Announcement

Collapse
No announcement yet.

AMD RDNA2 Vulkan: RADV vs. RADV+NGGC vs. AMDVLK vs. PRO Driver Benchmarks

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • MadCatX
    replied
    Originally posted by avem View Post

    In other words it's not "community" in any shape or form which means the whole idea of community driven GPU development is more likely a myth.
    Your definition of community is wrong. A FOSS community does not imply a community of unpaid volunteers. The idea is that everybody can chip in. RADV began as a hobby project of David Airlie (at least I think it was David). Zink was a hobby project of Mike Blumenkranz before Valve contracted him to work on it full time. If this is not the community spearheading important and impactful software projects, I don't know what is. Just because these projects later receive commercial backing to ensure further development doesn't change a thing.

    Leave a comment:


  • Venemo
    replied
    Originally posted by avem View Post
    Open Source Community means people coding (not working) in spare time for fun. Paid developers are paid developers. Funny how the terms have slowly been subverted. Subway workers doing kernel development? I'm not sure I get this lame attempt of a joke.
    I started working on Mesa as a hobby, because I was interested and enjoyed working on it. Do you believe I should stop enjoying it, just because I'm paid for it now? Did I stop being a member of the community in your opinion?

    Originally posted by avem View Post
    And not programmers? No way on Earth you're skilled enough to contribute.
    There are many ways to contribute, other than writing code. There are a good bunch of people who help development by doing game benchmarks, and/or create quality bug reports for the developers. This is very valuable, and I'm very grateful to everyone who helps us this way.

    Originally posted by avem View Post
    All I'm trying to say is that open source drivers vs closed source drivers means nada for the average Linux user since only a handful of people can actually fix issues in drivers and they do it only when a particular issue is critical enough. Go check i915/radeon bug trackers - hundreds of open issues and no one paying any attention to them since they are not widespread enough.
    If you're interested to hear about why open drivers are beneficial, here are a few thoughts: https://www.youtube.com/watch?v=QF-xHTKY4H4 which has many good points about how an open driver can be helpful, from the perspective of game developers, HW vendors and platform vendors. If you are a gamer, this benefits you indirectly, even if you don't hack on the driver yourself.

    Leave a comment:


  • darkbasic
    replied
    Originally posted by avem View Post

    In other words it's not "community" in any shape or form which means the whole idea of community driven GPU development is more likely a myth.
    Poor Valve, why wouldn't it qualify as community?
    Do you have something against Valve or your concept of community is flawed?

    Leave a comment:


  • avem
    replied
    Originally posted by bridgman View Post

    I have to disagree with you here. I see a whole class of what you might call "power users" who benefit from having access to source code and commit history, either from being able to bisect to determine exactly where a regression problem on their system occurred, or from being able to work directly with a developer to "try this patch".

    In both cases having open source effectively allows a degree of personalized support that has not been practical with closed-source code, making it possible for developers to quickly identify and fix problems that only appear on specific user systems.
    I cannot disagree with this but in my experience a vanishingly small number of "power users" actually have enough zeal and time to report and bisect major regressions. The Linux kernel architecture is not conducive to bisecting either because you cannot load and unload graphics drivers on the fly. Each bisect, go reboot. That's extremely tiresome.

    And then it's quite horrible that end users actually have to do QA/QC work for companies. It means open source software is not properly developed/funded in the first place. Regressions do happens with proprietary software either but they are a lot less frequent and normally affect a very small number of users.

    Originally posted by bridgman View Post
    If by "lately" you mean "in the last 30 years" then I guess I agree, although not completely even in that case.

    I don't think there have ever been a lot of cases where developers working in their spare time were able to start and maintain large projects - it can be done but it's very rough on the developer(s) - but what independent developers have done and continue to do is get projects through the "proof of concept" stage to the point where it makes sense for one or more companies to have their developers help with some of the heavy lifting.

    It's also worth noting that a lot of the "company developers" working on a project used to be independent developers until they were hired by a company to continue working on the project with more time than they could devote working just as a hobby.
    True so but the point stands: the parts of the open source stack with no major corporation backing are severely lacking.
    Last edited by avem; 18 August 2021, 02:24 AM.

    Leave a comment:


  • nadro
    replied
    Originally posted by nadro View Post
    Hello,
    Today I received my RX 6600 XT and I see some problems with video output during desktop environment start (I have to wait 10+ seconds until my monitor is able to detect video signal). With RX 580 I didn't notice similar issues. I use Ubuntu 21.04 + linux-firmware from git, mesa 21.2 from Kisak PPA + kernel 5.14rc6/5.13.11 from Ubuntu PPA. Someone noticed similar issues?
    Cheers,
    My issue was related to this bug:

    I removed problematic sensor module from /etc/modules and all works fine now.
    BTW. I added this reply to my own question, because it may be helpful for someone in future.

