14 Replies Latest reply on Feb 17, 2018 5:21 PM by hreich

    R9 380 - Screen goes black when GPU under stress - 17.7.2

    origbhod

      Hello

       

      When I am playing a game or the GPU is under a decent amount of stress (ran a GPU stress test program) after a very short duration the screen will go black for a moment. After which the screen returns however, whatever it was that I was playing has crashed. So far I have attempted to measure the temperature when it is under stress but it seems that there is no real correlation as it will occur sometimes when the GPU is at 65 degrees and sometimes when its at 75 degrees. Though it should be noted that it did reach 80 degrees at one stage before it had the black screen.

       

      I also tried to monitor the CPU, the temperature does not seem to be high enough to cause any issue, also the CPU usage doesn't seem to go high either when it occurs. However, the disk usage does quite often go to 100% for when the black screen occurs though it is quite hard to be concise due to the fact that the computer will go unresponsive for a while and be extremely slow even after the black screen.

       

      As I was writing this it happened again, this time while watching videos on youtube, not even full screen and what I was watching is not what I would call visually demanding. I have checked my WhoCrashed reports and found this:

      crash dump file: C:\WINDOWS\Minidump\072817-29359-01.dmp

      This was probably caused by the following module: atikmpag.sys (atikmpag+0xF7F8)

      Bugcheck code: 0x116 (0xFFFFBB0651BFF4A0, 0xFFFFF805CDC3F7F8, 0x0, 0xD)

      Error: VIDEO_TDR_ERROR

      file path: C:\WINDOWS\System32\DriverStore\FileRepository\c0316467.inf_amd64_3d8fbd78102e53d7\atikmpag.sys product: AMD driver company: Advanced Micro Devices, Inc. description: AMD multi-vendor Miniport Driver Bug check description: This indicates that an attempt to reset the display driver and recover from a timeout failed. A third party driver was identified as the probable root cause of this system error. It is suggested you look for an update for the following driver: atikmpag.sys (AMD multi-vendor Miniport Driver, Advanced Micro Devices, Inc.).

      Google query: Advanced Micro Devices, Inc. VIDEO_TDR_ERROR

       

      Did give google a go at what is suggested though it seems to be more aimed toward if there is an error that appears in the bottom right over the taskbar. If you think it may be worthwhile I will do what most seem to suggest if you receive that error.

       

      Any help is appreciated

      Many Thanks

       

      Specs:

      Windows 10 64bit

      Motherboard: ASUS ROG STRIX B350-F GAMING

      Bios Version: 0223 x64

      RAM: Corsair Vengeance LPX 2x4GB DDR4

      Processor: AMD Ryzen 5 1600x

      CPU Fan Hyper TX3 Evo

      GPU: AMD Sapphire R9 380 4gb

      PSU: Corsair 500w

      1TB Hard drive

       

      Radeon Software Version - 17.7.2

      Unfortunately not too sure how to check monitor in detail.

       

      Message was edited by: Matt B We have updated the title of this discussion with relevant details to better describe your issue.

        • Re: Screen goes black when GPU under stress
          kingfish

          Do you have the Power Limit set to +50 (all the way to the right)?

          Screenshot_1 (16).png

          • Re: Screen goes black when GPU under stress
            origbhod

            Updated the BIOS to the newest version, hasn't changed anything. Seems that being in full screen makes the problem occur more often.

            • Re: Screen goes black when GPU under stress
              origbhod

              Well now it seems the issue has gone further, now I can't even go into the games as they are blocked from accessing graphics hardware. Also happened when I attempted to go into wattman in Radeon Settings. I have tried to do a fresh install of the drivers using DDU to remove them and this has done nothing. Tried to make the TdrDelay 8 seconds and this has changed nothing as well.

               

              Also attempted a few other supposed fixes after searching up 'application blocked from accessing graphics hardware' yet so far nothing has worked.

               

              Is there a direct form of AMD support or an email that I can get in contact with?

               

              Anyway any help is appreciated,

              Many Thanks.

                  • Re: Screen goes black when GPU under stress
                    origbhod

                    The AMD issue reporting form says that it is not meant to gain technical support and refers me back to the community forum. So what can I actually do, do I just report the details and hope they address the issue in their next driver update? If that doesn't work what else can I do besides buy a new graphics card? Would this likely not fall under warranty and therefore be a valid way to get a refund?

                      • Re: Screen goes black when GPU under stress
                        spikeypup

                        TLDR:  Same issue here, TDR's, preceded or coincided with sound buffer stuttering, reset of relive wattman settings typically, have reinstalled 4 times, 3 times clean, post updated to latest WU WHQL pull, then clean then express, all variations.  Have checked all other dependencies involved for AMF and DX, seem ok, have DISM and SFC to be sure, also Chkdsk'd, cleared shader cache, and reset all games in question and/or validated files.  Cannot stream with OBS without immediate failure within 4-5 minutes of streaming, however preview is stable fine, after activating stream/record, crash will ensue within 1-5 minutes or as soon as 30 seconds.  Issues started at 17.7.1

                        Included full AIDA64 report with complete system info but there is a TLDR version at the end of this post as well.

