User Tag List

Results 1 to 5 of 5

Thread: First Frame latency Benchmarks for NAG Online

  1. #1

    Default First Frame latency Benchmarks for NAG Online

    So I've been figuring this out for half the day while working on another review and I think I've gotten it down. I'm testing out the Catalyst 13-04 drivers against the ones that were least released for the HD6000 generation - 12-10. I've never had reason to install the new ones since most of the time things went swimmingly.

    However, sometimes drivers do make things better and I'll put up my results so far. This is just the result of a day's testing and it's nothing extensive. I did find one interesting benefit from the drivers, though. I won't post up FPS averages for now, just the latency graphs. I'm still working out how to show things like the 99th percentile and **** like that, so this is a work in progress. I can sort of do scatter graphs as well, but I'm terrible at it.

    Component: Model: Speed:
    Processor AMD Athlon II X3 445 3.1GHz
    RAM 4GB Hynix and Kingston DDR3-1333
    Motherboard ECS A785GM-AD3 HT2000
    Graphics card AMD Radeon HD6870 1GB 900MHz core, 1050MHz VRAM
    Operating System Windows 8 Professional 64-bit


    GAME SETTINGS:

    Borderlands 2: Settings:
    Framerate Unlimited
    Resolution 1920 x 1080
    Ambient Occlusion On
    Depth of Field On
    Anisotropic Filtering 4x
    Bullet Decals High
    Foliage Distance Far
    Texture Quality High
    Game Detail High
    View Distance High
    Physc Low
    FXAA Off

    DiRT 3: Settings:
    Resolution 1920 x 1080
    Refresh rate 74Hz
    VSync Off
    Night Lighting High
    Shadows High
    Particles High
    Mirrors High
    Characters High
    Ground Cover High
    Distant Vehicles High
    Objects High
    Trees High
    Vehicle Reflections High
    Water High
    Post Process High
    Skidmarks On
    Ambient Occlusion High
    Cloth High

    Crysis 3: Settings:
    Resolution 1920 x 1080
    Anti-Aliasing SMAA Low
    Texture Resolution Low
    VSync Off
    Game Effects Medium
    Objects Low
    Particles Low
    Shading Medium
    Shadows Low
    Water Low
    Anisotropic Filtering 1x
    Motion Blur Low
    Lens Flare Yes

    I did the uninstall thing and it went fine. There's nothing particularly interesting about the new drivers from a functionality standpoint, but the Gaming section in CCC gets some changes. There's now an option for frame rate control, which is essentially VSync. You can now force Triple buffering globally for OpenGL titles. Not much else besides that the the usual other features.

    For Reference:

    Frame Time (ms) FPS Rate
    8.3 120
    16.7 60
    20 50
    25 40
    30
    50 20


    BORDERLANDS 2:


    I picked Sawtooth Cauldron for the test. From the beginning I run to the right and jump down the cliff into the landing of a double-story building. I kill one enemy coming out of the garage, hit the gas tank in front of me, turn around and kill two more, run over the guardrails and jump to the ground and make my way to the slope that finally takes me back to the entrance. During all this time there's lots of light rays, some lens flare, smoke and fire along with grenade effects, elemental effects from my weapons, a pretty large and detailed area as well as ****loads of enemies. I think it's pretty good.

    The results are actually pretty similar, so there's no performance increase from the drivers here. The longer frame times are when I got blocked in by enemies or I went into "Fight or Die" mode and I was being swarmed. They are long hiccups but I'm sure it's minor. I ran the test again to be sure and it wasn't as bad. There's perhaps a 2-3% increase here.

    DiRT 3:


    I wish I had Showdown right now, but this will have to do. I pretty much dialed in the settings, saved them and then ran the benchmark that's found in the graphics settings all the way at the bottom.

    Gameplay on both the 12-10 and 13-4 drivers was pretty good, although the high settings do seem to swamp my card a bit. I was averaging 50fps there, which is pretty good. Moving to 13-4 the variance became a little more pronounced on the graph but I didn't really notice it in the game. Considering that both DiRT 3 and my Radeon HD6870 came out over two years ago, this isn't really a surprise. I did, however, notice something when fiddling with things and turning the graphics details to Ultra where possible.



    Excuse the same colours, for now I just wanted to get this process finished so I can tweak it later. As you can see, enabling Ultra details doesn't seem to have impacted performance - in fact, most of the time it's actually running better than the High detail settings. I didn't make a third graph comparing 12-10 High to 13-4 Ultra, but you can see for yourself that it's a lot better and more stable. That's one win for the drivers there.

    Crysis 3:


    In some sections, my play-through of Crysis with 13-4 is better than 12-10. It pushes out a slightly higher amount of frames in the 60-second benchmark. Those two long ones stretching to 140ms actually measure 1800ms in length. This is similar behaviour to Crysis 2 and it's related to the game engine. Mind you, I'm also playing on low settings. Crysis 3 is really heavy on my triple-core processor and my 1GB of VRAM suffers wheny ou turn the details up to Medium.



    Just high frame times and stuttering everywhere. I'd hate to see numbers from Far Cry 3, I'm sure that's equally devastating. I'd probably benefit more from a new GPU than a processor upgrade at this point.

    So 13-4 is only a tiny amount better for HD6000 owners and AMD isn't fixing any latency numbers for the older GPU family with this driver set (if anything, latency for my games actually was a little worse). I have heard about the 13-5 Beta 2 drivers that are out and I will have a test soon, but I've got a feeling it'll only be the same, possibly worse. AMD's putting all their best into GCN optimisation and for those of you still on cards like mine, Catalyst 12-10 will be good enough. There might be bug fixes and the like for 13-4, though, so you might as well get it because sometimes AMD fixes things for one game that also fix issues for other titles at the same time.
    Last edited by Wesley; 28-04-2013 at 03:21 PM.

  2. #2

    Default Re: First Frame latency Benchmarks for NAG Online

    Thanks for the in-depth analysis.

  3. #3

    Default Re: First Frame latency Benchmarks for NAG Online

    I am a nVidia man myself, but I must commend you on your analysis.

  4. #4

    Default Re: First Frame latency Benchmarks for NAG Online

    Quote Originally Posted by Folset View Post
    I am a nVidia man myself, but I must commend you on your analysis.
    Thanks :) I could pull out the GT240 in the other machine in the house and do comparitive benchmarks :-P

    And hopefully in future I'll be able to do my reviews in the same manner as well.

  5. #5

    Default Re: First Frame latency Benchmarks for NAG Online

    COOL!

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •