Login
You're viewing the f.duriansoftware.com public feed.
  • Oct 16, 2026, 8:27 AM

    My husband got a new computer game that kept crashing. It turns out that the issue was that it has a system to dynamically drop the resolution and use AI upscaling to compensate when the video card can’t keep up, and this defaults to on. But… if you do in fact have a high-end video card that can keep up, and the AI upscaling never gets called off the bench, the game crashes on a scale of about 15 minutes.

    I suspect it’s just endlessly speculatively creating frames that never get used, and then it doesn’t free them because the freeing happens somewhere down the cold path where it’s called to action.

    Nnnnot a great bug to slip through QA, and surprising because usually the problem is that game studios do too much testing on high-end machines and not enough on mid and low-end ones. It suggests the upscaling system was slapped on last second because a manager heard it was AI.

    💬 3🔄 2⭐ 0

Replies

  • Oct 16, 2026, 8:35 AM

    @0xabad1dea that would be really weird. *usually* these things just have a small single digit array of frame buffers that are written to round robin, are allocated up front, and only relocated if the window is resized. you'd oom a lot faster than 15 minutes if it was allocating entire new full screen buffers every frame

    💬 1🔄 0⭐ 0
  • Oct 16, 2026, 8:37 AM

    @aeva if my guess is vaguely correct, then I assume it’s not happening every frame but when it expects there might be frame drops in the very near future

    of course it might not be correct at all, I’m just working with the strange surface symptoms

    💬 1🔄 1⭐ 0
  • Oct 16, 2026, 8:42 AM

    @0xabad1dea i think your theory about it being a memory leak associated to the upscaler is plausible. it might be leaking a smaller structure continuously

    💬 0🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0