16 Replies Latest reply on Aug 4, 2017 8:28 AM by mirh

    WattMan crashes Windows 7 and 10 desktop into black screen, corrupted screen or THREAD_STUCK_IN_DEVICE_DRIVER blue screen, never crashes in games and Linux.

    max0x7ba

      TLDR:

      • WattMan underclocks/undervolts my 290X too much when idle, it crashes at random. Fixed by installing ClockBlocker which prevents it from being idle.
      • WattMan-controlled fan is stuck at 20% regardless of the GPU temperature. Fixed by installing Afterburner and setting a manual curve fan profile.

      Update:

       

      I had to RMA the card back to ASUS. I conjecture what happened was:

       

      1. When 290X was released it had cooling problems. Which AMD resolved by updating the driver. Their workaround probably was to ignore some parameters from the GPU BIOS and use the driver-provided ones.
      2. ASUS tested the card with a driver that ignored some BIOS parameters and never realized that those ignored BIOS parameters were incorrect.
      3. WattMan was new software that did not integrate those old workarounds for 290X. WattMan uses the parameters from the BIOS. Which results in crashes in idle mode because these BIOS parameters were never actually tested by ASUS. People reported that editing the BIOS and increasing the memory frequency for the idle mode fixes the issue, but it voids the warranty.

      My ASUS ROG Matrix 290x crashes into black screen, corrupted screen or THREAD_STUCK_IN_DEVICE_DRIVER blue screen when in Windows desktop, but never when in games or Linux.

       

      I have tried clean installs of Windows 7 and 10, it did not change things. Windows 10 was installed, updated and the only 2 other programs I installed were Crimson driver Jan 2017 16.2 and Battlefield 1.

       

      I guess the fact that it never crashes in games (neither DirectX 11/12 nor Mantle) when GPU and CPU are maxed out rules out faulty PSU and overheating. I monitor CPU and GPU temperatures, they do not raise above 90°C. The GPU temperature when in Battlefield 4 and 1 is typically 85°C.

       

      It rather crashes when it is lightly loaded or when left idle. It often crashes during installing Jan 2017 16.2 drivers, but I managed to install it after a few tries. I only use Windows for games, and I basically boot into Windows and start a game as fast as possible before the desktop has a chance to crash into the black screen.

       

      It also crashed with series 15 drivers, however with 15.2 it never crashed and I kept using that driver for a few months. I had to update to the latest drivers for Battlefield 1 and the crashes returned.

       

      In the light of the above observations, I believe the root of the problem is the driver. Not the hardware.

       

      My system specs are:

      • Windows 10 Home 64-bit
      • Intel Core i7-3820 Processor @4.3GHz
      • Asus Radeon R9 290X ROG Matrix Platinum
      • Asus P9X79 PRO motherboard with P9X79 PRO BIOS 4801
      • 16GB(4GBx4) G.SKILL TridentX DDR3 PC19200 2400Mz 1.65v Cas10 - F3-2400C10Q-16GTX
      • BenQ XL2730Z 27" FREESYNC 144Hz, 2540x1440, DisplayPort connection, FreeSync is on.
      • Samsung 830 SSD 256GB

       

      Update:

       

      I have installed ClockBlocker that keeps the GPU and memory frequency at its maximum clocks/frequency, rather than downclocked when idle. This prevents my system from crashing.

       

      It looks like a problem with the Crimson drivers. They implement a new aggressive power management scheme (WattMan) that downclocks or undervolts the GPU/memory too much, causing crashes.

       

      A photo of a recent crush with Crimson 17.1.1.

       

      IMG_20170124_202159.jpg

      There are a lot of people suffering from similar issues:

      https://www.reddit.com/r/Amd/comments/5oykt8/bugworkaround_150300_mhz_memory_stuck_200300/

      https://www.reddit.com/r/Amd/comments/44ud42/is_amd_aware_of_the_downclocking_bug/

      https://www.reddit.com/r/Amd/comments/45w5s8/why_does_amd_not_care_about_customers_and_bugs/