    Leave a comment:


  • Linuxxx
    replied
    Originally posted by omer666 View Post
    I don't know what DE or WM you use, but I get some tearing with Nvidia blob, most notably when playing videos fullscreen.
    On Wayland, I don't get tearing but I can't disable VSync in-game, so that's a problem too.
    I have used a variety of DEs & WMs over the years on nVidia GPUs and actually can't remember the last time I saw any tearing.
    On my HTPC I'm running Xubuntu 20.04 LTS with Xfce 4.14 right now, with its own default compositor active while the option for fullscreen windows to be shown directly also being activated.

    As for fullscreen videos:
    No tearing with a default Firefox setup & naturally no tearing with mpv either.

    And as long as it's not a totally broken implementation, I usually keep Vsync activated in games, since it can actually help with frame-pacing, though of course there are exceptions to this rule.

    About Wayland:
    AFAIK, sticking to X11 for gaming is still the most viable option in 2021, even with Mesa drivers.

    Leave a comment:


  • omer666
    replied
    Originally posted by Linuxxx View Post
    Also, for the ill-informed naysayers without first hand experience for themselves:
    There is no tearing with the binary blob on X11 either, so stop spouting FUD, please!
    I don't know what DE or WM you use, but I get some tearing with Nvidia blob, most notably when playing videos fullscreen.
    On Wayland, I don't get tearing but I can't disable VSync in-game, so that's a problem too.

    Leave a comment:


  • Linuxxx
    replied
    Originally posted by Masush5 View Post

    They need massive improvements in certain cases though. For example, Death Stranding compiles all it's shaders on the first loading screen, which finishes in about 2-3 minutes on RADV and amdgpu-pro, but takes more than 20 minutes on AMDVLK (!).
    Coincidentally I tried out Death Stranding just yesterday as my very first VKD3D-Proton using title on my HTPC (i5-6500 + 1650), and with default settings, I seriously thought for the first few seconds where the landscapes are shown that this had to be a pre-rendered video, because it really did look that unbelievably good!
    The Decima engine out of the Netherlands is truly a technical marvel!

    And the first loading screen really did take only a few minutes here aswell, with the subsequent performance then being flawless ingame.

    It's just too bad that AMD still has nothing to offer that could counter nVidia's 1650 in the 75 Watts GPU space; hopefully at least Intel will fill this void next year...

    Also, for the ill-informed naysayers without first hand experience for themselves:
    There is no tearing with the binary blob on X11 either, so stop spouting FUD, please!

    Leave a comment:


  • bridgman
    replied
    Originally posted by avem View Post
    Amazing how you've skipped the most important part of my comment. My point stands: openness/closeness means nada for the average Linux user as in both cases you're 100% dependent on the will of developers despite many proclaiming that open source graphics drivers are better by definition because they are "open" (source).
    I have to disagree with you here. I see a whole class of what you might call "power users" who benefit from having access to source code and commit history, either from being able to bisect to determine exactly where a regression problem on their system occurred, or from being able to work directly with a developer to "try this patch".

    In both cases having open source effectively allows a degree of personalized support that has not been practical with closed-source code, making it possible for developers to quickly identify and fix problems that only appear on specific user systems.

    Originally posted by avem View Post
    Lately Linux has been driven by companies with a commercial interest not "community". The things developed solely by "community" are in such a dire state it's not even worth talking about. Go check the lm-sensors project. Or how Bluetooth works for people using Linux. No companies backing these features. So much success.
    If by "lately" you mean "in the last 30 years" then I guess I agree, although not completely even in that case.

    I don't think there have ever been a lot of cases where developers working in their spare time were able to start and maintain large projects - it can be done but it's very rough on the developer(s) - but what independent developers have done and continue to do is get projects through the "proof of concept" stage to the point where it makes sense for one or more companies to have their developers help with some of the heavy lifting.

    It's also worth noting that a lot of the "company developers" working on a project used to be independent developers until they were hired by a company to continue working on the project with more time than they could devote working just as a hobby.
    Last edited by bridgman; 17 August 2021, 08:06 PM.

    Leave a comment:


  • agd5f
    replied
    Originally posted by avem View Post

    Lately Linux has been driven by companies with a commercial interest not "community". The things developed solely by "community" are in such a dire state it's not even worth talking about. Go check the lm-sensors project. Or how Bluetooth works for people using Linux. No companies backing these features. So much success.
    The community for a project is made up of the people that are actively working on it. Projects with fewer active participants generally make less progress than projects with more active participants.

    Leave a comment:

Working...
X