980M Blackscreen on 84C+

Its is not dead, i order from them every now and then and they deliver super fast.

Not sure where you are getting that from.

On the other hand, i had to start a claim against Performance PCS twice. Once for a tube of Kryonaut and once for a water cooler.
When i called to check where my order was they gave me a fake tracking number and shouted at me calling me a liar and a "f u c king" homeless for bothering them.
Mind you i only called once. For the first time, after i received no reply from their sales and support email address.


@Ashtrix Fujipoly pads will be around the same cost every where, the one i have were bought from FrozenCPU.
I would avoid Performance PCS like plague.

 

Wherever it is, the price for some of these pieces of rubber isn't worth $100+++... Phobya XT pads are good enough and like 1/10 of the price..

 

With the stock un modded head stinks, yeah not worth it. But if you can get a tight fit, then its worth it.

Majority of the cases though, not needed by a long shot.

 

I still think Fujipoly is just too bloody expensive for some rubber pads.. Even for the improvement you get.. If it was easier to get and 50% cheaper, I could consider getting it....



Sent from my LG-H850 using Tapatalk

 

Update :
I guess my 980M is failing, Just now I've noticed some weird rendering, It's like all over the screen something like mosaic and a restart took it normal (My bad didn't took a snap) and also to mention ROTTR is crashing randomly (ran the game today only) dunno what's the deal with the ROTTR (Tried the 368.81 and the 369.x desktop driver but nope It just randomly crashes, Display Driver failure, says reduce the GFX effects and some resourcing memory error, not related to temperature doesn't happen with any game) Will try TW3 and max it out and then will update.

Ran Dota2, Technomancer, Hard Reset, Evolve stage 2, Borderlands and max mem reached to 7400 but later upon launching the MGS Ground Zeros I couldn't check for the GPUz and later on evolve auto closed maybe lack of vRAM but rest games were running without any crash runtime was for 5min-10mins avg.

The most intimidating part was that Desktop wonky mosaic rendering, WTF was that IDK, When I perform the heatsink swap or thermal pads I'll re-seat the card.

Edit : Could it be RAM ? Damn that Memtest will take a whole day out.

 

Did abuse that chip somehow? Cuz as far as I know you have been going stock (clocks, voltages ... etc.) ever since you got it so I find that very strange

 

Well I flashed Prema mod since day 1, I didnt experince any problems until this started happening, As you can see I'm just trying to stress test and a few basics to iron out any issues, conclude that the same - B$ yield of 980ms, new thermal pads set, heatsinks etc and. This one time bad rendering one is a completely new scenario again. No abuse at all, I didn't even touch any voltages racking em up too high or clocks, only stock with all vBIOSes both Prema and woodz OCed-stock vBIOS (both light OCed by default)

And the Benches I ran are Unigine Heaven, Firestrike Ultra and standard no Furmark, have 3DM11 got it from steam sale this year but this borked HW makes me not to test atm..

 

I know this is a really late reply, deep in the thread, but my friend had the same issue with one of their 980m's in SLI and it ended up being a warped and leaking heatsink.

 

Update 2 : Same windows issue, photos attached. But I forgot to check with HDMI T_T
https://imgur.com/a/8cDXk, did any of the vRAM modules got fried ? Ran Windows mem test and it reports fine, & a restart gets things back to normal WTF.

 

Do you have any MXM card to test or can you take the card completely out and try with intel iGPU. (On mobile so can't see signatures, if you are running 120HZ forget the intel iGPU idea).

 

No, my 860M is lying dead, and 60Hz only here, that windows rendering I couldn't notice it in BIOS iirc plus doesn't happen always, I found it (pattern maybe) happening when the Rottr crashes and next day morning while powering up it shows up this mess, I don't have the thermal pads for bottom and pads atm & the cleaning kit, I'm planning to get the repaste and heatsink test after 19 of this month.

