Asynchronous Compute

Discussion in 'Videocards - NVIDIA GeForce Drivers Section' started by Carfax, Feb 25, 2016.

  1. Redemption80

    Redemption80 Ancient Guru

    Messages:
    18,495
    Likes Received:
    266
    GPU:
    GALAX 970/ASUS 970
    Yeah, let's convince 980ti owners their cards aren't fast enough so we can get them cheap from them :)

    Hard to tell if Nvidia's DX12 is bad yet, it looks fine when you look at the 980ti, AMD is doing well with the horrible ports though, but these terrible ports can't last forever?
     
  2. Yeah, I understand that the developer has to add built-in support for UWP. I wonder if they will eventually add support for .dll injectors so, anyone that wants to use a third-party overlay, SweetFX, Nvidia Profile Inspector edits, etc (of course the tools will have to support DirectX 12 as well). I am glad to hear they are working on fixing the vsync problem and it will be out this May. Seems now that there is more compute heavy lighting in games that the next series of cards from Nvidia and AMD will focus more on compute performance especially with them wanting to provide real-time light ray tracing within a decade or so.
     
  3. Undying

    Undying Ancient Guru

    Messages:
    19,458
    Likes Received:
    7,795
    GPU:
    RTX 2080S AMP
    What next, Deus EX? AMD sponsored game.

    Star Citizen/DX12 update, again AMD sponsored.

    Warhammer Total War could be interesting. It will not be CPU heavy like before/less overhead again giving AMD the advantage or nvidia having none like before.

    What do you think it will happen? ;)
     
    Last edited: Apr 8, 2016
  4. Redemption80

    Redemption80 Ancient Guru

    Messages:
    18,495
    Likes Received:
    266
    GPU:
    GALAX 970/ASUS 970
    Probably the same situation with the Fury X and 980ti being around the same performance, hopefully a fallback to DX11 for those who want a less buggy version or are running SLI /CF.
     

  5. Kaarme

    Kaarme Ancient Guru

    Messages:
    2,978
    Likes Received:
    1,692
    GPU:
    Sapphire 390
    Games will only continue to be broken if there's a "fallback to dx11". It means they are dx11 games with halfassed dx12 icing causing more trouble than worth. As it is modern AAA titles are so complicated and expensive that the studios barely want to invest the effort to make them playable at all at launch. The last thing we need is the studios dividing their meager efforts between dx11 and dx12 graphics engines for the same game. Dx12 will get better in time if it's given the chance. Transition phase games will pay the price, but it can't be helped.
     
  6. Redemption80

    Redemption80 Ancient Guru

    Messages:
    18,495
    Likes Received:
    266
    GPU:
    GALAX 970/ASUS 970
    The best examples of DX12 have a DX11 version to fallback, it's the DX12 only games that appear to be broken.

    I understand and agree with the split resources reasoning, but not atat the expense of potentially endless poor ports that would have been better in a DX11 game.
     
  7. dr_rus

    dr_rus Ancient Guru

    Messages:
    3,328
    Likes Received:
    595
    GPU:
    RTX 3080
    It's C. - game's engine is just plain bad.

    No, next up is Doom, Homefront: The Revolution, Mirror's Edge Catalyst, No Man's Sky and only then Deus Ex. Doom may have Vulkan renderer but the rest will be OGL/DX11.

    What will happen though is Pascal's release in June.
     
    Last edited: Apr 8, 2016
  8. Carfax

    Carfax Ancient Guru

    Messages:
    2,913
    Likes Received:
    465
    GPU:
    NVidia Titan Xp
    At this moment, Quantum Break seems like a straight port from Xbox One with very little PC specific optimizations.

    That's why AMD has such a huge lead, since the Xbox One uses GCN architecture.

    To be fair though, NVidia does have it's work cut out when it comes to DX12 drivers. The good news is, that the drivers have been improving by leaps and bounds with every release.

    The 364.72 drivers are a very noteworthy improvement over the previous drivers in Rise of the Tomb Raider for instance.
     
