Linux Vulkan Dawn of War 3 alt tab crash system

Hi,

I looked into it and found an issue with the game. I contacted Feral and they will be releasing a fix soon.

Which drivers have you tried? And which desktop environment?

That is great! thanks

Same here even with the newest Vulkan developer drivers, switching away from Mad Max Vulkan beta lags the whole X session like crazy and produces the same Xid exceptions:

[    8.847858] NVRM: loading NVIDIA UNIX x86_64 Kernel Module  396.24.02  Thu May 24 03:48:07 PDT 2018 (using threaded interrupts)
[    9.650624] nvidia-modeset: Loading NVIDIA Kernel Mode Setting Driver for UNIX platforms  396.24.02  Thu May 24 03:44:56 PDT 2018
(...)
[ 2657.684107] NVRM: GPU at PCI:0000:01:00: GPU-5069632b-ce62-bb04-ec5b-8835feddef42
[ 2657.684111] NVRM: Xid (PCI:0000:01:00): 13, Graphics Exception on GPC 0: WIDTH CT Violation. Coordinates: (0x6f0, 0x0)
[ 2657.684118] NVRM: Xid (PCI:0000:01:00): 13, Graphics Exception: ESR 0x500420=0x80000010 0x500434=0x6f0 0x500438=0x1200 0x50043c=0x0
[ 2657.684131] NVRM: Xid (PCI:0000:01:00): 13, Graphics Exception on GPC 1: WIDTH CT Violation. Coordinates: (0x6c0, 0x0)
[ 2657.684136] NVRM: Xid (PCI:0000:01:00): 13, Graphics Exception: ESR 0x508420=0x80000010 0x508434=0x6c0 0x508438=0x1200 0x50843c=0x0
[ 2657.684149] NVRM: Xid (PCI:0000:01:00): 13, Graphics Exception on GPC 2: WIDTH CT Violation. Coordinates: (0x6d0, 0x0)
[ 2657.684155] NVRM: Xid (PCI:0000:01:00): 13, Graphics Exception: ESR 0x510420=0x80000010 0x510434=0x6d0 0x510438=0x1200 0x51043c=0x0
[ 2657.684167] NVRM: Xid (PCI:0000:01:00): 13, Graphics Exception on GPC 3: WIDTH CT Violation. Coordinates: (0x6e0, 0x0)
[ 2657.684173] NVRM: Xid (PCI:0000:01:00): 13, Graphics Exception: ESR 0x518420=0x80000010 0x518434=0x6e0 0x518438=0x1200 0x51843c=0x0
[ 2657.684183] NVRM: Xid (PCI:0000:01:00): 13, Graphics Exception: ChID 0027, Class 0000b197, Offset 000015cc, Data 00010000

This is with {ForceFullCompositionPipeline=On}, someone mentioned that above so probably relevant. WM is marco.

I am not able to reproduce that Mad Max crash on 396.24.02. I tried a variety of WMs including marco with ForceFullCompositionPipeline. What GPU are you using? Does it repro 100% of the time for you?

If I start it with “Run in a window” selected and then Alt-Enter to make it full screen, everything works fine, I can switch desktops and programs, no issue.

But if I run it full screen (which is the default and always resets in the Feral launcher), then switching away gets the short Xid spam, 100% reproducible.

Card is a GTX 980. I’ll attach nvidia-bug-report (disregard all the lib/module version mismatches in the logs, that’s just spam after installing different driver and before reboot :))

Edit: and apparently it doesn’t happen in Rise of the Tomb Raider which I just got today.
nvidia-bug-report.log.gz (148 KB)

Maybe depending on nvidia-drm.modeset=1?

Still no luck. :(

Rise of the Tomb Raider is NOT affected by the same problem as Mad Max Vulkan Beta and others.

396.24.02 (the Vulkan Driver) worked fine for RotTR and it still works fine with the new 396.45, which apparently includes 396.24.02’s fixes.

So, if you had problems on 396.24 (without the .02) - you’re safe to go for 396.45.