Spoiler: A theory...on win8.1

I think the Win8.1 may also be culprit while that rottr crash if its messing with the Windows memory pagefile and causing corruption, don't know because Crysis 3 back button bug here...

 

I personally have no vRam pads on the bottom of the card, didn't think it was this important.

 

Rottr again crashed (with settings - all maxed except AA which is on FXAA and no Motion blur), new install and now waiting for that messed up win rendering if it shows up, I think it's an OS game compatibility issue coz TW3 runs without any hassle and all those games too..

Update : Confirmed ! that desktop corrupted rendering happens only when this game crashes and after a full shutdown and removing the PSU and then powering on, doesn't happen in BIOS and doesn't repeat itself unless there's this Nvidia Driver error and revcovery. Happens only with this game, As I guessed that game corruption is leaving some mark even after uninstall and does that.

At-least this is for sure.

 

Most heat sinks (especially the thin ones that came with my P170SM-A are not flat. This makes for a very poor thermal interface. I shaved off several degrees by lapping my heatsink. I had about an hour in it total.

 

From what I understand, it's extremely important.

 

Even my old GTX880m didn't have bottom pads

 

My motto has always been having more vRAM pads does less harm than having none!!

Sent from my LG-H850 using Tapatalk

 

I also put pads on the bottom of the VRM chips on the nvidia cards. I also don't understand Dell's logic of "lets cool one side of VRMs with the bracket, but not the other!"

Seriously I don't understand how some people design things. Would it be that hard to make a x-bracket that covers BOTH sets of VRMS on the bottom of the card?

 

What's the point of having pads without a heatsink to dissipate the heat?
DELL did it completely wrong so I would say the space below the card is a good enough alternative to be a medium to dissipate the heat from the vRAMs because there is no metal to take the heat from the pads.

 

@Ashtrix did you find a solution for this blackscreen problem? I'm facing the same problem when using Prema v2 vbios. As soon as I OC 80mhz or more on clock it gives me blackscreen when running 3dmark test.

 

Premas Vbios runs @ 1.062v which could be high for many cards. Thanks to N$hittia the crappy 3 out of 6 MOSFETS quit on me after 50min of load.
I'm using johnksss & svl7 vbios @ 1.000v and never had a single blackscreen for several months now, except when running 1.087v and above.

 

What driver are you using?

 

I didn't get time to replace the thermal pads on my machine since that last not so good TIM job (If you read the thread you'd understand what mess I was being through). Better thermal pads and contact will reduce temps by a huge margin but need to be precise on that, And sad to hear that your 980M also Blackscreens, It's because of the MOSFETs As solowing said, That's true for all 980Ms, we are just unlucky but If you get them replaced the 980M would scream like Mr. Fox's hard modded 1500MHz OCed cards !!

And for the vBIOS I'm using the Woodstack's, So far everything is good since I have this Custom U3+

 

So sad that all 980Ms are like this. Can you tell me your score on 3dmark11 ?
And do you know what is the best score for one 980M with just OC vbios and without modding the MOSFETs ?

 

Here's my latest and last 3DM FS with the driver below I'm using - http://www.3dmark.com/3dm/13923375

 

eh? I managed 9400 on the AW18 4710MQ+980M

 

GPU Score only. That's On Win8.x It sucks with that OS afaik.

 

11.3K

http://www.3dmark.com/3dm/11992911

 

I'm not sure if the M18x cooling is different from other Alienwares, but with fresh antec formula 7, fresh ebay pads, slight elevation at the back, and undervolt to 1.000V max (courtesy of Prema), I top out at 65F now under full load.

 

Mine's pure stock (maybe a few ~100MHz slight OC clocks as iirc Svl7 vBIOS has same speed as 1202 on core), there are differences in those clock speeds on both. Maybe the readings are not precise too..

