Horizontal Flickering Lines present in Full-Screen Mode

Hi everyone, I had to get my M18x R2 up and running again recently and after re-installing Windows along with a re-paste with IC Diamond (secondary was hitting 85C in Unigine Heaven, primary only 75 @ +200 core / +400 memory 1.025V) I was thrilled to discover that the instability that previously existed has disappeared and now it will hold +250 core / +400 memory in all the various benchmarks, Valley, Heaven and Firestrike, even for extended periods (10 minutes or so) and +200 core / +400 memory in games, same voltage (so far only tested with GTA 5).

I am having another issue though, in full-screen mode there are occasional (and sometimes very present, as in the case of GTA 5) flickering horizontal lines mostly around some objects. Playing around trying to eliminate them I've discovered that they disappear if I run the game or application in windowed or full-screen windowed mode.

Unfortunately, running the game in windowed or full-screen windowed mode introduces a bit of stutter, so it's a hard choice between the lines or the stutter.

Other observations, in GTA 5 this phenomenon doesn't occur at night?

I've attempted to google this issue before coming here but haven't found a whole lot other than another observation that running the game or app. in windowed mode eliminates the issue.

I don't believe it's the OC or inadequate voltage because it's still present at default clocks with the additional voltage (1.025V). I don't believe it's heat related, as the avg. load in GTA 5 is usually only around 60% (even with everything on Very High, PCSS shadows on, Tesselation off, FXAA on) and the temperatures only peak in the mid 60's with averages in the low 60's and its far more of a problem than say Unigine Heaven where the temps will get up into the low 70's and the problem is virtually non-existent.

It's not V-Sync related as if I turn v-sync off or to adaptive in full-screen mode the flickering horizontal lines are still present.

I don't believe, but am not certain, that it is an issue with the SLI bridge. If it was the SLI bridge, wouldn't the problem remain irrespective of whether or not the game or application is in full-screen or windowed?

Is it a driver issue? How about vbios? (I'm on 350.12 and using SVL7's 1.025V vbios)

Also, if I run the game or app in single-GPU mode the problem also clears. I am tempted to try to swap the primary card for the secondary to see if the issue is related to the secondary but again, SLI is on and working when the game or app in question is in windowed mode so this leads me to conclude it's not an issue with the secondary. I've also just re-pasted, to include using IC Diamond on all the memory chips, so I'm not exactly thrilled of the idea of taking it all apart and wasting all of that TIM.

Otherwise, it is amazing to see GTA 5 running flawlessly with 680M SLI, even with memory at 4700MB out of a listed 4000MB lol. I'm seeing silky smooth 60FPS everywhere, even with PostFX on Ultra, PCSS Shadows, and shader quality, shadow quality, texture quality etc. (everything in that column) on Very High. I've pop diversity at about 65%, no reflection MSAA and Tesselation off as it's more resource intensive on Kepler and there is next-to-no benefit except the palm trees. I went by the Nvidia optimization guide.

Again, this issue isn't exactly new, but I don't remember it being this bad (over a year of gaming on the ROG Swift with my desktop might be why it's more noticeable).

Anyway, sorry for making this so long. Hopefully I'm not alone with this issue and there is a solution other than running the games in windowed mode with the stutters.

Thanks.

 

So you're telling you built that whole desktop for nothing?!

Oh and change the SLI bridge. Unlike CrossFire, both fullscreen and windowed utilize SLI.

 

lol, basically, going back and looking at my last exchange with Mr. Fox he even suggested reinstalling Windows as it is a potential culprit. At the time I didn't feel like doing that and didn't think it would alleviate the issues with the cards refusing to hold an OC.

Thing is as strong as ever with amazing temps to top it off (75C peak CPU and 70-72C peak GPU at 4.3GHz -1.266v and 1006MHz core / 1000Mhz mem):

http://www.3dmark.com/fs/4708907

 

For those that don't know:

http://forum.notebookreview.com/thr...monster-desktop-i-always-talked-about.752285/

Anyhow, Octiceps, it's good to see youre still stalking NBR, hope youre doing good. Oh and youre not going to believe it but I reinstalled PS2 about a few weeks ago and have been playing it a few nights a week. That was on my desktop, which can run the game no problem on a single overclocked 780 Ti, I am currently downloading PS2 for the M18x R2 but I have a bad feeling a single 680M is going to struggle with it at 1080p and as we both know SLI support has completely evaporated.

The game is STILL fun.

As far as replacing the SLI bridge goes, I apologize for the confusion but what I meant to say in my original post was that the flickering horizontal lines present in certain games and apps (GTA 5 and Unigine Valley) disappears when switching from full-screen to Windowed, either by way of Alt+Enter or through a setting within the game or app and that I don't believe the SLI bridge is the culprit because SLI is working whether it's in full-screen, windowed, or windowed borderless.

If it was the SLI bridge, wouldn't the phenomenon be present ALL the time?

Oh and the phenomenon isn't present it all apps, it is completely absent in Firestrike. I will try some other games and report back but right now this is driving me mad with GTA 5.