  9. Carfax

    Carfax Ancient Guru

    Messages:
    2,913
    Likes Received:
    465
    GPU:
    NVidia Titan Xp
    Whether a game is AMD or NVidia sponsored is irrelevant. It's the capabilities of the engine, and the optimizations the developers put in place that really matter..

    Quantum Break was basically an almost 1:1 port from the console, so it has very little optimization for non GCN hardware.

    Rise of the Tomb Raider is a much better showing of more balanced DX12 optimizations, and Fable Legends would be as well if it were still in development.

    Doom beta also appears to run very well on NVidia hardware, despite what the naysayers stated after the alpha benchmarks:

    Alpha:

    [​IMG]

    Beta:

    [​IMG]
     
  10. Ieldra

    Ieldra Banned

    Messages:
    3,490
    Likes Received:
    0
    GPU:
    GTX 980Ti G1 1500/8000
    [​IMG]

    Doom beta runs perfectly fine on NV hw, developer claims significant gains were made async shaders (on gcn hw) in this title

    AOTS:

    5820K @ 4.5
    32GB DDR4 2933
    980Ti

    1440p

    Some benchmark results with and without core and memory overclocks

    Highest memory controller load (reported by GPU-Z) was in CRAZY preset run with memory @ 1490/7000 (73%)

    DX12 - HIGH PRESET 980Ti @ 1210/7000

    == Configuration =========================================================
    API: DirectX 12
    ==========================================================================
    Quality Preset: High
    ==========================================================================

    Resolution: 2560x1440
    Fullscreen: True
    Bloom Quality: High
    PointLight Quality: High
    Glare Quality: Low
    Shading Samples: 8 million
    Terrain Shading Samples: 8 million
    Shadow Quality: Mid
    Temporal AA Duration: 0
    Temporal AA Time Slice: 0
    Multisample Anti-Aliasing: 2x
    Texture Rank : 1


    == Total Avg Results =================================================
    Total Time: 60.005051 ms per frame
    Avg Framerate: 61.550373 FPS (16.246855 ms)
    Weighted Framerate: 60.946045 FPS (16.407955 ms)
    CPU frame rate (estimated if not GPU bound): 142.365112 FPS (7.024193 ms)
    Percent GPU Bound: 99.982994 %
    Driver throughput (Batches per ms): 7991.057129 Batches
    Average Batches per frame: 14088.831055 Batches
    ==========================================================================
    DX12 - HIGH PRESET 980Ti @ 1490/7000

    == Configuration =========================================================
    API: DirectX 12
    ==========================================================================
    Quality Preset: High
    ==========================================================================

    Resolution: 2560x1440
    Fullscreen: True
    Bloom Quality: High
    PointLight Quality: High
    Glare Quality: Low
    Shading Samples: 8 million
    Terrain Shading Samples: 8 million
    Shadow Quality: Mid
    Temporal AA Duration: 0
    Temporal AA Time Slice: 0
    Multisample Anti-Aliasing: 2x
    Texture Rank : 1


    == Total Avg Results =================================================
    Total Time: 60.001251 ms per frame
    Avg Framerate: 71.991623 FPS (13.890505 ms)
    Weighted Framerate: 71.337791 FPS (14.017815 ms)
    CPU frame rate (estimated if not GPU bound): 145.114700 FPS (6.891101 ms)
    Percent GPU Bound: 99.961853 %
    Driver throughput (Batches per ms): 8080.241699 Batches
    Average Batches per frame: 14101.337891 Batches
    ==========================================================================
    DX12 - HIGH PRESET 980Ti @ 1490/8000
    == Configuration =========================================================
    API: DirectX 12
    ==========================================================================
    Quality Preset: High
    ==========================================================================

