High DX11 CPU overhead, very low performance.

Discussion in 'Videocards - AMD Radeon Drivers Section' started by PrMinisterGR, May 4, 2015.

  1. lorikano

    lorikano Master Guru

    Messages:
    323
    Likes Received:
    0
    GPU:
    2x Sapphire 7950@1150
    Has anyone with moded drivers benchmarked GTA V ? Any difference from 15.4b ?
     
  2. Blackfyre

    Blackfyre Maha Guru

    Messages:
    1,101
    Likes Received:
    89
    GPU:
    RTX 2070 Super
    15.4b was released for GTA V. So I don't think 15.2 would have the optimizations made for GTA V in it, but I'll give it a run anyway.

    EDIT #1:

    I'll put the results in this post after I have them with further edits below but I wanted to ask since this was the first time I ran the benchmark. Once the benchmark finished the game just loaded story mode. Did it save the benchmark results in a file or document somewhere?

    EDIT #2: Don't worry I found it in documents:

    Okay so 15.2 VBS Driver:

    Frames Per Second (Higher is better) Min, Max, Avg
    Pass 0, 64.997253, 127.447174, 102.546280
    Pass 1, 67.982399, 124.417198, 91.889137
    Pass 2, 56.416992, 186.804352, 101.175278
    Pass 3, 58.073906, 177.216156, 95.134003
    Pass 4, 27.611479, 166.088760, 98.744766

    Time in milliseconds(ms). (Lower is better). Min, Max, Avg
    Pass 0, 7.846388, 15.385265, 9.751695
    Pass 1, 8.037474, 14.709690, 10.882679
    Pass 2, 5.353194, 17.725157, 9.883838
    Pass 3, 5.642827, 17.219439, 10.511489
    Pass 4, 6.020877, 36.216820, 10.127119

    Frames under 16ms (for 60fps):
    Pass 0: 960/960 frames (100.00%)
    Pass 1: 865/865 frames (100.00%)
    Pass 2: 931/934 frames (99.68%)
    Pass 3: 884/887 frames (99.66%)
    Pass 4: 11201/11256 frames (99.51%)

    Frames under 33ms (for 30fps):
    Pass 0: 960/960 frames (100.00%)
    Pass 1: 865/865 frames (100.00%)
    Pass 2: 934/934 frames (100.00%)
    Pass 3: 887/887 frames (100.00%)
    Pass 4: 11255/11256 frames (99.99%)

    EDIT #3:

    And just finished benchmarking 15.4b under the exact same settings as above (honestly haven't even compared the results yet myself, just going to paste here and check them after I've posted).

    Frames Per Second (Higher is better) Min, Max, Avg
    Pass 0, 70.865593, 117.549049, 94.048401
    Pass 1, 54.564987, 145.592819, 91.238739
    Pass 2, 5.599658, 277.580963, 98.627220
    Pass 3, 65.156090, 157.126801, 94.229210
    Pass 4, 50.930817, 197.562988, 97.067764

    Time in milliseconds(ms). (Lower is better). Min, Max, Avg
    Pass 0, 8.507087, 14.111220, 10.632823
    Pass 1, 6.868471, 18.326771, 10.960257
    Pass 2, 3.602552, 178.582321, 10.139189
    Pass 3, 6.364286, 15.347759, 10.612420
    Pass 4, 5.061677, 19.634478, 10.302081

    Frames under 16ms (for 60fps):
    Pass 0: 883/883 frames (100.00%)
    Pass 1: 857/858 frames (99.88%)
    Pass 2: 882/887 frames (99.44%)
    Pass 3: 882/882 frames (100.00%)
    Pass 4: 11051/11083 frames (99.71%)

    Frames under 33ms (for 30fps):
    Pass 0: 883/883 frames (100.00%)
    Pass 1: 858/858 frames (100.00%)
    Pass 2: 882/887 frames (99.44%)
    Pass 3: 882/882 frames (100.00%)
    Pass 4: 11083/11083 frames (100.00%)
     
    Last edited: May 6, 2015
  3. Deathchild

    Deathchild Ancient Guru

    Messages:
    3,970
    Likes Received:
    2
    GPU:
    -
    +1

    whql/beta

    But anyways, it seems they've definitely done something to improve the overall overhead, SUPERB! ...

    Expect to see better performance in ALL DX11 GAMES, SICK. :D
     
  4. Blackfyre

    Blackfyre Maha Guru

    Messages:
    1,101
    Likes Received:
    89
    GPU:
    RTX 2070 Super
    I've just posted the results above... not sure what to make of them for GTA V if anyone is interested, some things are higher, other things are lower. But Average FPS is slightly higher using VBS 15.2 (in both instances the only things running in the background were steam and gta v after a fresh restart).

