• Surface Book 2 disappearing GPU problems

    Home » Forums » Newsletter and Homepage topics » Surface Book 2 disappearing GPU problems

    Author
    Topic
    #205478

    Lots of Surface Book 2 customers complain that their fancy extra-cost NVIDIA GPUs are disappearing — cause unknown. Just like Flickergate, Microsoft’
    [See the full post at: Surface Book 2 disappearing GPU problems]

    1 user thanked author for this post.
    Viewing 9 reply threads
    Author
    Replies
    • #205492

      A bad BIOS update could cause this if the video onboard EFI GOP drivers were not correctly updated.

    • #205495

      I can’t for the life of me understand why anyone would buy a piece of computer hardware from Microsoft anyway, I know I wouldn’t.  They seem to have a very poor overall record with their hardware products in my opinion.

      2 users thanked author for this post.
      • #205526

        Microsoft mouse is a good piece of hardware.

        Group "L" (Linux Mint)
        with Windows 10 running in a remote session on my file server
        • #205530

          Microsoft also made some good keyboards.

    • #205499

      I think it’s probably hardware-related.  I think it’s probably a bad solder joint or failing cap somewhere on the board.  Apple had similar issues with nVidia graphics on some of their Mac laptops, where they had used a smaller capacitor on the power circuit for the iGPU than was reasonable (but it fit better on the tight motherboard, so it was used anyway, according to the video I link at the bottom of this post).  In time, the cap would fail, causing the dGPU to cause bad artifacting or to simply stop working with no display.  That sounds a lot like what’s happening here (and note that this was not nVidia’s fault here– Apple made the choice of which capacitor to use on the motherboard).

      It sounds like the Surface devices in question are using nVidia Optimus (muxless), which is typical for laptops with Intel CPUs that have a discrete GPU.  The Intel iGPU (integrated GPU) is connected as normal to the various video outputs, including the internal port for the laptop display panel (usually eDP, or embedded DisplayPort, in newer models like this) and the various other ports the device may have (external DisplayPort, HDMI, etc.).  Sometimes one of the ports is hardwired directly to the dGPU (discrete GPU, the nVidia), but that’s kind of an edge case.

      The normal means of operation is that the iGPU handles the graphical display by itself, keeping the nVidia dGPU in a low-power state to save energy and extend battery life.  When the graphics load increases, getting near a point that the iGPU cannot handle the load by itself, the system enables the nVidia dGPU and sends the raw data there to be rendered, though the iGPU still remains active and in charge of the display.

      The nVidia card, once it is finished rendering a frame, sends the rendered frame as a bitmap to the iGPU over the PCIE bus.  It’s kind of an ingenious idea, since the back channel from the GPU to the CPU (remember that the iGPU is physically part of the CPU on modern designs, including all of the i3, i5, i7 series) is basically idle normally, so there’s no real downside to allowing the GPU to “hog” the bus like this.  It’s the lanes from the CPU to the GPU that are normally highly utilized during times of high graphical demand, but the ones going the other way are nearly idle.

      The Intel iGPU receives these frames from the dGPU and copies them to the frame buffer for display without having to process them.  Its workload drops significantly, though it is still doing the copying.

      A few systems don’t do it this way.  Some laptops use a mux, or multiplexer, for the various video outputs.  The iGPU and dGPU are both wired directly to the mux, and the mux switches to one or another to enable that unit to control the display.  This approach means that it is possible for the laptop maker to give the user the option to select the iGPU or dGPU in the UEFI settings, and it will have the same effect as plugging a video card into a desktop PC and plugging the monitor cable into that directly, while the iGPU’s port is unused.  The iGPU is still there, but it doesn’t have a connection to the display, while the dGPU has full control of the display with no “middleman.”  This is, of course, the usual way of doing it with a desktop PC.

      I am a little fuzzy about how it works when in Intel iGPU mode on a muxed laptop.  There is a good chance I will be buying one such unit in the next few months, so I will be able to give a more detailed/definitive report then.  My understanding is that when the mux-equipped laptop is set to use the Intel iGPU in UEFI, the standard Optimus mode described above is available, where the Intel iGPU remains in control of the display, but where the nVidia dGPU is there if needed. I don’t know if the mux is visible to the operating system or not, or if it is, whether using that instead of sending rendered frames over the PCIE GPU back-channel is possible.

      I am guessing the Surface in question is muxless (by far the more common setup).

      The reason for this long post (other than the fact that all my posts are long, even when I try to make them short) is to answer the questions people may have about how the device can still be displaying anything if the nVidia is not recognized or functional.  Not sure if anyone wondered that, but if they do, hopefully this will clear it up.

      As for the Surface in question: Wow, MS wanted to be like Apple with the Surface line, and it seems that they really are.  Despite their reputation for having the best hardware designs around, Apple… well, isn’t that great in a lot of ways, hardware-wise.  It looks like Microsoft is following in their footsteps, not only in terms of making hardware that fails before its time, but in terms of refusing to own up and support their customers until they’re forced to do so.  Any hardware maker can make the occasional blunder, but when that happens, they should admit the failure and make it right for their customers.