    Resolution: 2560x1440
    Fullscreen: True
    Bloom Quality: High
    PointLight Quality: High
    Glare Quality: Low
    Shading Samples: 8 million
    Terrain Shading Samples: 8 million
    Shadow Quality: Mid
    Temporal AA Duration: 0
    Temporal AA Time Slice: 0
    Multisample Anti-Aliasing: 2x
    Texture Rank : 1


    == Total Avg Results =================================================
    Total Time: 59.999844 ms per frame
    Avg Framerate: 74.839638 FPS (13.361903 ms)
    Weighted Framerate: 73.669044 FPS (13.574223 ms)
    CPU frame rate (estimated if not GPU bound): 136.416306 FPS (7.330503 ms)
    Percent GPU Bound: 95.914856 %
    Driver throughput (Batches per ms): 7934.309570 Batches
    Average Batches per frame: 14281.052734 Batches
    ==========================================================================
    DX12 - CRAZY PRESET - 980Ti @ 1210/7000
    == Configuration =========================================================
    API: DirectX 12
    ==========================================================================
    Quality Preset: Crazy
    ==========================================================================

    Resolution: 2560x1440
    Fullscreen: True
    Bloom Quality: High
    PointLight Quality: High
    Glare Quality: High
    Shading Samples: 16 million
    Terrain Shading Samples: 16 million
    Shadow Quality: High
    Temporal AA Duration: 0
    Temporal AA Time Slice: 0
    Multisample Anti-Aliasing: 4x
    Texture Rank : 1


    == Total Avg Results =================================================
    Total Time: 60.002960 ms per frame
    Avg Framerate: 37.833069 FPS (26.431904 ms)
    Weighted Framerate: 37.195988 FPS (26.884621 ms)
    CPU frame rate (estimated if not GPU bound): 123.674149 FPS (8.085764 ms)
    Percent GPU Bound: 100.000000 %
    Driver throughput (Batches per ms): 7912.587402 Batches
    Average Batches per frame: 17555.273438 Batches
    ==========================================================================
    DX12 - CRAZY PRESET - 980Ti @ 1490/7000
    == Configuration =========================================================
    API: DirectX 12
    ==========================================================================
    Quality Preset: Crazy
    ==========================================================================

    Resolution: 2560x1440
    Fullscreen: True
    Bloom Quality: High
    PointLight Quality: High
    Glare Quality: High
    Shading Samples: 16 million
    Terrain Shading Samples: 16 million
    Shadow Quality: High
    Temporal AA Duration: 0
    Temporal AA Time Slice: 0
    Multisample Anti-Aliasing: 4x
    Texture Rank : 1


    == Total Avg Results =================================================
    Total Time: 60.002106 ms per frame
    Avg Framerate: 43.745224 FPS (22.859638 ms)
    Weighted Framerate: 43.088013 FPS (23.208311 ms)
    CPU frame rate (estimated if not GPU bound): 129.148865 FPS (7.743002 ms)
    Percent GPU Bound: 100.000000 %
    Driver throughput (Batches per ms): 8550.075195 Batches
    Average Batches per frame: 17746.050781 Batches
    ==========================================================================
    DX12 - CRAZY PRESET - 980Ti @ 1490/8000
    == Configuration =========================================================
    API: DirectX 12
    ==========================================================================
    Quality Preset: Crazy
    ==========================================================================

    Resolution: 2560x1440
    Fullscreen: True
    Bloom Quality: High
    PointLight Quality: High
    Glare Quality: High
    Shading Samples: 16 million
    Terrain Shading Samples: 16 million
    Shadow Quality: High
    Temporal AA Duration: 0
    Temporal AA Time Slice: 0
    Multisample Anti-Aliasing: 4x
    Texture Rank : 1


    == Total Avg Results =================================================
    Total Time: 60.002808 ms per frame
    Avg Framerate: 45.875206 FPS (21.798267 ms)
    Weighted Framerate: 45.184284 FPS (22.131588 ms)
    CPU frame rate (estimated if not GPU bound): 132.483719 FPS (7.548098 ms)
    Percent GPU Bound: 99.924889 %
    Driver throughput (Batches per ms): 8930.305664 Batches
    Average Batches per frame: 17745.410156 Batches
    ==========================================================================

    [H] results for comparison[​IMG]

    Guru3D
    [​IMG]
     
    Last edited: Apr 9, 2016

  11. niczerus

    niczerus Master Guru

    Messages:
    290
    Likes Received:
    3
    GPU:
    MSI GamingX 580 4GB
    Quantum break runs very very bad on nvidia hardware , its for async ?
     
  12. dr_rus

    dr_rus Ancient Guru

    Messages:
    3,328
    Likes Received:
    595
    GPU:
    RTX 3080
    Nobody knows but highly doubtful. Async can result in +5-10% of performance.
     
  13. Keesberenburg

    Keesberenburg Master Guru

    Messages:
    857
    Likes Received:
    27
    GPU:
    EVGA GTX 980 TI sc
    Last edited: Apr 12, 2016
  14. fantaskarsef

    fantaskarsef Ancient Guru

    Messages:
    13,295
    Likes Received:
    6,046
    GPU:
    2080Ti @h2o
  15. Barry J

    Barry J Ancient Guru

    Messages:
    2,796
    Likes Received:
    137
    GPU:
    RTX2080 TRIO Super
    DX12 is as far as I am concerned in beta maybe even alpha Devs need to up there game they seem to be coasting producing console ports and doing the minimum they can get away with. Until DX12 is truly a viable option we need DX11 version to compare there DX12 work I figure nvidia DX11 against DX12 is a good way to compare the api as AMD DX11 needs work.

    DX12 only game has nothing to compare it against = Devs get away with sloppy work
     

  16. Spets

    Spets Ancient Guru

    Messages:
    3,257
    Likes Received:
    394
    GPU:
    RTX 3090
    Surprised this thread is still going after that post and the interview with IO Interactive:
    With all the extra work needed for negative/no/small increase in performance, it doesn't seem worth it for devs. Considering amd also claims Hitman to feature best implementation of DX12 async compute just goes to show how out of proportion everyone is blowing async compute.
     
  17. Ieldra

    Ieldra Banned

    Messages:
    3,490
    Likes Received:
    0
    GPU:
    GTX 980Ti G1 1500/8000
    I can't measure framerate in QB as shadowplay fps monitor isn't working, but it runs fine no crashes so far.

    Game a blurry-edged mess and the story is cringe-worthy stupid

    As for async being hard to tune, I'm not even bothering trying to convince people of that :p I'm just sticking to showing it provides no tangible advantage over competing nvidia hw
     
  18. Darren Hodgson

    Darren Hodgson Ancient Guru

    Messages:
    16,114
    Likes Received:
    609
    GPU:
    Palit RTX 3080
    Why? I'm curious to know why (use spoiler tags where appropriate if you are going to reply).

    The only stupid bit I came across was where one of the characters in the TV kills someone in a hospital in full view of the people there, including (what looks like) security staff!!! Apart from that, I found the story to be fine and the writing/acting is a cut above what we normally get in these kinds of games (the ones that don't have the words "Grand", "Theft" and "Auto" in them, anyway).
     
  19. lantian

    lantian Master Guru

    Messages:
    292
    Likes Received:
    1
    GPU:
    MSI GTX 970 GAMING 4gb
    Have to agree, the game has it's problems but in my opinion story isn't one of them, it had one or two strange moments, but other than that is pretty enjoyable
     
  20. Ieldra

    Ieldra Banned

    Messages:
    3,490
    Likes Received:
    0
    GPU:
    GTX 980Ti G1 1500/8000
    I haven't played much to be honest, so I'll correct my previous statement

    The dialogue ive heard is cringe-worthy, the only saving grace (in my opinion) for the introductory sequence is the guy who enters the corridor, I love that actor.

    "TIME IS GOING TO END"

    I hate those lines, but I'll admit I'm particularly critical and picky when it comes to story driven games, I'm not a fan of the genre
     

Share This Page