The Witcher III Wild Hunt VGA Performance review

Discussion in 'Frontpage news' started by Hilbert Hagedoorn, May 20, 2015.

  1. meth curd

    meth curd Active Member

    Messages:
    94
    Likes Received:
    2
    GPU:
    2080TI FE
    Any chance of seeing some comparison between these results and the post 1.03 (which supposedly brings performance boosts) performance?
     
  2. -Tj-

    -Tj- Ancient Guru

    Messages:
    18,103
    Likes Received:
    2,606
    GPU:
    3080TI iChill Black
    ?

    I dont see it.. Or you mean 770GTX vs 960GTX now with patch 1.3?



    TressFx got better @ AMD and NV later with newer drivers, find something with newer driver not 314.xx.. And my point is mostly how it looks, namely it looks 10x better vs Witcher, at least it has proper 3d hair sprites each individual hair, like with Hairworks @ COD Ghosts for example.
    Not this @ witcher with mostly 2d texture sprites and 2 years newer tech and huge perf. impact. You know TressFx2.0 got better too with even less impact, I tested one tech demo back then when it got released, so idk what's your point here, are you seriously defending this bull @ witcher now?


    btw this is my last response to you, I blocked you for a reason.
     
    Last edited: May 21, 2015
  3. Noisiv

    Noisiv Ancient Guru

    Messages:
    8,230
    Likes Received:
    1,494
    GPU:
    2070 Super
    here ya go:


    Hairworks™ :
    http://www.pcgameshardware.de/The-Witcher-3-PC-237266/Specials/Grafikkarten-Benchmarks-1159196/

    GTX 970 HW off = 100% = 44.7 fps
    GTX 970 HW on = 81.7% = 36.5 fps
    perf. hit: 18.3%

    GTX 770 HW off = 100% = 25.9 fps
    GTX 770 HW on = 82.2% = 21.3 fps
    perf. hit = 17.8%

    R9 285 HW off = 100% = 28.1 fps
    R9 285 HW on = 70% =19.6 fps
    perf. hit = 30%


    TressFX™ :
    http://www.pcgameshardware.de/AMD-R...mb-Raider-PC-Grafikkarten-Benchmarks-1058878/

    GTX 670 TressFX off = 100% = 52.3 fps
    GTX 670 TressFX on = 71% = 37.1 fps
    perf. hit = 29%

    HD 7970 TressFX off = 100% = 72.8 fps
    HD 7970 TressFX on = 72% = 52.3 fps
    perf. hit = 28%

    You can see that hairworks has virtually the same (30%, GCN 1.2) or smaller performance hit (18% Maxwell, Kepler) than TressFX (28%).
    Despite TressFX doing only Lara's hair. And despite no Hairworks optimizations(!!) whatsoever on AMD (according to AMD/CDP)

    Unoptimized, yet faster(or same) on all hardware, except on slow*** tesselation GCN<1.2, and doing much more targets than TressFX.
     
    Last edited: May 21, 2015
  4. Yxskaft

    Yxskaft Maha Guru

    Messages:
    1,495
    Likes Received:
    124
    GPU:
    GTX Titan Sli
    Tomb Raider used the first version of TressFX, AMD claims TressFX 2.0 and 3.0 are much faster
     

  5. Reddoguk

    Reddoguk Ancient Guru

    Messages:
    2,665
    Likes Received:
    597
    GPU:
    RTX3090 GB GamingOC
    This update introduces a range of stability and performances ehancements as well as allows users to take advantage of some additional graphics settings.

    Please find the detailed list of fixes below:

    Improves stability in game-play and the UI
    Improves performance especially in cut-scenes and game-play
    Fixes grass and foliage popping that could occur after density parameters were changed
    Improves Nvidia Hairworks performance
    Boosted texture anisotropy sampling to 16x on Ultra preset
    Sharpen Post-process settings extended from Off/On to Off/Low/High
    Blood particles will now properly appear after killing enemies on the water
    Corrects a bug where player was able to shoot bolts at friendly NPCs
    Improves menu handling
    Corrects an issue with Stamina regeneration while sprinting
    Fixes a cursor lock issue that sometimes occurred when scrolling the map
    Generally improves world map focus
    Improves input responsiveness when using keyboard
    Corrects some missing translations in the UI
    Corrects an issue in dialogue selections
    Rostan Muggs is back
    Minor SFX improvements

    1.03 may skew all of H's hard work and mean he has to do the benches all over again :(

    It would be interesting to see the patches result actually and see if nv HW has less of a hit now.
     
  6. Dorlor

    Dorlor Guest

    Messages:
    1,706
    Likes Received:
    0
    GPU:
    2x evga 780 ti sc acx
    Very interesting - and there is certaintly a pattern here, however, looking at the numbers, the 970 and 780 ti should be MUCH closer than they are in terms of fps. So there must be more to it than just pixelfill rate.
     