Maybe since the old machines had the much more robust cooling with high CFM fans, But to let you guys know these are my current temps with the M17x R4's GPU and CPU fans (Ranger's fans are really low CFM) with the Custom Cooler Pad on top of it, DOOM maxes out (95% ultra) with 90-100FPS OpenGL at 64-67C no AC, Ambient maybe arnd 60-70F, And 65F/18C on the GPU ??

 

Hey you all have good score! Is it because you're using 330W psu ?

@Ashtrix so this score is achieved with svl7 vbios? The one we can get on techinferno right?

@Mobius 1 what vbios are you using? What driver version?

 

No the 330W PSU is only a safe investment, 240W one should suffice, 330W is for moar power when you OC the CPU like 3940XM on R4 and 4930MX on Ranger systems. Plus the 330W will have less load (more life expectancy) compared to the 240W PSU in a similar setup of mine with a 47xxCPU, the 4910MQ might be also able to pull off with a 240W but with over 4GHz things need more juice.

Use the One I linked (That's from woodzstack, He's a reputed vendor here at NBR for the GPUs and His team made that vBIOS for the Ranger and Voyager machines and It has (the one which says 980MOC is what I'm using) that core clock same as Svl7 default one, and It also has the OC option to crank up speeds) if you want or try the Svl7 from T|I. And that Blackscreen does it happen only when you OC or when it reaches certain temperature or the first resulting in the second causing it to choke ?

I hope you know how to Flash on Ranger, the 120Hz machines use different methods, the 60Hz machines do it this way -

Make sure you did safemode --> DDU - Then go to Optimus mode (fn-f5, AW OSD), Device manager --> IF 2 display adapters that means it's in optimus, if you only see 1 and it's not the intel one then you're in dedicated mode.

In that device manager find and disable the 980M and flash the card with vBIOS use the commands from the modme.bat (modme2 is for SLI) from Prema mod v2 Zip - that script doesn't work sometimes so Manual flashing. Then fire up the machine into the PEG mode and Install the INF drivers, Profit

P.S - The reason I'm not using Prema mod v2 should be already in your mind, as most of us have experienced hot temperatures with 1.062v on that vBIOS that's why people are running their cards on 1.000v vBIOSes to make em run cool (Prema's work is fantastic but that extrapolated ASIC based code doesn't hold it's efficiency for most of the chips out there imo)

 

@Prema standard

 

Blackscreen happens only when I OC more than 80mhz using prema v2. I think it's because the too high voltage. I like using prema v2 because it give me 12500 on 3dmark11 without OC. But if I can manage the woodz vbios to go here I'll prefer it for the low temp. And yes I know how to flash vbios

By standard you mean prema v1?

 

@XGIII Current driver 372.54.
This was the highest I could get on svl vbios.
http://www.3dmark.com/fs/9870707
http://www.3dmark.com/3dm11/11513875
Back in March I reached 1402/2905 Mhz stable in 3Dmark, but couldn't achieve that ever again.

 

That's probably because it is the @svl7 / @Johnksss@iBUYPOWER vBIOS...

 

What ?! Dayum.
I didn't know it. Sorry

 

980MOC? I presume you mean OC980M?

I've tried it, but my card crashes with 1.0V after a short while. Bumped the volume to 1.050V and it seemed stable, but didn't have time to test with full GPU load (only ran some game with vsync on). But, if it's the same as SLV7 vBIOS from TI (just with stock clocks), I suppose it won't work stable with my card. The default vBIOS starts throttling due to TDP limit at some point I presume, while Prema's BIOS does indeed overheat my card at full load which leads to throttling again. Perhaps lowering the voltage on Prema's vBIOS might help, but it's locked of course. Damn.

 

Ashtrix
I got good news for you - or bad actually, depends how to look at it. My 980M 8GB G-Sync in P870DM-G does exactly the same.
Here are some details.