      Apple apparently can get away with that because they’re the single vendor for any and all MacOS products… if you don’t want to completely abandon the Mac platform, you don’t have any other choice but Apple.  That, along with the cultish devotion some Apple fans have to Apple, means they can treat their customers poorly and still count on having them as repeat buyers, most of the time, which is unfortunate (why should Apple do any better if they don’t have to?).

      Microsoft… is one of many vendors selling Windows 10 laptops and all-in-ones.  If they want to charge premium prices, there needs to be a perception that theirs are premium products, since there are many alternatives.  That perception certainly doesn’t exist in my mind after what we’ve all seen!  If MS admitted it was an issue and offered a recall to fix or replace the units when they were defective, even if they were out of warranty, it would go a long way in fixing this, but as it stands… I wouldn’t buy one (even if I liked Windows 10).

      Dell XPS 13/9310, i5-1135G7/16GB, KDE Neon 6.2
      XPG Xenia 15, i7-9750H/32GB & GTX1660ti, Kubuntu 24.04
      Acer Swift Go 14, i5-1335U/16GB, Kubuntu 24.04 (and Win 11)

      3 users thanked author for this post.
    • #205501

      As a software engineer who writes pixel-wrangling software and does customer support, I can tell you that it’s nothing short of amazing that a dual GPU (Intel and in this case nVidia) system can ever be made to work to begin with. When you stop to think of the complexity that has to be managed to get the output from both GPUs back on the same display, it’s mind-boggling to say the least.

      A second comment, and this is not against Microsoft… If you’ve ever taken a portable device apart, you’ll never have an expectation that it could be a reliable computing platform. Virtually everything is paper-thin, flimsy, and failure-prone in a flexible, temperature-variant, and physcially jarring (portable) environment (not to mention caustic – think about batteries). Not exactly a recipe for high reliability.

      -Noel

      4 users thanked author for this post.
      • #205509

        Noel Carboni: “If you’ve ever taken a portable device apart, you’ll never have an expectation that it could be a reliable computing platform. Virtually everything is paper-thin, flimsy, and failure-prone in a flexible, temperature-variant, and physcially jarring (portable) environment (not to mention caustic – think about batteries). Not exactly a recipe for high reliability.

        For decades now I have only used laptops, always without hardware problems other than some literally sticky key in the keyboard, until I had to retire each of them after no less than six years of service, because of too little RAM memory and hard disk capacity, and eventual unsuitability to do certain things with new software, particularly online, because of evolving CPUs and Internet technologies. So perhaps you are referring to cellphones, tablets and the like? Or to very cheap laptops?

        Ex-Windows user (Win. 98, XP, 7); since mid-2017 using also macOS. Presently on Monterey 12.15 & sometimes running also Linux (Mint).

        MacBook Pro circa mid-2015, 15" display, with 16GB 1600 GHz DDR3 RAM, 1 TB SSD, a Haswell architecture Intel CPU with 4 Cores and 8 Threads model i7-4870HQ @ 2.50GHz.
        Intel Iris Pro GPU with Built-in Bus, VRAM 1.5 GB, Display 2880 x 1800 Retina, 24-Bit color.
        macOS Monterey; browsers: Waterfox "Current", Vivaldi and (now and then) Chrome; security apps. Intego AV

        2 users thanked author for this post.
        • #205612

          No, I’m speaking of laptops. Quite expensive ones. Most recently extracting literally 100 rivets to replace a MacBook Pro keyboard. They’re not made to be serviced. I hate this modern “throw away” philosophy.

          I stand by what I said, though it’s encouraging to hear you have had better experiences that I and my family have had. I admit I’ve had desktop hardware fail too, but not like laptops. That being said, we have an iPad 2 that won’t give up, and even the battery is still good.

          As a workstation class machine user I tend to measure reliability on the capability to run error-free for literally months on end, and be ready for computing chores 24/7 and on a moment’s notice. There is room in the world for desktop computing, but the manufacturers clearly feel glued together, throwaway hardware is where people need to be herded. That’s killing quality.

          -Noel

          2 users thanked author for this post.
          • #205616

            I have an iPad 2 Mini. The battery is still great and it just took on iOS11.4.1. I got it with cellular, and it’s great for traveling – bigger than the cell for maps and lookups.

            2 users thanked author for this post.
      • #205528

        Some laptops have been very serviceable – the IBM Thinkpad 760 comes to mind. You removed the keyboard (very easy to do so) and you found the hard drive and the optical drive. Simplicity.

        Group "L" (Linux Mint)
        with Windows 10 running in a remote session on my file server
        2 users thanked author for this post.
        • #205603

          My Core 2 Duo laptop from 2008 (Asus F8Sn) is very upgradeable/repairable by laptop standards.  I’ve swapped the CPU, the GPU (on a MXM II card, flipped over in the usual Asus manner), added twice the RAM that Intel says will even work in it, swapped the wifi card, and swapped the HDD for a Samsung 850 Evo 1TB SSD.  I’ve also replaced the keyboard and the wifi antennae (when I broke the tiny coax cables they use).  I’ve tried liquid metal TIM to see if it would reduce temps (it didn’t) and swapped the tiny fan for variants with different blade counts to see if they blew more air (they don’t).

          It’s a thick, chunky, heavy unit by modern standards.  Some of that is attributable to the presence of the optical drive.  Undoubtedly, the addon-card style GPU adds some thickness, and certainly the 70W heat sink contributes too (35w TDP CPU and 35 TDP GPU, with a unified heat sink design).  It gets hot when pushed to its limit (running WoW pushes the CPU to the mid-80s on both cores in CPU utilization percentage, while the GPU hits 99% and mostly stays there), but it doesn’t throttle.  It’s about 10 C away from throttling, which is not much thermal headroom, but it’s better than more recent models that throttle pretty easily under load.

           

          Dell XPS 13/9310, i5-1135G7/16GB, KDE Neon 6.2
          XPG Xenia 15, i7-9750H/32GB & GTX1660ti, Kubuntu 24.04
          Acer Swift Go 14, i5-1335U/16GB, Kubuntu 24.04 (and Win 11)

          2 users thanked author for this post.
    • #205508

      This did happen once to me. From what I can tell, the system was just not recognizing that it was there until I undocked and docked it again. I was able to force the machine to shut down with the power key then force it into bios with the volume and power key right after that, save and exit. Then it has been working perfectly since then.

    • #205534

      There’s a rather weird keyboard combination that’s supposed to reset NVIDIA drivers: Shift-Ctrl-WinKey-B. I have to use it after almost every startup because the display usually ignores the NVIDIA settings and reverts to the default settings instead; something to do with Win 10, I believe, because it has been an on-and-off issue ever since the original Win 10 install.

      Edit to remove HTML

    • #205532

      I have a first gen Surface Book and it works perfectly.  On occasion I’ve had it not detect the base, but it’s rare and all it takes is to reattach it and it’s fine.  I suspect this is a bunch of uproar over a fairly isolated issue.

      • #205550

        Could be, but why the cat-and-mouse responses from MS?

    • #205552

      It’s not hardware related. I had the issue for months after the initial release. Hours of time wasted with Microsoft reps till the point it got escalated to a rep giving me their personal number. Downloading a new image from.microsoft.com and reimaginf your machine from a flash drive works. I had constant blue screens before, 3-10 per week. After reimage, 3 months no issues and GPU is always recognized. Also with the new image the Nvidia controls changed.

      • #205608

        If it’s not hardware related, why would it only affect Surface devices?  Optimus is used in a lot of different Windows 10 laptops, but this report is just about Surfaces.  Did MS manage to mess something up so badly in their own preinstalled image that they cause BSODs and disappearing GPUs?  If Microsoft themselves can’t properly install Windows on a Microsoft device… well, I don’t know what to make of that.

        I was thinking this was an older model when I wrote the post above, but I see that this model is a current one.  Surface BOOK 2, not Surface 2.

        Not doubting your report here at all… just trying to think how MS could have messed that up.

        Dell XPS 13/9310, i5-1135G7/16GB, KDE Neon 6.2
        XPG Xenia 15, i7-9750H/32GB & GTX1660ti, Kubuntu 24.04
        Acer Swift Go 14, i5-1335U/16GB, Kubuntu 24.04 (and Win 11)

        1 user thanked author for this post.
    • #205624

      Hi, I had the same problem for about a month, trying to see what kinda weird hardware issue was the cause.

      For me, after starting computer, GPU would basically disappear after 30 seconds or so. Redocking would re-enable GPU, but starting a game, letting it idle, etc no matter on plugged in or power setting it would eventually disappear.

      I had updated the NVidia drivers from the NVidia website a while and after a bit I had a feeling that there was some kinda GPU power time-out glitch going on. So what I did was completely uninstall the NVidia drivers, and reinstalled the base Surface Book 2 driver package MS has on their site, and have no touched any non-MS-SB2 only gpu drivers since. I think main NVidia drivers just isn’t really good with detatchable 2-in-1s like MS surface.

      I have had 0 problems since then, and if anything disconnecting is faster/less glitchy and GPU is always shown when docked. I hope this helps some of you.

    • #207888

      As nice as the Surface devices may be, and I have some the company has bought for me, this is typical Microsoft.

      Sooner or later the support will be lacking unless you can get extended support where tehy have to fix it.

       

    Viewing 9 reply threads
    Reply To: Surface Book 2 disappearing GPU problems

    You can use BBCodes to format your content.
    Your account can't use all available BBCodes, they will be stripped before saving.

    Your information: