Announcement

Collapse
No announcement yet.

NVIDIA Releases New Vulkan Linux Driver With Better Multi-Threaded Scaling

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

  • NVIDIA Releases New Vulkan Linux Driver With Better Multi-Threaded Scaling

    Phoronix: NVIDIA Releases New Vulkan Linux Driver With Better Multi-Threaded Scaling

    While NVIDIA mainlined their Vulkan driver support in the NVIDIA 364 driver series, they issued another Vulkan-focused driver update yesterday for Linux and Windows for developers and enthusiasts wanting to try out the latest support for this high-performance graphics API...

    http://www.phoronix.com/scan.php?pag...6-Linux-Vulkan

  • #2
    Be careful about upgrading the Windows driver though, I mean their last update was like this, maybe they haven't fixed the oh I don't know, bug that can fry your card... (anything like this ever happen on their Linux driver stack btw? They release broken drivers like this all the time on Windows it seems)
    Last edited by rabcor; 09 April 2016, 01:54 PM.

    Comment


    • #3
      I just run The Talos Principle benchmark on my Optimus laptop:

      20:26:11 INF: - benchmark results 364.15 -
      20:26:11 INF:
      20:26:11 INF: Gfx API: Vulkan
      20:26:11 INF: Duration: 60.0 seconds (3959 frames)
      20:26:11 INF: Average: 66.0 FPS (66.9 w/o extremes)
      20:26:11 INF: Extremes: 166.3 max, 5.4 min
      20:26:11 INF: Sections: AI=8%, physics=1%, sound=1%, scene=72%, shadows=11%, misc=7%
      20:26:11 INF: Highs: 39 in 0.3 seconds (123.2 FPS)
      20:26:11 INF: Lows: 24 in 1.5 seconds (16.4 FPS)
      20:26:11 INF: 30-60 FPS: 21%
      20:26:11 INF: > 60 FPS: 79%
      20:26:11 INF:
      ---------------------------------------------------------------------------------------------------------------------------------
      20:49:59 INF: - benchmark results 364.16 -
      20:49:59 INF:
      20:49:59 INF: Gfx API: Vulkan
      20:49:59 INF: Duration: 59.9 seconds (3890 frames)
      20:49:59 INF: Average: 65.0 FPS (67.4 w/o extremes)
      20:49:59 INF: Extremes: 191.5 max, 6.4 min
      20:49:59 INF: Sections: AI=7%, physics=1%, sound=1%, scene=75%, shadows=9%, misc=6%
      20:49:59 INF: Highs: 55 in 0.4 seconds (144.5 FPS)
      20:49:59 INF: Lows: 55 in 3.4 seconds (16.3 FPS)
      20:49:59 INF: < 20 FPS: 1%
      20:49:59 INF: 30-60 FPS: 19%
      20:49:59 INF: > 60 FPS: 80%
      ---------------------------------------------------------------------------------------------------------------------------------

      20:57:09 INF: - benchmark results 364.16 OpenGL -
      20:57:09 INF:
      20:57:09 INF: Gfx API: OpenGL
      20:57:09 INF: Duration: 58.7 seconds (3187 frames)
      20:57:09 INF: Average: 54.3 FPS (61.2 w/o extremes)
      20:57:09 INF: Extremes: 145.4 max, 5.0 min
      20:57:09 INF: Sections: AI=5%, physics=1%, sound=1%, scene=72%, shadows=16%, misc=5%
      20:57:09 INF: Highs: 18 in 0.1 seconds (125.1 FPS)
      20:57:09 INF: Lows: 315 in 11.9 seconds (26.5 FPS)
      20:57:09 INF: < 20 FPS: 1%
      20:57:09 INF: 20-30 FPS: 3%
      20:57:09 INF: 30-60 FPS: 39%
      20:57:09 INF: > 60 FPS: 57%

      Comment


      • #4
        Originally posted by rabcor View Post
        Be careful about upgrading the Windows driver though, I mean their last update was like this, maybe they haven't fixed the oh I don't know, bug that can fry your card... (anything like this ever happen on their Linux driver stack btw? They release broken drivers like this all the time on Windows it seems)

        Doubt anyone's card was fried from a bug. Not sure that can even happen, even from driver code, in a modern card.