When flashed with Prema v2 vBIOS, 980M is working without any restrictions. Full total power blast, 101% performance. When I`m pushing it REALLY hard, it shows peaks above 140 Watts! Each time, when GPU core reach 86C BANG! black screen - each and every time. What`s interesting, GPU clock is at turbo speed all the time, there is no sight of throttling, no PerfCap. Full blast all the time, card is not trying to cool itself, because it`s far from GPU temp limit @91C. It`s like GPU is not aware that some other card element(s) are already in thermal trouble. Game/bench/Windows/whole computer continues to operate normally (sound, disk or network activity etc) but picture is gone. Restart and all is fine.

My theory is that:
A) some other component on the card is getting much hotter than GPU and shuts card down or
B) stock MOSFETs are crap and too weak for Prema vBIOS lol and can`t provide such current (much above 100W) for longer periods of time.

If I lower GPU temp limit to 82C using Nvidia inspector profile I`m seeing what I should - as soon as GPU hits 82C, card starts to throttle, lowering GPU clock, power draw and temps. Set it to run at Windows startup and enjoy Prema vBIOS. No black screen. In coming days I plan to repaste*, so I will look closer to heatsink and thermopads placement to see if something is wrong with them and A) is true. If all is fine, than B) is most likely and some vBIOS restrictions for 980M are needed for normal gaming usage (prolonged GPU load as opposed for benchmarking).

Next I have flashed stock Clevo vBIOS to see what`s going on with it.
Clevo did a good job, as they exactly knew the problem and hid it with vBIOS. Did they deliberately used worst and cheaper components for 980M to maximize profits margin and then dodged it with vBIOS? Tin foil hat theory, I know. But Clevo vBIOS does all it can to keep card from hitting temps above 80C.
Made many test and monitoring, I can see two built in limits, YMMV:
1) power limit: as soon as power draw reaches 105W card starts to throttle, regardless of GPU temp; 68C on GPU and I`m not in turbo clock anymore because of 110W power
2) temp limit: as soon as GPU reaches 76C card starts to throttle, regardless of load or power draw.
So with Clevo stock vBIOS it`s absolutely impossible to hit this "hot spot" 86C issue. Coincidence?

Two options here: run stock Clevo vBIOS and accept lower performance (because of two mentioned limits) or use Prema vBIOS and limit it with NVidia Inspector (lower temp from 91C to 82C).
Third option is still pending (repaste, repad, alter heatsink - will see if something is wrong when I will be repasting*).

* I`m putting my second 980M into P870DM-G for 980M SLI monster again. But before that I will take out currently mounted 980M and put in second 980M alone (as a single GPU setup) and test it to see if it behaves the same as above.
Will report back.

 

High performance thermal pads on the MOSFETs and some liquid metal on the core, svl7 vBios delivering at 1.000 volts @1202 MHz core clock stable ... & increase the voltage as you require for your overclock.

Much lower power draw and temps and significantly better performance when compared to stock.

 

It`s not that easy because there is no "g-sync" version of that vBIOS, not that I`m aware of.

 

Didn't know you required G-Sync version ... But as far as I know, you can for a very small donation ask @Prema to make you an undervolted vBios tailored to your GPU

 

@artpra , thanks for your findings, Yeah the MOSFETs are a bust on the 980Ms, they didn't add the 3VRMs to the GPU PCB, Mr. Fox and Nikos and others have hard modded their cards with extra MOSFETs to OC to 1500MHz stability. For now I'm managing the card with previous paste job and custom cooler with Prema and Svl7 whenever there comes a need to OC I use the Global 980M Prema mod and crank the clocks to match the high voltage default on the Prema Mod rest Svl7-John vBIOS does the work and temp never reach 84C the trigger point of my card until now.

Phobya pads are here but need to do the TIM and Pad job, still pending... & should definitely improve once I get the pads done

 

Ok, new news.

