Announcement

Collapse
No announcement yet.

Intel's Mesa On-Disk Shader Cache Maturing, Radeon Devs Not Yet Convinced

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

  • dungeon
    replied
    Originally posted by Kano View Post
    I don't know why you always mention the Doom Vulkan benchmark article here.
    Because you asked "% of what?" and i explained before of what that is There you can see, how usually people present perf boost with percentages and that is mine vocabular too

    This game does not even run with Wine die to the copy protection used!
    And where i said that you should run that game under wine?

    But as you like to discuss it:
    Nope that is not my point and i don't want to disscuss it, link is just for your desire as i know you have strong nVidia driven fanboy libido Due to that, you can't even focus on subject here and neither there - so instead you focus on trollism
    Last edited by dungeon; 15 July 2016, 03:29 PM.

    Leave a comment:


  • Kano
    replied
    I don't know why you always mention the Doom Vulkan benchmark article here. This game does not even run with Wine die to the copy protection used! But as you like to discuss it: if you don't see a (huge) diff between OpenGL and Vulkan the hardware was already maxed out. This happens more often with Nvidia and a fast CPU. The Doom Vulkan port seems to be able to use more threads and therefore slower CPUs get a much huger gain. AMD's OpenGL code is/was much more Single Core speed limited that's why Mantle, Vulkan, DX12 helps even with relatively fast CPUs. Simple logic to define GPU vs CPU limited benchmarks. This of course also changes with the used res, the higher the more likely you hit GPU limits.

    Leave a comment:


  • dungeon
    replied
    Who cares about benchmarks Kano here Before you mentioned it, i did not mentioned benchmarks at all - go back and find one . Forget about benchmarks in this thread, BREXIT I talked how it goes with amdgpu-pro and i guess you can test it with your nVidia card too... disable game MT and inject __GL_THREADED_OPTIMIZATIONS=1 and compare diff for yourself, maybe you will have some diff maybe not... just like in Doom 4 article about new vulkan patch, where nvidia get less and amd much more, where even cards of the same vendor diff by their % boost, so there is no proof there that what % i get you will get exactly same % too, you might not see anything, etc... so try it for yourself

    People care about or want to have shader cache driver feature with open drivers, because it provides smoother gameplay... intel implemented seems something for intel (how good that cache is i dunno) but that is how they can test it for real with Talos by disabling internal one too (as regardless of having disk cache, by default game will still compiling them during gameplay and will do it again on next run so once you go back to play a game...) For radeonsi user should not disable this, Marek said now up to 4 cores compile shaders, which is cool and faster/better then before but still not ideal if you value smoothness
    Last edited by dungeon; 15 July 2016, 12:00 PM.

    Leave a comment:


  • Kano
    replied
    This can not be seen during benchmarks as you run the same scene after the 2nd run. But you said that MT would be better to be turned off and provided no results. Only relative performance figues of an UNDEFINED AMDGPU-PRO gpu without mentioning OpenGL or Vulkan and without any other system settings!!! That's pure bullshit!

    Leave a comment:


  • dungeon
    replied
    And i will not pay more then one minute to serve wishes of known nVidia driven trolls

    Long story short for this thread and about Talos... Talos compile shaders during startup, but also during gameplay too and it do it every time you run the game and during gameplay... so if you driver have good disk shader cache, disable this and enjoy .

    ogl_bDisableShaderCaching = 1
    Last edited by dungeon; 15 July 2016, 10:19 AM.

    Leave a comment:


  • Kano
    replied
    I want to see YOUR results. Doom does not matter at all because wine can not run it because of Denuvo. You always throw in numbers and you can not show what you benchmarked! Talos is supported by pts, you could create a new (local) profile from the existing one and add your settings switchable as option the test-definition.xml. Show a an updated test-profile and YOUR benchmarks. But it seems you prefer to discuss theoretical options. I can not test AMDGPU-PRO, nobody sent me a card and i will not pay 1 cent to buy one.

    Leave a comment:


  • dungeon
    replied
    Originally posted by Kano View Post
    Usually all shaders are compiled at startup, but that takes only a few mb space to cache em, it really doesn't matter if you do that twice.
    Show me how much people has Xeon CPUs and if number is enough big then i would agree that is not much of a issue

    But here we talking of conflict of features, which can easely be avoided as there is an option in case you have driver feature which serve it better.

    Where are your benchmarks? % of what?
    Perf diff my Kano nVidia's funboy Here is the benchmark result for you, Doom 4 vulkan and RX 480 on the GTX 980 level

    http://www.eurogamer.net/articles/di...formance-gains
    Last edited by dungeon; 14 July 2016, 06:09 PM.

    Leave a comment:


  • Kano
    replied
    Usually all shaders are compiled at startup, but that takes only a few mb space to cache em, it really doesn't matter if you do that twice. Where are your benchmarks? % of what?

    Leave a comment:


  • dungeon
    replied
    Yes, it is optimal - it will compile it once (as it must do that) then cache it to the disk once in the lifetime and you will never ever see it again... after that all is butter smooth

    So if you are amdgpu-pro user like you isn't , do this an enjoy

    ogl_bDisableShaderCaching = 1
    ren_bMultiThreadedRendering = 0
    Originally posted by Kano View Post
    No idea why MT should be bad, did you do any Talos benchmarks?
    Because driver has threading profiles for both those games with amdgpu-pro and work faster with it of course, peak at plus 40% or 25% on average Something like 10 or 15% is diff (game MT vs profile MT) if i remember right and if you have enabled game MT together with profile MT it is sort of 5% slower, etc...

    So the best is to disable both if you use amdgpu-pro.
    Last edited by dungeon; 14 July 2016, 03:15 PM.

    Leave a comment:


  • Kano
    replied
    Well your logic is not optimal, it would still need to compile it on the fly when it is used first, then it will stutter. No idea why MT should be bad, did you do any Talos benchmarks?

    Leave a comment:

Working...
X