  7. SoloCreep

    SoloCreep Guest

    Messages:
    686
    Likes Received:
    13
    GPU:
    RTX 2080 TI
    Holy crap the conspiracies around these forums is getting bad.. Saying NVidia is crippling performance on purpose.. Dumbest **** I have heard in awhile.
     
  8. Undying

    Undying Ancient Guru

    Messages:
    25,478
    Likes Received:
    12,883
    GPU:
    XFX RX6800XT 16GB
  9. Dorlor

    Dorlor Guest

    Messages:
    1,706
    Likes Received:
    0
    GPU:
    2x evga 780 ti sc acx
    As dumb as nvidia lying about gpu specs? That was also called out as bs by the likes of you, and look how that turned out...

    It has already been proven several times that nvidia plays dirty, so i wouldnt be the least bit surprised if that was the case here aswell.
     
  10. Kwee

    Kwee Guest

    Messages:
    2
    Likes Received:
    0
    GPU:
    8GB Gskill 2133 Mhz CL9
    Gpu-z made a error and i copy this mistake without verifying. The real fillrate of GTX 780 Ti is 44 Gpixels/s not 52.
     

  11. Kwee

    Kwee Guest

    Messages:
    2
    Likes Received:
    0
    GPU:
    8GB Gskill 2133 Mhz CL9

    And GTX Titan pixel fillrate is 42 Gpixels/s. There is others little mistake(blame gpu-z) but the order is correct.
     
  12. Lane

    Lane Guest

    Messages:
    6,361
    Likes Received:
    3
    GPU:
    2x HD7970 - EK Waterblock
    Well, YOu should know that TressFX 3.0 is way different now .. and the result have largely changed after some patch current 2013 in TR.

    Basically tressFX have completely move to a different way of compute the hair and fur, Everything have been largely moved to another level.

    And for compare you need to understand that TressFX was really close of how is made hairwork today, but tressFX 3 use a completely different pipeline.

    - no tesselation+GS, instead, extrusion of quad by vertex shader, no MSAA need to be applied ( analytic computing coverage AA is used ), tile deffered rendering method applied, PPL shading etc. At absolutely every level of the pipeline, they have change the method, from the isoline creation to lighting method, shadows, AA method, shading etc.

    AMD presentation of TressFX3.0
    http://amd-dev.wpengine.netdna-cdn....The-Fast-and-The-Furry-Nicolas-Thibieroz.ppsx
     
    Last edited: May 22, 2015
  13. -Tj-

    -Tj- Ancient Guru

    Messages:
    18,103
    Likes Received:
    2,606
    GPU:
    3080TI iChill Black
    @ Kwee

    Idk, gpu-z reads base freq., something like Aida64 takes actual boost frequency into account

    actual boost
    [​IMG]

    base frequency
    [​IMG]


    1215mhz boost
    [​IMG]

    This glitch you talk about has been fixed since ~ 0.60, I remember all the sudden Fermi 570GTX didnt have as high pixel fillrate anymore, was ~ 32ish gpxiel, after fix ~ 22gpixel I think,
    http://www.techpowerup.com/downloads/1921/techpowerup-gpu-z-v0-6-0/


    EDIT: ok found my old OC'ed 570GTX screens in some backup, also tested 780gtx

    before gpixel fix 38gpixel
    [​IMG]

    after gpixel fix 29gpixel
    [​IMG]


    old 0.56 reads half boost 1150mhz freq, that number 2x is 55gpixel , actual boost in aida64
    [​IMG]

    newer 0.64
    [​IMG]
     
    Last edited: May 22, 2015
  14. tsk2k

    tsk2k Guest

    Messages:
    17
    Likes Received:
    0
    GPU:
    Adreno 330
    Excellent review as always Hilbert. Do you have any statement as far as radon vs geforce drivers go in general?
     
  15. Monchis

    Monchis Guest

    Messages:
    1,303
    Likes Received:
    36
    GPU:
    GTX 950
    Frame pacing is just horrible, CDPR fail:


     

  16. ivymike10mt

    ivymike10mt Guest

    Messages:
    226
    Likes Received:
    12
    GPU:
    GTX 1080Ti SLI
    Also GTX 770 perform not good, when compare with GTX 960!
    Many people still have that GPU.
    I still remember b4 W3 comes out. They say that W3 will have around 40-50 FPS on GTX780Ti maxed with all physics on FHD.
    It was also post here on guru 3d in news...

    Anyway Witcher 3 works fine to me.
    I have above 45 FPS maxed out at 4K without AA.
    Properly SLI scaling in day 1 is nice!
     
  17. Medic101

    Medic101 Guest

    Messages:
    3
    Likes Received:
    0
    GPU:
    970M
    Any chance you can do a mobile GPU test? I'd be really interested to see how a the 900M cards and the R9-M cards do in a real world comparison with their desktop siblings.
     

Share This Page