That blackscreen`ing 980M taken out of P870DM-G, second 980M putted in its place. Repasted, thermal pads stayed the same (stock from Clevo factory) as they were attached to heatsink and I didn`t bother. Flashed Prema v2 vBIOS in it and ran exactly the same test on it (same as with first 980M). Guess what, as soon as I seen it hitting 86C KABOOM! black screen again.
Well, that`s interesting. 4 x 980M cards behaving exactly the same (1 yours, my 2 and 1 of my buddy), that`s not what I can call a coincidence!

As my second 980M (then mounted) have higher ASIC score then the first one (unmounted) I wanted to put it in GPU Slave slot, because card it there have worst cooling than in GPU Master slot. So higher ASIC score card in hotter place and lower ASIC score card in cooler place. So removed second 980M and mounted first one in GPU Master again, repasted AND (just because, for sport) I removed all of the factory thermal pads from heatsink. I swapped them with my cheap "no name" blue 3,2WmK ones. As then I didn`t have time to mount the second (higher ASIC) 980M in GPU Slave slot, P870DM-G returned to the exact same configuration as per my previous post, only difference being in new thermal pads.

Few days later, I`m driving PCars@Nordschleife with storm, max opponents number and totally maxed settings - to see how far P870DM-G can be pushed (that`s the most demanding conditions I could find, taking absolutely everything from your GPUs and yes, it can be pushed FAR ). I was on auto fan profile and was watching my GPU temp is rising. After 80C I realized that I`m still on Prema vBIOS - black screen is imminent! At 84C I went from my driving wheel setup to the computer to restart from soon to be black screen...

LO AND BEHOLD!

980M is at (previously critical) 86C and is still operating! Temp is rising again and again, hitting 90C AND STILL NO BLACK SCREEN! At 92C started to drop turbo clocks, as ordered by the vBIOS.
You know what that means... Card is fine and is finally operating as it should be. Stock thermal pads were to blame for its previous blackscreen`ing! Odd, unbelievable, shocking even but that`s the truth. Simply swapping them cured my 980M! My guess is that factory thermal pads were not transfering heat from one of the MOSFETs properly.

Now I`m running SLI again, with the second (higher ASIC) 980M in GPU Slave slot mounted, of course with changed thermal pads and it can break that 86C temp barrier too, no problem. Black screen is nowhere to be seen! Fine and dandy.

Factory thermal pads are trash and it`s where they should be putted, not on the card. That was my cure. My two GPU beast are healthy

@Ashtrix
so yeah, change them or (if you are not using stock ones already) rearrange them better, because from my own experience and experiments I believe that is the reason for yours black screens too. Try it!

 

ahhhh overvolted bios meant for lower quality cards, or chinese imported cards. might as well go BGA if your doing that.

now for me to run off and read some more of this thread..

 

Excellent 980M I got from you , 77% ASIC quality, SVL7 vbios, 1252/2802 @ 1.025v daily gaming= SUPER COOL.
I used to have the black screen issue when I was using Prema's 1.062v vbios. Ditching stock thermal pads helped to some extent, on the high OC range 1320+.
We wouldn't be on this mess if N$hitia/ MSI / clevo or who ever is responsible put 6 MOSFETs instead of 3.

 

Update time (woodz bumped it anyways) - Well I used Phobya XT 7W/mk pads all over the HS and below VRAMs as well and I can see the heatsink to get more hotter now also transferring it tothe bottom side of the GPU MXM area on the mobo, not a biggie. I'm happy with the 980M, Since I can keep it cooler now with the Phobya and winter FTW, Hope it stays like that. Didn't run the machine off the U3+ mod or stress it the same way to check this BSOD hard crash point, Also running Svl7 modded one will use Prema when I need moar perf, And Maybe sometime in future a cool Uv vBIOS would be good. No plans to change the HS as of now, Keeping it for the 1070 Zotac or AMD Vega if they make it to this last true Alienware beauty roar again


Further if I manage to find anymore with anything from sw / vBIOS / hw regarding the VRMs and MOSFETs I'll update, Thanks. Peace !

