900M series it is a worthwhile thing to mention. Prevention is always good.

From what I understand with ASIC% the temps will bleed easier at lower % , meaning the card will get hot all over unless you pull it (the heat) away quickly, and with higher ASIC% it's a tighter release, more important to make proper contact, because it will get hot where it is supposed to and less elsewhere, at least in practice this is what I have understood.
Higher ASIC% is still better, but just cause a Ferrari is a better, maybe more efficient car, doesn't mean we should all drive one either.
On another science, my sources suggest ASIC% is equal to or greater than SNAKEOIL%.

 

+ rep for sharing this info, thank you, I think we are having a MOSFET issue which seems more likely over damaged core / vrms.

But those temps on 980M, Amazing !!
I'm waiting for the heatsink now and will pick up some prep pads free of lint too, gotta test many things now..

Eurocom gave me the price of that thermal cooling pad for 45USD plus shipping, Since I don't have the fat pads like the ones that AW give for 780M / 680M /880M, I think I may have to get that Eurocom's pad, maybe the stock pads at back are not enough for cooling. But that's ridiculously expensive.

 

Look for Phobya XT pads.. Got them off ebay, German seller.. They will be cheaper then Eurocom's and better.. You can get 0.5/1mm ones and stack accordingly..

Sent from my LG-H850 using Tapatalk

 

Update : Modded a U3 Plus with EK Vardar 2200RPM 120mm x3 fans & Silverstone filters, first test results.

Ambient 72-75F.

IDLE (notebook fans on auto)
33-36C GPU
40-44C CPU

Load (notebook fans maxed)
Heaven all maxed out.
61-62C max CPU
73C max GPU (Ran a 35-40Min session on Heaven, with all tesselation size / distance / factor jacked to highest, before that temps started at 63C and went all the way upto 73C when I was looking at that dragon and the houses with a wide fov point making the GPU stress a lot for nearly 15-20 mins & then ran a bench)

Heaven Benchmark (Since the GPU ws hot before, upto 73C but during the Bench temps dropped to 68-69C)
HWInfo reading

Will try to run Evolve maxed now,

P.S : The underlying issue needs to be fixed can't carry the cooler pad always..., Waiting for the heatsink.

 

Holy crap that's a huge drop in temps! I should look into that mod myself...

 

firestrike ultra stress test will draw much more power than heaven on 4k + max tess


there's also that one scene in metro last light redux where you're asked to board regina, go behind it and stare at the light. on my 1.274v TX I remember it hitting 390-400w compared to heaven max 360w

I'd try not to stack pads, but avoid the fujipoly 11w/mk and thermal grizzly minus 8, they don't compress very well

 

Now THOSE temps look like what an AW17 R1 should be getting.

 

So after a long session of gaming the PCH temperature was 94.6c. Something is not right, a bit of searching around I found this:
http://forum.notebookreview.com/threads/m17x-r4-pch-temps.678988/
This bloke ↑↑↑ mentioned black screen/locking up.
http://forum.notebookreview.com/thr...g-high-pch-temperature-in-w860-w870cu.709654/
After Heaven/valley benchmark + GPU-z render test I noticed PCH temps increasing rapidly into the 90s. Temperature increase is directly related to GPU load, and has nothing to do with loading the CPU.
I guess it's due to the die being under the GPU on the opposite side, close to a mechanical drive and below the keyboard.
There's a large silicon pad under the GPU that came with the stock 680M, I thought it might be transferring heat between the PCH & GPU. Replaced it with 1.5mm vram pads, the PCH ran 8 degrees cooler at 86.5c !
My next step would be to completely dismantle the computer, taking the outer shell and replacing the pads/paste or whatever dell put there.

 

http://forum.notebookreview.com/threads/best-notebook-coolers-for-18-alienware-laptops.593626/

Dell gimped the EC completely from 2013, the auto fans don't spin when needed properly, Else the idle temps would've been fine. Load temps on GPU and CPU those need AC and the GPU (upgraded) one is really a careful work, else the temps would be like mine... :-(

That's a nice find, Unfortunately the Hotwell has PCH inside the CPU thus it operates like a furnace, I have the old thermal pads and 1 is low quality of 2 below the GPU, need to get a new set of good pads for this bottom one or that Eurocom option, I think the bottom vRAMs are getting hotter, So much of ambiguity and variables to iron out, but hey It's good to know things, the way I see it now.

Hope you can fix that Issue else It might be the first case, GL !

 

I actually automatically assumed anyone upgrading an AW17 R1 would be using a HWiNFO to control their fanspeeds. I'm pretty sure @Mr. Fox always talked about that. Maybe you should look into it if you aren't already using it?

No, this isn't correct. Blast Furnace has the PCH on the motherboard; I know exactly where my PCH is in my laptop. What it has inside the CPU is the VRMs, and the only ones on the board simply translate power from 12v to 1.8v to the chip. PCH cooling is indeed something that should happen. However, I don't know if it's what causes your problems. I once put a video to render, loaded up a livestream to watch, and proceeded to play CoD: Black Ops 2 multiplayer at 120fps maxed out. At the same time. My PCH arrived at a lovely 105c and sat there for about half hour while I did everything. If 95c PCH is cutting off your machine, I'd say the problem lies elsewhere (like a defective board).

 

I'm using the HWinfo, But here's the catch 2 ECs - Dell EC / Compal EC - Compal EC - more granular RPM but same slider for CPU and GPU, Also won't work with GPU fan IF Maxwell for both custom or auto profiles, Dell EC - Only 3 RPM presets - 0,2k and highest 4K for CPU / 3.4K for GPU. So cannot use that Compal (Once you start a custom RPM table It gets bugged out need to get fresh INF again) and It introduces the DPC also Dell does but disabling monitoring from HWInfo fixes the hiccups. Finally I use Dell one and set it to Auto and when gaming or loads I max the fans.