However, for Mad Max, 396.45 still doesn’t fix the Xid’s, now they say:

[11126.256121] NVRM: Xid (PCI:0000:01:00): 13, Graphics Exception on GPC 0: WIDTH CT Violation. Coordinates: (0x6f0, 0x0)
[11126.256128] NVRM: Xid (PCI:0000:01:00): 13, Graphics Exception: ESR 0x500420=0x80000010 0x500434=0x6f0 0x500438=0x1200 0x50043c=0x0
[11126.256141] NVRM: Xid (PCI:0000:01:00): 13, Graphics Exception on GPC 1: WIDTH CT Violation. Coordinates: (0x6c0, 0x0)
[11126.256146] NVRM: Xid (PCI:0000:01:00): 13, Graphics Exception: ESR 0x508420=0x80000010 0x508434=0x6c0 0x508438=0x1200 0x50843c=0x0
[11126.256158] NVRM: Xid (PCI:0000:01:00): 13, Graphics Exception on GPC 2: WIDTH CT Violation. Coordinates: (0x6d0, 0x0)
[11126.256164] NVRM: Xid (PCI:0000:01:00): 13, Graphics Exception: ESR 0x510420=0x80000010 0x510434=0x6d0 0x510438=0x1200 0x51043c=0x0
[11126.256176] NVRM: Xid (PCI:0000:01:00): 13, Graphics Exception on GPC 3: WIDTH CT Violation. Coordinates: (0x6e0, 0x0)
[11126.256182] NVRM: Xid (PCI:0000:01:00): 13, Graphics Exception: ESR 0x518420=0x80000010 0x518434=0x6e0 0x518438=0x1200 0x51843c=0x0
[11126.256193] NVRM: Xid (PCI:0000:01:00): 13, Graphics Exception: ChID 004f, Class 0000b197, Offset 000015cc, Data 00010000

Rise of Tomb Raider sometimes crash is because there is a fix in 396.24.10 with fullscreen flipping on “Fixed an intermittent hang of Vulkan applications running fullscreen when flipping is allowed on Linux”
396.45 include the previous fixes of vulkan.

Dawn of War has a bug that is confirmed by Nvidia and Feral should release a fix.

Rise of Tomb Raider I also experienced some random crashes even with the latest 396.45 The Rise of Tomb Raider launcher sends a bug report several times.

Nvidia has full support for Linux since TNT cards. AMD is the worst support for Linux. The AMD driver installer only works for certain version of Xorg. Some of the AMD cards that are not that old were previously supported on Ubuntu 16 are not supported anymore in Ubuntu 18. AMD Linux has the worst performance and rendering bugs than Nvidia. All the Linux users has been buying Nvidia cards which is the case for many years. Opensource does not equate to better driver since the community does not design or develop the hardware. Only the creator knows best.

Whilst I’ll agree that the ‘radeon’ and ‘fglrx’ generations of AMD/ATI gpus had shocking support and crappy drivers, the new generation with open-source drivers does not suffer from the driver-fix fragmentation I’ve experienced with NVIDIA.

There definitely was a time I would have agreed with you, but just two bugs that I’m following, since they affect me:

  • incomplete Vulkan support which causes game crashes (this one)
  • display does not blank if using Display Port (quite a big deal: this crappy driver has the ability to significantly shorten my monitor’s life-span – and even induce burn-in, which I was very lucky to recover from because the monitor had only been stuck on the image for a few hours overnight)

These, imo, are great examples of reasons why, next time I buy a card, it’s unlikely to be NVIDIA.

We all know the history of the former bug (fixed on arbitrary, non-merged branches and left unavailable to the consumer!); and the latter has been open for OVER TWO YEARS, whilst no-one gives a crap about the many users who have added a “me too”. And for every user who reports the issue, there are users who can’t be bothered and just assume that “linux is crap” – and there are those who haven’t encountered the issue simply because they use HDMI – but I had to buy an extra hdmi cable because I couldn’t use the display-port cable that came with my Dell monitor – and the writing has been on the wall for ages for HDMI, so I can’t blame Dell, or any other monitor manufacturer, for giving me a superior cable.