        Comment


        • #5
          Originally posted by computerquip View Post
          Doubt anyone's card was fried from a bug. Not sure that can even happen, even from driver code, in a modern card.
          There was a bug in acpi code few years back that caused HDDs to be aggressively powersaving by default e.g. parking off magnetic needle immediatelly after freeing IO queue.
          That bug actually caused some real world damage to some harddrives and reduced their lifetime etc. I don't remember exact details, but I would suppose that If you tried hard enough, most components could be damaged.

          Even software can cause damage, for example running newer versions of prime95 on overclocked machines or skylake processors is not reccomended since this synthetic benchmark can fry your processors FPU.

          Comment


          • #6
            Originally posted by computerquip View Post
            Doubt anyone's card was fried from a bug.
            Reality says otherwise:

            http://www.tomshardware.com/news/Nvi...ting,9802.html
            http://forum.notebookreview.com/thre...panels.779449/

            Originally posted by tpruzina
            There was a bug in acpi code few years back that caused HDDs to be aggressively powersaving by default e.g. parking off magnetic needle immediatelly after freeing IO queue.
            That was a problem that affected only Western Digital Green drives. It wasn't a bug in ACPI code, but in the drive's own firmware.

            Comment


            • #7
              That doesn't really say anything other than heresay.
              Even in the case you ran your card as hot as possible, it would likely restart itself (at the hardware level) before it got to dangerous levels. There are sometimes bugs within hardware but there's nothing indicating that some bogus driver actually fried the hardware other than someone saying it does. Unfortunately, I don't tend to take things at face value and immediately put on my tin foil hat.

              Comment


              • #8
                Originally posted by rabcor View Post
                Be careful about upgrading the Windows driver though, I mean their last update was like this, maybe they haven't fixed the oh I don't know, bug that can fry your card... (anything like this ever happen on their Linux driver stack btw? They release broken drivers like this all the time on Windows it seems)

                It's rare for major problems to hit a WHQL driver though, since each driver goes through several Beta releases before an official driver is made. Sounds like a last minute fix went in that wasn't properly tested or something. The WHQL drivers tend to be VERY stable.

                Comment


                • #9
                  Originally posted by computerquip View Post
                  Even in the case you ran your card as hot as possible, it would likely restart itself (at the hardware level) before it got to dangerous levels.
                  Well, I agree with you that it sounds strange, but what we think will happen in theory based on previous experiences and what really happens aren't always the same thing. That said, I also never had a video card, or any other piece of equipment, fried from software alone.

                  Originally posted by computerquip View Post
                  Unfortunately, I don't tend to take things at face value and immediately put on my tin foil hat.
                  I can relate to that. I also read a lot about how terrible Seagate HDDs are, but I've been a PC enthusiast since 1994 and only a single Seagate failed on me ever, and it was a very old drive (from the early 90's), which is precisely how many IBM, Fujitsu, Western Digital and Samsung HDDs also failed on me.

                  Comment


                  • #10
                    Originally posted by gamerk2 View Post


                    It's rare for major problems to hit a WHQL driver though, since each driver goes through several Beta releases before an official driver is made. Sounds like a last minute fix went in that wasn't properly tested or something. The WHQL drivers tend to be VERY stable.
                    Ha! Tell that to AMD, they've had many,many whql drivers with serious bugs. You obviously don't remember the Radeon 2000 series launch, those drivers were whql certified too. They never really recovered their drivers after that. Catalyst is today just as bad as it was then and they are still whql certified.

                    Shit, in fact the 64bit builds of windows all require whql certified drivers and the fact is MS still blames stability issues on faulty drivers even though they know they certified them. I don't understand how in the hell anybody could consider that to be a better model.

                    Comment

                    Working...
                    X