Announcement

Collapse
No announcement yet.

A18 Optimization issues persist. 130 FPS indoors to 10 FPS outside

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

    #31
    Originally posted by erlgrey View Post
    tbh. providing a slider for adjusting af is a workaround, not a fix. And just for convenience because you could force 4x af in driver too and have exactly the same effect. And i think, usually you have in games a really low performance difference between 0 and 16x AF, so its not like it is a common issue that you loose 50% by enabling 16x af..

    terrain quality setting is kind of useful, because you can turn down the terrain quality, even switch to legacy shaders for really weak systems, without having to make too many sacrifices regarding e.g. model quality. (tools < full texture quality look extremely ugly)

    what really surprised me, was an insane performance boost if i just switched from dx11 to vulkan in exactly these situations (outdoor, desert or snow + 16x AF), i got similar fps in vulkan with 16x af compared to <4x AF dx11..
    this should not be the case at all, but this also means, there is hope.
    I'm guessing this is addressed to me. I was not talking about the AF option they included in...A18.1 I believe. I'm talking about the new terrain option. You can search faatals post history on the forum and learn what it is and does.

    Comment


      #32
      Originally posted by erlgrey View Post
      tbh. providing a slider for adjusting af is a workaround, not a fix. And just for convenience because you could force 4x af in driver too and have exactly the same effect. And i think, usually you have in games a really low performance difference between 0 and 16x AF, so its not like it is a common issue that you loose 50% by enabling 16x af..

      terrain quality setting is kind of useful, because you can turn down the terrain quality, even switch to legacy shaders for really weak systems, without having to make too many sacrifices regarding e.g. model quality. (tools < full texture quality look extremely ugly)

      what really surprised me, was an insane performance boost if i just switched from dx11 to vulkan in exactly these situations (outdoor, desert or snow + 16x AF), i got similar fps in vulkan with 16x af compared to <4x AF dx11.. (edit: it was actually a +47% gain from dx11 too vulkan with same settings)
      this should not be the case at all, but this also means, there is hope.
      What we need is a way to see how many draw calls are being issued. I wonder if it's a problem of too many blocks being rendered unnecessarily, as that is exactly where Vulkan excels. When there's a large number of objects being drawn, Direct3D 11 and OpenGL crater once you pass 6,000 draw calls. Vulkan, on the other hand, will eat as many draw calls as you can throw at it without any driver-related performance drain. And if the renderer uses multiple draw call submission threads (Vulkan has parallelizable draw call processing that scales almost 1:1 with core count), you can throw more cores and see your performance shine even in ungodly complex scenes.

      The other possibility, is that the Vulkan shaders are far more efficient than the Direct3D 11 ones. I'm guessing that the Direct3D 11 shaders got borked somehow, and either the Vulkan shaders themselves are just better written, or the Vulkan driver is so much more clean that it chews threw the shader that much faster. Could even be both.

      Comment


        #33
        Originally posted by FiftyTifty View Post
        The other possibility, is that the Vulkan shaders are far more efficient than the Direct3D 11 ones. I'm guessing that the Direct3D 11 shaders got borked somehow, and either the Vulkan shaders themselves are just better written, or the Vulkan driver is so much more clean that it chews threw the shader that much faster.
        i guess its just bit borked in dx11 and vulkan fixes something by "accident", not intentional.
        i would somehow understand the performance diff. if i'm running an e.g. 2500k with my vega56, but i'm actually using a 8700k@4.9.
        But yeah, maybe it generates way more drawcalls as i would expect, but don't really know why performance reports from different people vary so extremely. I mean, there are people which claim to get >60fps constant on way weaker hardware and this damn far away from my experience so far.

        Comment


          #34
          Originally posted by SylenThunder View Post
          Having your GPU at 100% is a good thing. You want that.
          Why... why would I want that?
          That's like saying you should drive at maxspeed in your car at all times.

          Comment


            #35
            Originally posted by -Holo- View Post
            Why... why would I want that?
            That's like saying you should drive at maxspeed in your car at all times.
            You want to use 100% of your GPU while gaming. If you aren't at 100% at some point while playing the game then you are either playing a much older game, using too low of graphics settings, or your CPU isn't strong enough. It is completely normal for the GPU usage to bounce around during a game. However your GPU is actually designed to be used at 100%.

            Not having your GPU at 100% is like trying to launch the space shuttle without booster rockets.

            Comment


              #36
              yes, you really can't compare cars to gpus, that's just ridiculous.

              there is only one legit reason why you don't want your gpu at 99%, they currently (amd+nv) seem to have a inputlag penalty compared to having a few % headroom, but in general 99% gpu usage is good because this is more a competitive fps thing.

              Comment

              Working...
              X