On top of which, performance is a constant debate with NVIDIA vs AMD, with neither winning hands-down and the AMD offerings often being better on the pocket. Where AMD shines is that at least, with the ENTIRE DRIVER being open-source, people can patch the driver if there is an issue, and contribute upstream. Neither vendor is perfect, but I’d much rather be in the house where anyone can have a crack at fixing the problem – so someone is likely to, instead of problems just being left to rot because the platform isn’t considered important enough (or so it seems). The windows driver is regularly updated and kept in good form. The linux driver, not so much, at least from the perspective of someone who consumes both.

@thedf413

Do you still get a repro if you disable “Allow Flipping” in nvidia-settings (X Screen -> OpenGL Settings)?

Note: you need to reapply the setting each reboot, unless you have a script that runs nvidia-settings -d each time you start X.

@davydm The point of possibility of bug fixed only if several people reports exact problem and is reproduceable by developers, otherwise is not fix. people don’t have the problems you are having. Monitor does not blank is not necessarily a driver problem. Have you actually isolated and pinpoint is a driver problem? And monitor today do not suffer from burn-ins, that is in the past long ago.

AMD Linux is not a good performer. There is no debate even needed, Linux games on Steam are recommended system with Nvidia. You would say the same negative things to AMD if AMD had bugs.

@thedf413 the vulkan alt-tab is already fix. If you still have problem with the game, then is a game bug. Dawn of War 3 for example has already been confirmed is a game bug, and that might be the case too for the other games you mentioned

@howdev:

  1. DPMS bug report: https://devtalk.nvidia.com/default/topic/968059/linux/dpms-not-working-with-displayport-monitor (open since 2016; some users report that it’s fixed on their setup with a driver update (so, driver bug!) others still have the issue). HDMI works fine – I don’t know how a user could possibly have configured the system to have DPMS only for HDMI

  2. Monitors do still suffer from burn-in, though not as bad as in the past (read: it takes much longer with greater differentials). An overnight stint with the screen on, on a page which was largely dark, with a white patch in the middle, left a discernable ghost image on my Dell 27" when I came in in the morning and started using the machine. I could undo the damage with a quick color-flipping program, mostly because it hadn’t been that way for long. If that was a machine in an office, left over a weekend, I don’t think I would have been so fortunate.

  3. Yes, NVIDIA is a good performer on Linux but what good is performance when it brings your whole system down?. I’d rather have something reliable than have to figure out system lock-ups every half an hour. I was fortunate enough to have only X lockups (could switch to vt & restart xdm), but they would happen within 10-40 minutes on ROTTR, and would happen immediately if the game had any graphics interruption – alt-tab, change volume (and get OSD). ANYTHING. Unless you’re willing to drop top dollar on a 1080ti, performance of cards in the mid-range between AMD and NVIDIA is comparable: https://www.phoronix.com/scan.php?page=article&item=2018-start-mesanv&num=1

But hey, what do I know? I’m only on the discussion for bug (1), have personally experienced (2) (and you can google about it), and have managed to use minimal google-fu to get to an article pitting AMD vs NVIDIA, on Linux, for a bunch of games.

@davydm I am using Display port and I don’t have DPMS problem. I am using the same driver too. I think is to do with the OS power management

To be fair, I haven’t re-tested DP vs HDMI, but I’d like to know how OS power management would be different between the two. The problem was that the correct command is sent to blank the monitor (so it does), but a second later, it comes back online. This was a known problem with NVIDIA drivers for many (but not all) Linux users. It seems to affect certain cards. Just because it’s working for you, it doesn’t necessarily mean it’s working for everyone.

seems like those more recent people in the DPMS bug thread uses kernel 4.16. try 4.15

unfortunately the 396.45 that supposedly fixed all previous vulkan bugs crashes many more games. Deus Ex Mankind not vulkan but now crashes not long after loading the game when accessing the menu or change resolution.
Still have to revert back to 396.24 which I had backed up