    EDIT: I still don't think this is a fair test because 15.2 driver wouldn't have any of the GTA Optimizations made in 15.4 (Unless someone can transfer those into the 15.2 driver and mod it so that it has everything GTA related from 15.4b, it won't be a fair test)

    EDIT #2:

    Also to note, in benchmark #2 which is the 15.4b Driver benchmark, the final benchmark (Pass #4) the character was not shooting at the police from the car at all (no breaking windows, etc) could explain why it ran better on 15.4b.

    EDIT #3:

    Just noticed even though min/max fps were lower, average FPS was still higher using 15.2 (With regards to Pass #4).

    EDIT #4:

    Would be great if someone with a much slower CPU than mine and a cross-fire setup can test the VBS Drivers from Page 5 and compare them to 15.4b...

    Once the GTA Benchmark finishes *which you will find under graphics settings in the settings menu* Copy the results the way I did *which you will find under documents/rockstar games/gtav/benchmarks*
     
    Last edited: May 6, 2015

  5. Deathchild

    Deathchild Ancient Guru

    Messages:
    3,970
    Likes Received:
    2
    GPU:
    -
    So VBS 15.2 seems higher to you?

    Yeah.. you're right, if anyone could take the good stuff outta 15.4b and put it in the 15.2 vbs driver lol, would be awesome.

    See if you guys have any other benchmark tests though.
     
  6. Blackfyre

    Blackfyre Maha Guru

    Messages:
    1,101
    Likes Received:
    89
    GPU:
    RTX 2070 Super
    During the (most graphically demanding scene I believe) it definitely felt smoother. The benchmark results back that up I believe, minimum FPS was 56 FPS instead of 5 FPS on the 15.4b (look at Pass #2).

    Basically with 15.2 VBS there was 0 micro-stuttering, with 15.4b there was a slight micro-stutter in that plane scene I think... with the overall feel being that 15.2 is smoother, but feelings are subjective :p numbers speak better.
     
    Last edited: May 6, 2015
  7. Deathchild

    Deathchild Ancient Guru

    Messages:
    3,970
    Likes Received:
    2
    GPU:
    -
    Hmm alright.. might give it a try then, thanks for that anyways.
     
  8. Blackfyre

    Blackfyre Maha Guru

    Messages:
    1,101
    Likes Received:
    89
    GPU:
    RTX 2070 Super
    Yeah check out my EDIT #4 from above (two or three posts ago) which I added later. Would be great if you can benchmark it with crossfire (might be an even bigger difference).
     
    Last edited: May 6, 2015
  9. Deathchild

    Deathchild Ancient Guru

    Messages:
    3,970
    Likes Received:
    2
    GPU:
    -
    Yeah I'll do it.
     
  10. Demrock

    Demrock Member

    Messages:
    45
    Likes Received:
    0
    GPU:
    Sapphire Boost 7950x2
    Holy **** PrMinisterGR what a awesome thread !

    +10 !!!!
     

  11. Deathchild

    Deathchild Ancient Guru

    Messages:
    3,970
    Likes Received:
    2
    GPU:
    -
    Demrock if you have GTA V could you please do the benchmark tests too? Let's test together.

    Would be super interesting to see, sad though that the 15.2 don't have the 15.4b optimizations.
     
  12. theoneofgod

    theoneofgod Ancient Guru

    Messages:
    4,623
    Likes Received:
    259
    GPU:
    RX 580 8GB
    I'll do them too in a few hours.
     
  13. Deathchild

    Deathchild Ancient Guru

    Messages:
    3,970
    Likes Received:
    2
    GPU:
    -
    We need to use same settings then, or I'll post my settings here.
     
  14. theoneofgod

    theoneofgod Ancient Guru

    Messages:
    4,623
    Likes Received:
    259
    GPU:
    RX 580 8GB
    Post your settings.
     
  15. PrMinisterGR

    PrMinisterGR Ancient Guru

    Messages:
    7,890
    Likes Received:
    755
    GPU:
    Inno3D RTX 3090
    You do have DX12 support, and Tier 3 binding resources at that. That screenshot was taken with a driver that does not include the DX12 driver in it. You can see the actual hardware support (that is complete in hardware for DX12 and GCN), in previous posts.

    GTA resets the graphics settings between drivers. I learned it the hard way when I did my benchmark run with them. Make absolutely sure that the ==SETTINGS== part of the benchmark.txt is absolutely the same for both drivers. After I did it, the Vbs driver was slower. But it felt less juddery still, which I can't back up with numbers :(

    EDIT: Since most people ask for GTA V Benchmarks, here are mine. My specs are on the left. I originally posted them here.

    I have a single gpu. I did two runs with each driver, and I kept the best. Here are some results. My system spec is on the left. RadeonPro was NOT running while running the tests. My OS is Windows 8.1.1 x64, and I'm running Process Lasso in the background and I have "Gaming" mode enabled for GTAV.exe, which disables core parking and runs my CPU at minimum 3.4GHz all the time.

