P775DM3-G graphic artifacts, stuttering and no GPU compute

Discussion in 'Sager and Clevo' started by Mnddbbl, Jan 9, 2018.

  1. Mnddbbl

    Mnddbbl Newbie

    Reputations:
    0
    Messages:
    6
    Likes Received:
    0
    Trophy Points:
    5
    Hello,

    I've bought the machine (with a i7-6700k and a GTX1080) about one and a half year ago, from the german XMG branch. After ten month or so, I ended up with an issue on boot and video playback that forced me to RMA the damn thing and led, i believe, to a new mobo and GPU.

    We are 4 months later and a similar yet new issue has come : startup is now slow and glitchy (the screen flickers to black), i'm getting small artifacts flashing the screen when doing various tasks (seems somewhat linked to windows with transparency and stuff like that), video playback show green and magenta lines all over (chrome sometimes notifies me that WebGL has crashed or something), the screen often stutter, showing a black screen for a few seconds as if the display driver was updating. And the most problematic issue is that the card doesn't seem to be able to compute anymore. Rendering an image in Blender returns a CUDA error, Photoscan (photogrammetric software) log shows that the GPU processing fails, and OpenCL doesn't seem to work.

    On the other end, I'm getting what seems to be a nominal performance for the card on Unigine Heaven, and I have not yet tested an actual sizable game, but the small title "Besiège" ran without issue. The card seems to be recognized and GPU-Z shows normal stats. Graphic memory reading, however, isn't the same in GPU-Z and the clevo GPU OC thingy (With unigine matching readings from clevo).

    I'm fairly dissappointed as the computer wasn't really stressed during those four months (was mostly used for a small size Revit project, a couple reasonnable renders and barely enought gaming hours to complete AC:Origins storyline. Nothing was ever overclocked or tinkered with) That sound like mundane tasks to me...).

    I'm not sure where the issue is from, and don't understand why the card can run games without trouble but can't even render the default cube in Blender. This makes the machine basically a brick to me...

    Is there anything more I can do to diagnose the problem ? What could be the cause of that kind of issue ?
    Hopefully you'll be able to help, but thank you anyway !
     
  2. Danishblunt

    Danishblunt Notebook Virtuoso

    Reputations:
    492
    Messages:
    3,155
    Likes Received:
    1,288
    Trophy Points:
    231
    Could you make a photo of the artifact you have on the screen? That way we can determine what is causing this issue.
     
  3. Mnddbbl

    Mnddbbl Newbie

    Reputations:
    0
    Messages:
    6
    Likes Received:
    0
    Trophy Points:
    5
    Of course, I just forgot...

    [​IMG]
    This one correspond to web video playback, and does not seem to affect all videos. Maybe it depends on the format.

    [​IMG]
    That one shows the smalls artifacts that affect the whole system. These happen when loading a webpage, opening a window, and basically anything. There is many more than the 3 visible here, but they flash really fast, I could not capture anything better. You can also notice the different memory clock readings.

    I don't have a picture of the third kind (I can't reproduce it right now) but W10 start menu and notification panel sometime appear with a white checker pattern on top of it, and most clickable things won't respond when it happens.

    I also tried to watch a 4K video (.mov) from my SSD, and it does play back with minor stuttering, but it does trigger the small flickering black artifacts.

    Is there any other useful information I can provide ?
     
  4. Chastity

    Chastity Company Representative

    Reputations:
    1,170
    Messages:
    6,163
    Likes Received:
    117
    Trophy Points:
    231
    Boot to memtest and check memory stability. If it's stable, then I fear the GPU needs replacement. Looks to me like VRAM issues, maybe the thermal pads for the VRAM aren't doing their job? I would pop off the heatsink and check, and maybe even replace the thermal pads.
     
    Dr. AMK likes this.
  5. Danishblunt

    Danishblunt Notebook Virtuoso

    Reputations:
    492
    Messages:
    3,155
    Likes Received:
    1,288
    Trophy Points:
    231
    The issues u're seeing are most likely vram related.

    Go to afterburner and downclock your vram to something very low and see if the problem still persists. If it doesn't a simple replacement of the thermal pads will do just fine, otherwise I'm afraid your graphicscards vram are dying on your, if you know a proper repairshop you can get your vram replaced for cheap.