 

thats pretty crazy....

 

man am i glad i found this thread.... another black screen victim here with a 980M!

so, heres the background info on my case (specs in my sig):

- installed public Prema vbios mod for non-gsync clevo mxm 980M from day one (1.062 V at stock)
- first noticed black screens when running firestrike stress test (stock clocks, stock voltage mind you!). black screens manifest in machine switching off instantly, with one of the two power LEDs blinking green. situation is only recoverable by removing psu and pulling battery.
- switched out stock thermal pads with Phobya Ultra for 3/4 mm thicknesses and Arctic thermal pads for 1/2 mm thicknesses. at first i though this would take care of the issue, but no dice, unfortunately.
- talked to @Prema about this, he provided me with a UV Vbios mod (1.000 V for stock), since my card apparently has quite the insane ASIC value = 78.0 % (higher than 91.9% in GPU-Z database). Flashed it, still no dice *sigh* I rechecked my pad job, seems to be good, although I´ll probably check it again just to be sure...

Things ive noticed / checked:

- Flashing the stock vbios takes care of the black screens, but all of a sudden CPU clocks start dropping randomly all over the place during gaming / benching. this didnt happen before with the vbios mod! (i also have Prema Bios Mod flashed and tweaked it for the CPU to hold its clocks solidly, running 4.5 Ghz everyday btw...) Then i saw @woodzstack ´s post about CPU throttling being an indicator for GPU throttling! and sure enough, whenever the GPU reached 100W+ the CPU started throttling. that only recovered once the GPU fell below 80W for longer periods of time. Still, the black screens seemed to be gone with the stock vbios.
- with the UV Prema Vbios mod, black screens seem to take longer to appear when i set my fans to go full blast (Fn + 1), so I am suspecting that its a MOSFET / VRM acting up (the reason why id like to double and triple check my pad job even though im pretty confident its fine...)
- GPU Core temps while running auto fans were in the high 60s, low 70s running games and benches at 4K resolution

any input on this would be much appreciated! @Prema is also very welcome to chime in on this Would I have to go even lower with the GPU voltage due to this freakishly hihg ASIC value maybe?

 

@jaybee83 , the CPU clock dropping with stock vBIOS part is hard to understand for me since I don't have any experience with Clevos (I've seen the 140W power pull and never seen a CPU throttle), Next that Blackscreen crash of yours is temperature independent ? That's different because 3-4 cases in the thread all point to the " temperature trigger", some have that happen at low to high 70's. I highly suspect & a strong gut feeling it's the VRM or MOSFET quality issues among the 980M chips. Since most of them who posted here all are having the Clevo MXM (2 AW17, one M17x R4, One 870DM - all Clevo MXM cards since AW18 is the only one which shipped with the 980M MXM, limited run), Might also see the same case with the Dell/MSI but no reason not to speculate with the Clevo's MXM, Pure speculation but the Dell / MSI 980Ms are less in number compared to these, so that we can have more concrete opinion on these 980M MXMs.

Also the hardmods done to the cards from what I read, Pre modded they used to crash with Blackscreen due to high Overclock but after the VRM/MOSFET (added an extra one at the top of the card, Also to mention Khenglish - The HW Expert FrankenDoc, said the cards are already robust enough with the technology put into might need to speak to a hardware specialist at this point since the Zotac MXM 1070 also has 3, correction it has 7) mod they don't crash like that anymore. So far only 2 users I've seen running the modded cards - Nikos and Fox (Trivia - Mr. Fox said that these 980Ms are mostly like this due to electronics on the card if you can read his posts), The Prema vBIOS as he always mentioned is based off the numerous tests, I won't doubt him in any case. But the heat might just be adding to the card's circuitry and trip it fast enough that's it due to this faulty electronic part not the GPU core. The low voltage is preferred for most of us while some who get good thermals are happy with either of the vBIOS public mods from Svl7 and Prema and private Uv BIOS as well.