Results 1 to 2 of 2

Thread: Gathering benchmark results

  1. #1
    Basic Member
    Join Date
    Nov 2011
    Posts
    25

    Gathering benchmark results

    I think proper benchmark results can help realize if your low fps is a game/driver/hardware problem. If everyone will do the benchmark on same scene, results will actually mean something to everyone.

    This method is not a perfect one, but its better than regular frapslog.txt results (min fps value is plain wrong, doesn't show most of the fps drops). One bad thing that it cant test vulcan mode, but doing a benchmark with a software that can measure vulcan frame-times is way more difficult.

    What software you're going to need:

    1) dota 2 with console enabled

    2) fraps 3.5.99 http://www.fraps.com/download.php

    3) 'FrafsBenchViewer' to build a graph based on the results you get from fraps, allowing us to analyze the frame-times and see the frame drops. Get it here https://sourceforge.net/projects/frafsbenchview/ . For those who don't want to install random apps on their pc, I made a portable version https://mega.nz/#!ewAj0IyT!Zi7O7dJPf...c-IDqVCsUKv22I (virustotal https://www.virustotal.com/en/file/e...is/1481671358/ )

    4) (optional) GPU-Z to monitor your GPU load. Get the latest version here https://www.techpowerup.com/downloads/SysInfo/GPU-Z/

    5) (optional) notepad to copy instructions there and close the browser and all the other apps, if you usually play with no apps open to get the max performance


    Setting up the fraps

    Press Disable on Movies tab-Video capture hotkey to prevent occasional video recording

    FPS tab http://imagizer.imageshack.com/img924/1553/BbOmI4.jpg

    The most important settings here is the Frametimes checkbox and 10 seconds duration of the benchmark. Make sure to make your settings the same.

    Now how to do a benchmark that would actually allow to compare results from different settings and PC's:

    1) Download replay ID 2844309781 and open it

    2) Move to middle of the game, select player perspective and double-click on underlord, pause the game

    3) Copy this command demo_goto 103000 and execute it in dota 2 console

    4) Unpause the game, wait for 52:37 in-game time and press the benchmark key

    5) Wait until its over (10 seconds)

    6) Open frafs bench viewer, it should open your fraps benchmarks directory. Drag the latest csv file with word 'frametimes' at the end of it into the frafs bench viewer

    7) Look for the avg.time, 1% time and 0.1% time FPS values. That's your average fps, average-low fps and the absolute worst fps drops values.

    example on the picture: http://imagizer.imageshack.com/img923/884/IhehgU.jpg

    Post your CPU model, frequency if its overclocked, approximate CPU load value if you measured it during the test, GPU model, approximate GPU load value if you measured it during the test, gpu driver version, amount of RAM, your OS version, specify if its installed on ssd or hdd

    then write the resolution and settings you've run benchmark on and the three values you got.


    My results:
    i7-4770k, 41% load, gtx 1060 3gb, 35% load, 375.95, 8 gb ram, win7x64sp1, hdd

    1080p -gl high no vsync all settings enabled render quality 100% shadow quality high: 44/29/18

    1080p -dx11 high no vsync all settings enabled render quality 100% shadow quality high: 44/31/19

    1080p -dx11 high no vsync all settings except ambient cloth simulation and grass, render quality 100% shadow quality high: 44/31/20


    Looks like dx11 is better than gl on the low fps, but not that much. Ambient cloth simulation and grass don't really impact performance on my configuration.

  2. #2
    Basic Member aveyo's Avatar
    Join Date
    Aug 2012
    Location
    EU West
    Posts
    2,922
    I've seen your reddit thread but you have deleted it before I could reply.
    I had around a benchmark.cfg that automates the job more and does not need any external tools - simply uses Source 2 built-in benchmark and reporting.
    I've updated and published it with your match id, as it seems to be spot-on at identifying particle spam issues.

    benchmark.cfg
    Code:
    ///////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
    ////
    ////     AveYo's DOTA 2 BENCHMARK.CFG v2 - inspired by the inovative D-OPTIMIZER V2 cfg engine (RIP)
    ////
    ////     1: remove .txt extension and copy this file to \Steam\steamapps\common\dota 2 beta\game\dota\cfg\benchmark.cfg
    ////     2: change Dota 2 launch options (-dx11 / -dx9 / -vulkan / -nod3d9ex / -nosound )
    ////     3: open Dota 2 and go to Watch - Replays - enter 2844309781 in Search for MatchIDs then download it
    ////        credits for the matchid goes to dev.dota user: jstq
    ////     4: to start, enter into console: exec benchmark
    ////     5: change video settings and repeat step 4
    ////
    ////     Results are saved to \Steam\steamapps\common\dota 2 beta\game\dota\ Source2Bench.csv and particle_profile.csv
    ////     and vprof summary is also printed to console when finished
    ////     more details can be viewed in console.log if adding extra launch options: -console -condebug -conclearlog
    ////
    ////     v2 Changes:
    ////     - initial short warmup now uses normal demo playback
    ////     - removed demo_goto usage (caused Source2Bench.csv not writing fps number - it was reported fine in console)
    ////     - timedemo now runs uninterrupted for more reliable results  
    ////     - added particle usage reporting
    ////     - improved ui feedback
    ////
    ///////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
    
    
    ///////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
    ////  BENCHMARK TARGET - IF YOU CHANGE IT, ALSO ADJUST BENCHMARK PARAMETERS                                             DEFAULT
    ///////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
    alias warmup_target "playdemo replays\2844309781";//                                REPLACE replays\2844309781 WITH YOUR CHOICE
    alias benchmark_target "timedemo replays\2844309781 fps.csv";//                     REPLACE replays\2844309781 WITH YOUR CHOICE
    
    ///////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
    ////  BENCHMARK PARAMETERS
    ////  version 1 reporting was fine for console but did not write actual fps to Source2Bench.csv due to demo_goto usage
    ///////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
    alias benchmark_start_at "timedemo_start 103000";// replace -1 (from 0) with an actual tick from your benchmark_target     USED
    alias benchmark_end_at "timedemo_end 104200";// replace -1 (until end) with an actual tick from your benchmark_target      USED
    pastebin mirror

    Update:
    Benchmark.cfg v2 with more reliable reporting and improved ui feedback

Posting Permissions

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