                         

                        Unfortunately I am sad to report that I am having the same issue as well as of 17.7.1, short TDR's, some cause a full crash to desktop with wattman reset and Radeon settings failure to load afterwards (via context menu and tray unless the process is terminated manually and then boo-time CNEXT executable ran to restore access to context RS and Tray RS app).  Usually it is accompanied by the sound buffer issue as you described where the sound is rapidly stuttering until the screen blanks and is about the return, other times the sound will start with that and the video is fine, but then I know it is about to crash within a few seconds.

                         

                        The problem is really killing it for me because it has pretty much caused a complete inability for me to operate my Livestream whatsoever, it stays stable for about a 2-5 minutes, then the VCE starts to choke, then eventually starves, and causes a full crash of the game being played, then an abnormal unhandled error in OBS which will result in a need to manually terminate OBS (19.0.3 with AMF Encoder Plugin Version 2.2.0 for compatibility with 1.4.4 AMF changes within 17.7.2) even after an exit from the app, as the process stays hung even though it appears as successfully closed to the user, more than likely because it is still waiting to send or verify receipt of frames from the encoder engine.

                         

                        During Forza Horizon 3, I'm having random crashes, within a few moments to within 30 minutes, performance is excellent and monitoring the GPU shows no abnormal temps, throttling, etc.  I have noticed that sometimes after a wattman failure, and BTW I recommend that anytime your settings get reset in wattman, especially if you notice that they are only partially reset (i.e. your power settings still show 50% but the clocks have been reset) that you definitely restore them to default using the little arrow-ma-jig (super technical term) in the upper right, usually if I don't do this, then wattman doesn't seem to hold the changes after changing clocks etc. and hitting apply.  In fact the clocks will go allover the place rapidly sometimes afterwards when testing in benchmarks if this is the case.

                         

                        I checked to see if it was the Enhanced Sync/Turbo Sync (which is working REALLY well, so pleased with it, other issues aside), but no, whether VSYNC ON via app or Forced in Relive Drivers or no VSYNC In App or Forced Off, or Enhanced On, crashes still occurring.  Tested with all defaults set, still problematic, tested with no Overclocks and max power, same problem, tested all with Relive Recording off as I always have the global set to off anyway since I use OBS, but when local recording I do turn it on because it performs transparently and delivers excellent quality at lower bitrates and honestly in general they don't really (OBS/Relive)interfere with each other as long as they are not both recording, in my case anyway.

                         

                        The crashes happen to me during Forza H3, Asphalt 8, Sniper Fury, Overwatch, Diablo 3 and several others.  Oddly GTA V, Doom (Vulkan), and a couple others seem unaffected, so I can't really correlate.  I checked for any particular ISR/DPC issues but haven't really setup an intense profiling and diagnostic and monitoring group yet to narrow down the problem. I'll also attach a debugger to see if the games can spit out any useful info as they crash, but not so sure yet.  I will try to do more diagnostics and update any findings if worthwhile.

                         

                         

                        Operating System  Microsoft Windows 10 Pro 10.0.15063.483 (Win10 RS2 BUILD 1703)
                        Version  AIDA64 v5.92.4312 Beta

                         

                         

                        Computer:
                        Computer Type  ACPI x64-based PC
                        Misc Device  Alienware Master I/O Comm. Board [MIO HID-VDD ON [INTEL ICH-3A34]
                        Operating System  Microsoft Windows 10 Pro
                        DirectX  DirectX 12.0
                        Motherboard:
                        CPU Type  QuadCore Intel Core i7-960, 4200 MHz (21 x 200)
                        Motherboard Name  Alienware Aurora
                        Motherboard Chipset  Intel Tylersburg X58, Intel Nehalem
                        System Memory  24566 MB (DDR3 SDRAM)
                        DIMM1: Kingston HyperX KHX1600C9D3/4GX  4 GB DDR3-1333 DDR3 SDRAM (9-9-9-24 @ 666 MHz) (8-8-8-22 @ 609 MHz) (7-7-7-20 @ 533 MHz) (6-6-6-17 @ 457 MHz)
                        DIMM2: Kingston HyperX KHX1600C9D3/4GX  4 GB DDR3-1333 DDR3 SDRAM (9-9-9-24 @ 666 MHz) (8-8-8-22 @ 609 MHz) (7-7-7-20 @ 533 MHz) (6-6-6-17 @ 457 MHz)
                        DIMM3: Kingston HyperX KHX1600C9D3/4GX  4 GB DDR3-1333 DDR3 SDRAM (9-9-9-24 @ 666 MHz) (8-8-8-22 @ 609 MHz) (7-7-7-20 @ 533 MHz) (6-6-6-17 @ 457 MHz)
                        DIMM4: Kingston HyperX KHX1600C9D3/4GX  4 GB DDR3-1333 DDR3 SDRAM (9-9-9-24 @ 666 MHz) (8-8-8-22 @ 609 MHz) (7-7-7-20 @ 533 MHz) (6-6-6-17 @ 457 MHz)
                        DIMM5: Kingston HyperX KHX1600C9D3/4GX  4 GB DDR3-1333 DDR3 SDRAM (9-9-9-24 @ 666 MHz) (8-8-8-22 @ 609 MHz) (7-7-7-20 @ 533 MHz) (6-6-6-17 @ 457 MHz)
                        DIMM6: Kingston HyperX KHX1600C9D3/4GX  4 GB DDR3-1333 DDR3 SDRAM (9-9-9-24 @ 666 MHz) (8-8-8-22 @ 609 MHz) (7-7-7-20 @ 533 MHz) (6-6-6-17 @ 457 MHz)
                        BIOS Type  AMI (11/10/10)
                        Display:
                        Video Adapter  Radeon (TM) RX 480 Graphics (8 GB) (17.7.2 [27thJuly Release, however I DID Install the BAD 26th when it was released]
                        3D Accelerator  AMD Radeon RX 480 (Polaris 10)
                        Monitor  Asus MX279 [27" AH-IPS LCD] (F5LMRS034590) VIA HDMI
                        Monitor  Benq GL2760H [27" LED LCD] VIA HDMI
                        Multimedia:
                        Audio Adapter  ATI Radeon HDMI @ AMD Ellesmere - High Definition Audio Controller
                        Audio Adapter  Creative SB X-Fi Audio Controller
                        Storage:
                        IDE Controller  Intel(R) ICH10 Family 6 Port SATA AHCI Controller - 3A22
                        Disk Drive  Samsung SSD 850 EVO 1TB (1000 GB, SATA-III)
                        Disk Drive  ST31000524AS (1 TB, 7200 RPM, SATA-III)
                        Disk Drive  WDC WD10EVDS-63U8B0 (1 TB, SATA-II)
                        Disk Drive  WDC WD10EZEX-00BN5A0 (1 TB, 7200 RPM, SATA-III)
                        Optical Drive  HL-DT-ST DVDRWBD CH20N
                        Optical Drive  PLDS DVD+-RW DH-24ABS
                        SMART Hard Disks Status  OK
                        1 of 1 people found this helpful
                  • Re: Screen goes black when GPU under stress
                    danieldeguesm├╝ller

                    I have suffered with this problem too, but it occurs with more violence and more frequency if you watch any strem video, on youtube and more unusual and rarely occurs, but other sites have a more frequent occurrence, and in a few times the driver can recover from the crash, but in the vast majority the screen ends up generating the minidump and causing a blue screen on windows, the next startup gets very slow.

                     

                    Even with the 17.11.4 driver, I still have issues with that.

                    • Re: R9 380 - Screen goes black when GPU under stress - 17.7.2
                      julip

                      Why has this not been addressed professionally yet? For how many posts I've seen about this issue this seems to be pretty wide spread among the 4 series.

                       

                      I get the black screen at least a few times a day and from what I've seen there has been no resolution on this forum.

                        • Re: R9 380 - Screen goes black when GPU under stress - 17.7.2
                          pokester

                          Yah this is a frustrating issue. I went through thinking it was overheating, overloading, power supply issue or bottleneck issue before realizing nope it's just an issue with the new power saving feature AMD put in the drivers. Setting the power to 50+ I think helps but alone won't do much.

                           

                          This is a speed/power phasing issue. Do to new features introduced in 17.7.2 and cemented in the 17.7.3 drivers.  The drivers since march by default vary your speed based on load to save power and keep cool the gpu. However on many cards the issue seems worse. In Wattman change to manual settings and look to the last state. This is the true clock of your card, (mine is 1405, use the correct max speed for your card) now set all the editable states to the same top speed, so it's not variable anymore. Now do the same with the power, look for the highest voltage and change all the states to that number. Yes you will burn more energy, but also hopefully be stable.

                           

                          I also disable overlay in Steam as this seems to help. I am theorizing that the overlay at time may be causing the drivers to adjust speed hence adding the instability.

                           

                          Many have found the newest Adrenaline drivers 17.12.1 & 17.12.2 have fixed this.  However they also seem to introduce various other bugs, so YMMV. But they are the only ones that have worked without setting all the states the same since driver version 17.7.1. Definitely uninstall and run DDU from safe mode before installing the 17.12.2 drivers.

                           

                          If none of this helps try regressing to 17.7.1 they don't have the new features that started this mess. However it just won't have fixes for games or other issues since March.

                        • Re: Screen goes black when GPU under stress
                          hreich

                          I disabled sleep otpions in WIN10 and since then i am not having problem with black screen