580 release feedback & discussion

Another quick update on the new 580.95.05 and the Xid 13 errors in Elite Dangerous—perhaps rather interesting, too.

TL;DR: Besides resolution/supersampling, the issue seems affected by refresh rate / fps. Limiting frame rate with DXVK_FRAME_RATE to 60 (from 144) seems to eliminate the Xid 13 errors.

Let me emphasize right away that these are very preliminary results, mostly from AFK testing with Elite staying in the main menu. I have never experienced Xid 13 error if the game made it through the first couple of minutes, but IIRC other people reported encountering them well into a gaming session.

Discovering the possible frame rate involvement was by accident. The 3070 I mentioned is connected to a 60 Hz monitor and thus has a DXVK_FRAME_RATE=59.999 setting on most games. Unexpectedly, it happily ran 2.0x SS at 1080p for more than an hour, which made me wonder why would the 3080/12 behave differently. The two cards are, in fact, in a single, multi-seat, system, so the difference—besides the fps cap, that is—is mostly in how logind sets the user sessions.

In any case, when I set DXVK_FRAME_RATE=59.999 also on the 3080/12 and the 2070 Max-Q, they, too, ceased to produce the Xid 13 errors, even at 2.0x SS (@ 1080p).

But again: I haven’t had the time for repeated tests. Those errors have always tended to be somewhat non-deterministic on the outside, so it’s possible that these initial tests were pure luck. Then again, too, I do see some consistency—just like I saw the opposite one without a 60 fps cap on the frame rate, when both the 3080/12 and, especially, the 2070 repeatedly froze fairly quickly.

If frame rate does turn out to be a contributing factor, it also perhaps makes sense. From my basic understanding, Xid 13 is too generic of an error to draw specific conclusions, but I wouldn’t be surprised if some timing/racing issues might be involved.

In any case, those playing Elite might want to check if capping the frame rate to a lower value helps them to sidestep the problem while a true solution is—hopefully—in the works.