    Here they are:
    [​IMG]

    [​IMG]

    [​IMG]

    [​IMG]

    [​IMG]

    I played a bit, it somehow feels smoother here, less juddery, but the numbers don't seem to be backing it up. There is also that max framerate anomaly on the second pass.

    Keep in mind that this is not the "GTA V" driver, like the 15.4 is. I wonder what the combination would give to my poor, old, slow 7970. :D

    These were my game settings for testing both drivers.

    Code:
    === SETTINGS ===
    Display: 1920x1080 (FullScreen) @ 59Hz VSync OFF
    Tessellation: 2
    LodScale: 0.700000
    PedLodBias: 0.200000
    VehicleLodBias: 0.000000
    ShadowQuality: 3
    ReflectionQuality: 2
    ReflectionMSAA: 0
    SSAO: 0
    AnisotropicFiltering: 16
    MSAA: 0
    MSAAFragments: 0
    MSAAQuality: 0
    TextureQuality: 2
    ParticleQuality: 0
    WaterQuality: 1
    GrassQuality: 1
    ShaderQuality: 1
    Shadow_SoftShadows: 4
    UltraShadows_Enabled: false
    Shadow_ParticleShadows: true
    Shadow_Distance: 1.000000
    Shadow_LongShadows: false
    Shadow_SplitZStart: 0.930000
    Shadow_SplitZEnd: 0.890000
    Shadow_aircraftExpWeight: 0.990000
    Shadow_DisableScreenSizeCheck: false
    Reflection_MipBlur: true
    FXAA_Enabled: true
    TXAA_Enabled: false
    Lighting_FogVolumes: true
    Shader_SSA: true
    DX_Version: 2
    CityDensity: 1.000000
    PedVarietyMultiplier: 1.000000
    VehicleVarietyMultiplier: 1.000000
    PostFX: 3
    DoF: true
    HdStreamingInFlight: false
    MaxLodScale: 0.000000
    MotionBlurStrength: 1.000000
     
    Last edited: May 6, 2015

  16. Vbs

    Vbs Master Guru

    Messages:
    291
    Likes Received:
    0
    GPU:
    Asus Strix 970, 1506/7806
    Statistically, both driver results are the same. Next time, add the Standard Error bars to the graph. ;)

    Your average for 15.4 is 80.27 fps with a 95% Confidence Interval between 75.14 and 85.39 fps. For 1018.1 it's an average of 79.68 fps, 95% CI [74.63 84.73]. The confidence interval is too big to draw any comparisons between both, you would need a much larger sample.

    If you can get the standard deviations from the runs, those would better represent frame variations, as a low std will represent a low frame time variation. :)
     
    Last edited: May 6, 2015
  17. semitope

    semitope Member

    Messages:
    36
    Likes Received:
    0
    GPU:
    iGPU

    You've provided the "evidence", I'm just pointing out the questionable nature of your interpretation of it.

    You really need to examine your reasoning on this. Consider the 260x benchmarks that lost much less. This is simple bottlenecking that you are turning into a big deal. The higher end nvidia cards lose similar performance on the i3, lower end amd cards lose less performance. etc etc. The logic is invalid. You're probably posting this on several forums I bet.
     
  18. Deathchild

    Deathchild Ancient Guru

    Messages:
    3,970
    Likes Received:
    2
    GPU:
    -
    Wtf is this semitope guy talking about, PrMinisterGR is talking about overhead, dear lord, it's a pretty known fact now that the overhead on the amd drivers (pre win10) are worse than with the nvidia counterpart. Simply the nvidia drivers are a little more efficient, in cpu situations.

    So what he speaks sorta is the truth, you get less if you have a weaker cpu, whereas if you had a stronger one it wouldn't really maybe matter. (Although I'm sure it does on all levels). Any cpu heavy situation.

    So, you with 4 posts or PrMinisterGR who is a pretty known guy here now, who the f are w supposed to believe lol, ofc PrMinisterGR... wtf... troll.

    Are you from AMD lol? Or an AMD fan.
     
  19. theoneofgod

    theoneofgod Ancient Guru

    Messages:
    4,623
    Likes Received:
    259
    GPU:
    RX 580 8GB
    They should be posting it everywhere possible as it needs to be known.
     
  20. Deathchild

    Deathchild Ancient Guru

    Messages:
    3,970
    Likes Received:
    2
    GPU:
    -
    Dude how do you have such low drawcalls? rofl look at mine, I have 10x more than you, no wonder GTA V working **** for you man omfg. :D
    ///
    edit: Wait why did you do the test @1080p, didn't it default at 720p, like me? o,O Weird.
    ///
    It's probably your 3770 @ 4.1GHz, you need 4.8GHz like me. :D

    So my test with current 15.4b drivers, will do vbs drivers in a sec
    [​IMG]

    semitope is just probably someone from amd lol or doesn't understand what we're talking about .. :D
     
    Last edited: May 6, 2015

Share This Page