Thanks for correcting, I just checked the Mobo pictures It sits right on the other side of the board (Keyboard area, right side) where GPU sits, And Ran Evolve Stage 2 maxedout, temps were fine with a small tutorial for 10mins then GPU and PCH temps are mostly directly proportional, When the GPU was at 81C PCH was also at same, I didn't want to do the crash again so stopped there (This run was w.o CM U3 mod) now with the Fan mod temps sat at 71C GPU at that point PCH was also at 70-71C and PCH was slowly increasing compared to GPU temp curve, took 4-5 tutorial matches (monster vs hunter).

I don't get the PCH temps because there's a plastic film sandwiched between the GPU also no big thermal pads on the back of the GPU to transfer the heat, only the regular pads. Confusing as hell.

EDIT : Tried to pressure the machine, Ran FS Ultra (Steam) GPU peaked at 123W, way more than Prema mod at 114 (Does still MOSFET fault can be there or is it the Core / vRAM cooling ?) and PCH maxed at 64C, running a BF1 4K video on Youtube (12 tabs open) with a Twitch stream running, a FHD movie conversion to x265 (from ext HDD to int HDD, USB 2.0), Running a full modded F2K (FLAC) with waveform which uses DX & CPU tops at 78C-80C, 100% usage, couldn't recall the GPU max, here's the graph. I should have reset the HW counters for the perfect reading, My bad.
http://www.3dmark.com/3dm/13556912?

 

Well looks like you're on the right track. Still check the chip itself and see what's going on. I hope you can find a better fix.

 

Waiting for the heatsink, Just to mention the PCH temps when GPU is on 55C in Dota2 it goes over 63C. So It's almost in the same ballpark.

 

Those seem to be pretty good for the AW PCH.

 

have you tried to use thermal pads to connect the PCH to the back of kb (p370)

 

I know I could, but I haven't bothered. I usually don't get any problems with its heat, though it is better that it's cooled than it isn't. It was only that one time =D. I was only using it as a point of reference though that a PCH hitting 85c and causing a blackscreen is a defective mobo of some kind.

 

Haswell PCH is rated for 6W and approximately 106C max. If you aren't hitting 100s, you should not have any problems. The tjmax on the PCH is always higher than the CPU.

 

@Astrix

Black screen here at 82 Degree's without fail! WTF

 

Sad, This Ngreedia and their dirty practices..as Mr. Fox said the mobile cards of Maxwell were dumpyard of the GM204, and I thought only 980Ms are affected seems like 970's are also hit, Ngreedia skimped out on everything thinking we don't know like that 970 Fiasco, what a bunch cash grabbing fools..

I still don't get whether it's the VRMs or the vRAMs / MOSFETs or the Core, Do you have the rear pads ?

If Yes then I presume It's either the MOSFET or the VRMs mostly else the Core, Since we know if vRAMs are at fault we can see artifacts or crashes, Also Mr Fox's card died prematurely and with a replaced MOSFET and an extra one (dunno whether it was vrm or mosfet) It does well, and Nikos from the M18x R2 he got his 980M's Hardmodded with some semiconductor company with extra VRMs for maximum stability.

Some one more with hardware expertise / personal experience can sort this out clearer.

I got the heatsink here but since travel plans, work and all I will do it after 19th, Eurocom thermal solution is very expensive with shipping of 20$+49$, Ill get Phobya XT 7 mW/k pads, for the back. **** you Nvidia for being a douchebag !!

Spoiler: OT

That DP of yours it let me to this track, Thought it's related to watchdogs's Defalt lol & It's Insanely good, given It's a hybrid trap with the superb visuals, great channel for electronic music lovers with fantastic visuals too , Give it a shot you might like it...

EDIT:
@Mr. Fox , If you don't mind can you please share your experience with the GM204 mobile card failure and fix, thank you.

 

Check the RJTech website.

http://rjtech.com/shop/index.php?dispatch=categories.view&category_id=179

Much more reasonable prices.

 

I told you to get the XT pads, they're much cheaper then the crap Eurocom offers.. Also get some good ICD paste...

Sent from my LG-H850 using Tapatalk

 

Do they also do that enhanced cooling pads ? Will look at that inventory. Thanks !

Yeah, I'll get those Phobya pads. ICD7 I guess I'll go with Kyronaut this time

 

No idea about the pads, but the Phobya's are way better than the stock ones.

Main thing with whatever pads you use is that every component that needs to be cooled has to make contact with the heatsink.

 

Also see the link in the post about pads http://forum.notebookreview.com/thr...aptops-notebooks.784799/page-16#post-10306750

 

It says those Fujipoly again, I guess If I have to get those I need to get them from Frozen CPU which is expensive with shipping more or less at the Eurocom's price (I asked Eurocom for discount but they said we could only offer NBR discount to upgradekits or laptops), Plus Amazon shipped these (a seller) with a damn cardboard sandwiched in b/w there was this thermal pad. I was shocked to see that packaging and returned it, That's why I'm looking at those Phobya atm.

 

I thought more about to show you what decent better thermal pads can do for hardware. Sorry I don't explained that

 

frozencpu is dead
here you can buy the fujipoly from a real shop, just check if it's 17w/mk or not: http://www.performance-pcs.com/fuji...ize-sheet-60-x-50-x-1-5mm.html#Specifications