    And the memory are not different readings. One is real clock , the other one which is around 5ghz is the effective clock, your memory is quadpumped hence both are correct. Meaning you can either say your memory is running at real 1.2ghz or effective 5ghz.
     
    Last edited: Jan 9, 2018
  6. Mnddbbl

    Mnddbbl Newbie

    Reputations:
    0
    Messages:
    6
    Likes Received:
    0
    Trophy Points:
    5
    Hey, thanks for the info on vram readings, even if I'm not sure what "quadpumped" means. Is that like some sort of engine with four things running at the same speed but with an offset for each ? Anyway, that was not the point.

    Fairly obviously, simply clocking down to minimum on afterburner (-500 offset) didn't help, and merely introduced a new kind of artifacts in 3D apps (flashing green squares more or less in 3D space, It seems to be expected i gather). I'll try the memtest thing (the issue arose while computing photogrammetry, and that's known to be ram intensive, hopefully it's just that, despite your quite definitive answer), and will have a look at the heatsink, but i'd rather avoir tearing it down myself, that thing look sensitive...

    Now that you mention vram, Cuda-Z reported only 4098mb of it, rather than the 8gig expected. I've ignored that as the software seems shady, but could it mean that one vram chip got fried and the other is fine ? I guess that could explain the inapropriate behavior in some softwares, and proper behavior in other, depending on the way it uses it ?

    And, as a mean of learning from this, what could cause that kind of issue ? Overheat ? The laptop had proper temps in most graphic intensive apps I used, but I actually never checked or benchmarked a full load on both the CPU and the GPU (Photoscan does load both, and reported CPU performance quite lower than expected, indicating thermal throttling I guess - about 10 times less computation than the GPU, when a typical raytracing CUDA application should yield about a three fold perfomance on the GPU at best, from my limited experience of it). If this is the cause, it actually means that servicing the machine won't fix it for long. Is that shared heatsink an actual design flaw than can't cool both components a the same time ?

    Anyway, it doesn't sound like good news for me, but thanks a lot for the insights !
     
  7. Danishblunt

    Danishblunt Notebook Virtuoso

    Reputations:
    492
    Messages:
    3,155
    Likes Received:
    1,288
    Trophy Points:
    231
    The thing is you assume the temps were fine but in reality all you knew was that the GPU temps were fine, the vRAM could still be overheating. How it happened, we don't know, at this point it's speculation. Also your card has not 1 vram chips, it has 16 of them, each containing 512MB vRAM. Your theory about the heatsink design could be correct, since the CPU is the biggest problem in those notebooks, it's entirely possible that your CPU has killed the vrams, but could also have died on their own for other reasons.

    Just RMA it, get it fixed and then enjoy or sell it and get another one, since apparently you already had the same issue 2 times in a row and understadibly might be scared and frustrated with your notebook.
     
  8. Mnddbbl

    Mnddbbl Newbie

    Reputations:
    0
    Messages:
    6
    Likes Received:
    0
    Trophy Points:
    5
    Yep, that sounds like the only reasonnable advice. I'll try to have XMG explain to me what happened, and will consider a CPU downgrade as well.

    Thanks for the help.
     
  9. Danishblunt

    Danishblunt Notebook Virtuoso

    Reputations:
    492
    Messages:
    3,155
    Likes Received:
    1,288
    Trophy Points:
    231
    You can undervolt your 6700K and keep it stock to keep the temps low. It's a waste since the CPU is capable of much more, but at least it would run cool. Better than playing wannabe desktop and burn your entire system down tho. Maybe you can switch up the 6700K to a 6700 non K, as far as I know they run much cooler.
     
  10. Meaker@Sager

    Meaker@Sager Company Representative

    Reputations:
    7,335
    Messages:
    46,106
    Likes Received:
    12,174
    Trophy Points:
    931
    DDR = Double data rate
    Quad pumped = Quadruple data rate.

    In DDR 2 bits are transfered every clock cycle, in quad pumped or QDR it's 4 bits.

    The clock is basically a square wave:

    [​IMG]

    (From wikipedia)

    Usually in DDR a bit is sent on both the top edge and bottom edge of the clock. In QDR it's sent twice each top and bottom edge.

    Thus the data clock is given as 4x the quoted speed.
     
Loading...

Share This Page