 

Not necessarily. It could also be driver issues, but given the number of SLI users not reporting this problem in GTA V, it seems unlikely. Best to eliminate all variables.

 

Yeah, that's a pricey troubleshooting attempt ($30 ebay). I just played about 10 minutes each of both Skyrim and Metal Gear Solid: Ground Zeroes, neither of which exhibited the issue. Also, I even tried turning down the VRAM intensive settings in GTA 5 to well under the 4000MB (2GB actual) limit to see if the issue was related to exceeding available memory and the problem was still just as present and annoying. Looking at the utilization from MGS: Ground Zeroes, it's showing a full 98% memory and 84% peak core utilization. No flickering horizontal lines around objects.

I know Skyrim with an ENB does this, it drove me nuts the last time, and I know GTA 5 and Unigine Heaven does it. Unigine Heaven does it in the beginning of the bench sequence around the ropes and captains wheel as youre boarding the blimp. Sometimes its not present. In fact the phenomenon is intermittent, it isn't always doing it, it comes and goes.

If it was the SLI bridge, wouldn't it be doing it in all games all the time?

BTW, I had to get my R2 up and running again as I recently moved from Nevada to Oregon and will be leading a nomadic existence until I secure a job and a place to live, which could take a few months. And boy am I glad I did, I can't believe how fast this thing is. 680M SLI overclocked is nearly as fast as a single 780 Ti (11.5k GPU vs. 12k Firestrike). I'm running GTA 5 nearly maxed out and am experiencing a silky smooth 60 FPS everywhere.

If money were no issue I would simply replace the pair of 680's with a single GTX 980 as the problem is gone in the games where it's present turning SLI off. Going from 2 to 8GB of VRAM would also be nice.

As a matter of fact, I think once I get things going I will in fact do just that. This thing is still an amazing laptop. Who knows when else it will come in handy down the road.

 

Just to clarify my last post above, the problem was present with Skyrim in conjunction with an ENB and is absent now with Skyrim out-of-the-box with no ENB (seeing as how I deleted the entire game I have to start from the beginning and pick and choose various mods all over again).

 

You mentioned when you repasted you used IC Diamond on the VRAM chips as well? Big no-no. Poor contact and it'll make a mess. You're supposed to use thermal pads. What happened to the original pads?

 

they were in need of replacement, im seeing amazing temps, about 5C lower than last repaste so i imagine the TIM is working better. Ive used TIM on memory chips in the past (5870M).

 

How can you tell? VRAM has no thermal sensor. If you're referring to GPU die itself then yeah lower temp after repasting is typical.

 

Yes youre right, I was under the assumption that lower memory chip temps would some how equate to lower GPU core temps but that is probably not the case.

Is the GPU die height the same as that of the memory chips? Because the memory chips and the GPU all use(d) the same thickness thermal pads. So I'm assuming that if there is enough pressure for the heatsink to make proper contact with the GPU die shouldn't it be the same for the memory chips?

 

No there is a bigger gap between memory chips and heatsink than between GPU die and heatsink. That's why you use thermal pads on memory and thermal paste on GPU.

 

Yes I know, thank-you for pointing this out though as I could be flirting with disaster here. According to web research on the matter, some have replaced the thermal pads on their memory chips along with the thermal pad on the GPU die with TIM and have only seen positive results. Others sternly advise against this. My understanding is that you should only really use a TIM of high-viscosity and you need to be absolutely certain that the height of the memory chips is the same as that of the GPU die. My reasoning is thus, the copper heat-sink that covers both the GPU die and the memory chips is of the same plane and the thermal tape applied to the memory chips and GPU die the same thickness, so I'm under the assumption that it would be safe to apply TIM to the memory chips as well. I've already run it like this for a solid 10-15 hours of 99% load but failure could be long in coming. It is also interesting to note that GTX Titan X (as with GK204) has it's memory chips on the top of the card, with no back-plate, and the only cooling being the memory chips direct exposure to open air:

http://www.legitreviews.com/hands-nvidia-geforce-gtx-titan-x-12gb-video-card_159519

What I MIGHT do (probably will) is remove (again) the heat-sinks to ensure that the TIM is making adequate contact when my replacement SLI bridge and keyboard arrive (I'm currently using the keyboard from my M17x R2, it's actually compatible with M18x R2, whose keyboard suffered the fate of having rubbing alcohol accidentally spilled on it during the last repaste).

I will even order some thermal tape (any recommendations?) to have on hand in case the TIM isn't making proper contact.

Thank-you for the concern though Octiceps, only a friend would point out a potential disaster such as this.

 

You're not supposed to put thermal pad on the GPU die though. Have I not been clear about this? LOL.

Paste for GPU die, pads for everything else (memory chips and vregs). Paste is there to fill in the invisible microfissures on the surfaces of the GPU die and contact plate because air is a poor conductor of heat. Pads do not do this, they're meant for bridging larger gaps such as you'd find between the heatsink and other components on the PCB which paste is not thick/hard enough to fill.

If you're a visual learner, here's what a Titan X looks like with its cooler popped off:

Spoiler

Notice the paste for the GPU die and the pads for the VRAM chips on the top side and the vregs. The VRAM chips on the bottom side of the board are not covered since the assumption with a desktop is that there is adequate airflow.

As for thermal pads, I like the ones from Fujipoly and Phobya. Just make sure you're getting the correct thickness (this is why you don't throw out your old pads) and don't overpay for something with a super-high advertised thermal conductivity as it's snake oil.

 

It's been so long since I initially repasted my GPU's that I can't remember if there were thermal pads or if it was thermal compound on there. Anyhow, I'm going to be ordering some thermal pads, problem is that I don't actually know the thickness, but if they are either 1 or 2 mm I'm certain they are the 1mm variety as they were a bit thinner than the 1.5mm copper shim I used on my desktop cards recently.

I suppose I will just put in Fujipoly pads on ebay and seen what comes back then. Thanks again.

 

Here is what comes on Dell/Alienware heat sinks from the factory, or on new replacement heat sinks purchased from Dell Spare Parts. New NVIDIA 100W heat sinks that work perfectly for 680M/780M/880M/980M with new thermal pads are under $30 (USD) each for the M18xR1/R2 when they are in stock.

 

I'm almost certain Dell put thermal paste, not pad, on the GPU die at the factory

 

Yeah youre probably right, although they did put a big fat thermal pad on the CPU lol.

Anyhow, here's what I'm about to order, I'm fairly certain 1mm is the correct thickness:

http://www.performance-pcs.com/fuji...padding-quarter-sheet-150-x-100-x-1-0-mm.html

 

Dell put a thermal pad on the CPU? Thought they used to be better than that...

 

Lol, yeah, thermal pad with thermal pad temps (100C under load):

http://forum.notebookreview.com/threads/getting-sick-of-haswell-extreme-4940mx.775549/

 

You don't need to spend that much. This is good enough:

http://www.performance-pcs.com/fuji...padding-quarter-sheet-150-x-100-x-1-0-mm.html

 

Not a thermal pad like those on GPU components. They have always used a TIM pad designed for a CPU rather than paste. If it had a regular thermal pad, then somebody assembled it using the wrong product. The photo of the GPU heat sink shows the type of TIM (thermal phase change material) used on both CPU and GPU. It is highly effective for all but significantly overclocked Extreme CPUs, unless or until the machine gets jarred or something else disturbs the bond between the heat sink and die, then it's no longer effective. On normal QM/MQ CPUs and GPU, I have never experienced any thermal problems with the factory TIM. The exception is overclocked Haswell blast furnace CPUs. It does not work well on them at all, but neither do most of the popular thermal pastes.

 

I see, so it's like the same pre-applied compound as you'd find on the water blocks of AIO liquid coolers

 

Yes, exactly. Really good stuff for most applications. For heavy overclocking, not so much.

 

I guess my OCD makes me reapply the thermal paste on any new CPU & GPU as soon as I get it. Even on desktop cards, where the improvement is very slight. Oh well.

 

Mr Fox that thing is useless... Even with my 3630QM which tops out at 3.5GHz, I used to hit 90C easily in games... IC Diamond has been a blessing in comparison although I do need to repaste soon as I might have worn it out

 

OMG guys, I am on cloud 9 right now, I got my replacement SLI bridge in the mail and fully skeptical that it was the problem (I thought secondary was going out /gone out) the flickering horizontal lines have completely disappeared! Thanks Mr. Fox, you were right! I'm about to tear it all the way down and replace the TIM I placed on the memory chips with 1mm Fujipoly thermal tape but couldn't wait to relay the good news. I am totally ecstatic right now.

 

WOOT! Happy to hear it, bro. Don't you just love it when you can fix such a big problem for little more than pocket change? Did you get it directly from Dell for about the same price as I did a while back?

 

Yeah Fox, it is so awesome when you have a serious issue and it's an inexpensive and simple fix. Again, I was totally skeptical that the problem was the SLI cable.

Also, I want to thank Octiceps for adamantly insisting that TIM not be used on the memory chips. Although the TIM did make contact between the heat-sink and the memory chips on the secondary, mainly because I used the dot / blob method here, it was poor contact as the dots of TIM weren't even spread out but out the primary, where I used the spread method, there was zero contact between the memory chips and the heat-sink. Fortunately I've been so busy with the recent move and trying to get situated and finding a job that I haven't spent more than maybe 15-20 hours over the past few weeks actually gaming. Either way, I'm incredibly thankful for the community here.

I also replaced the keyboard, which was a spare out of my M17x R2, with a new M18x keyboard as I accidentally spilled rubbing alcohol on the original.

All in all it was a great maintenance tear down. I am so grateful to have my M18x R2 still going strong, it is really comforting to be able to catch a few hours here and there of my favorite past-time while in transition (again).

When I eventually get things going again I will be replacing the 680's with a single 980 for the next time things get rough again.