4 ms·

Updated WebGL Benchmark Results

41 points by Aissen 11y ago

brudgers 11y ago
My understanding is that Unity3d integrates well with asm.js and Mozilla sort of has a head start in that.
- drawkbox 11y ago
  Yes both Unity and Unreal started their WebGL engine output with asm.js/emscripten and Mozilla, they both announced it at GDC with Mozilla that year (2014)[1]. [1] https://blog.mozilla.org/blog/2014/03/18/gdc-2014-mozilla-and-partners-prove-the-web-is-the-platform-for-gaming/ https://blog.mozilla.org/blog/2014/03/18/gdc-2014-mozilla-an...
- clouddrover 11y ago
  On the other hand, Edge's asm.js implementation is younger than Chrome's but Edge is second fastest in Unity's benchmark while Chrome is third. Chrome's head start hasn't helped it beat Edge in this benchmark.
  - FeepingCreature 11y ago
    Chrome does not specifically optimize for asm.js.
    - fulafel 11y ago
      It does kinda, according to https://hacks.mozilla.org/2015/03/asm-speedups-everywhere/ https://hacks.mozilla.org/2015/03/asm-speedups-everywhere/
    - clouddrover 11y ago
      Sure it does. Chrome has been specifically optimizing for asm.js for over two years (see [1]). Google's own Octane browser benchmark has included an asm.js benchmark for over two years (see [2][3][4]). Game developers have been commenting on Chrome's asm.js performance for over two years (see [5]). Chrome is just slower. Microsoft has done a good job with Edge's JavaScript engine. [1] http://blog.chromium.org/2013/05/chrome-28-beta-more-immersive-web.html http://blog.chromium.org/2013/05/chrome-28-beta-more-immersi... [2] http://blog.chromium.org/2013/11/announcing-octane-20.html http://blog.chromium.org/2013/11/announcing-octane-20.html [3] http://techcrunch.com/2013/11/06/google-updates-its-octane-javascript-benchmark-adds-asm-js-and-typescript-tests/ http://techcrunch.com/2013/11/06/google-updates-its-octane-j... [4] https://developers.google.com/octane/ https://developers.google.com/octane/ [5] https://www.scirra.com/blog/132/more-about-asmjs-powered-physics https://www.scirra.com/blog/132/more-about-asmjs-powered-phy...
      - FeepingCreature 11y ago
        Let me clarify: Chrome does not to my knowledge contain any special optimizations that are only active in asm.js mode. Firefox does, to the extent of having an entire separate codepath for asm.js AOT compilation.
        clouddrover 11y ago
        Chrome contains special optimizations that are only active in asm.js mode: http://blog.chromium.org/2015/07/revving-up-javascript-performance-with.html http://blog.chromium.org/2015/07/revving-up-javascript-perfo... I don't know what you're arguing here. It doesn't matter what approach is taken to optimizing performance, all that matters is the performance achieved. Chrome is simply slower. Microsoft is to open source Edge's JavaScript engine (https://blogs.windows.com/msedgedev/2015/12/05/open-source-chakra-core/ https://blogs.windows.com/msedgedev/2015/12/05/open-source-c...), so perhaps the Chrome developers can study that for ideas on speeding up Chrome.
        FeepingCreature 11y ago
        Huh. Fair enough, cool I guess. It was my impression that for a long time, Chrome's stance was basically "We aren't gonna bother optimizing specifically for asm.js; instead, we're just gonna make JS faster in general." Which would seem to explain why > Chrome's head start hasn't helped it beat Edge in this benchmark.
vegabook 11y ago
Amazing to see Firefox Nightly more than double the speed of Safari on my late 2013 i7 mac mini.
tgb 11y ago
Unitless scores with no "higher is better" label meant I spent far too long trying to guess what the graphs meant.
throwaway13337 11y ago
Unfortunately, the packed WebGL file size for a scene with nothing in it comes at ~80mbs which is still way too big to be a reasonable alternative to three.js-based libs. I hope they'll be able to trim it down in the future because I'd really like to use unity to deploy to WebGL. It just doesn't seem to be something that they can do, for whatever reason. Edit: The benchmark seems to have a minimalist packed size of just under 10mbs. Cool that they achieved that. I haven't been able to with my project.
- golergka 11y ago
  Right now making a mature Unity3d project that was developed for mobile platforms run on WebGL takes 2-4 weeks of developer time to reach feature-complete, semi-stable state. And still, it runs abysmally slow, and hogging at least 1Gb of memory. After Unity browser plugin was retired a year ago with the whole Netscape plugin architecture, teams that use Unity just better accept the fact that they can't realistically deploy their projects in the browser anymore.
  - throwaway13337 11y ago
    I just spent the last 20 minutes compiling my unity project with 5.3 webgl. It's a voxel game so it's quite small, asset-wise and fits into under 20mbs! This is a huge leap. The upstart performance feels a bit like a java applet (freezes for a second... not great) but once it gets started, it's performant for my use case. In my brief test, I only notice one weird thing (the input sometimes got screwy). This is a huge improvement. Give it another try?
    - golergka 11y ago
      Uhm, is your game a commercial project, with a lot of art, complicated requirements involving SDKs from social networks as well as marketing, analytical and bug report tools, commits and random code from incompetent developers who were fired a year ago, urgent gufixes that were put in as "temporary" and stayed forever? Because of course it would work nicely with a small hobby project ran by a single programmer of adequate level.

41 points by Aissen 11y ago

brudgers 11y ago

My understanding is that Unity3d integrates well with asm.js and Mozilla sort of has a head start in that.

drawkbox 11y ago
Yes both Unity and Unreal started their WebGL engine output with asm.js/emscripten and Mozilla, they both announced it at GDC with Mozilla that year (2014)[1]. [1] https://blog.mozilla.org/blog/2014/03/18/gdc-2014-mozilla-and-partners-prove-the-web-is-the-platform-for-gaming/ https://blog.mozilla.org/blog/2014/03/18/gdc-2014-mozilla-an...
clouddrover 11y ago
On the other hand, Edge's asm.js implementation is younger than Chrome's but Edge is second fastest in Unity's benchmark while Chrome is third. Chrome's head start hasn't helped it beat Edge in this benchmark.
- FeepingCreature 11y ago
  Chrome does not specifically optimize for asm.js.
  - fulafel 11y ago
    It does kinda, according to https://hacks.mozilla.org/2015/03/asm-speedups-everywhere/ https://hacks.mozilla.org/2015/03/asm-speedups-everywhere/
  - clouddrover 11y ago
    Sure it does. Chrome has been specifically optimizing for asm.js for over two years (see [1]). Google's own Octane browser benchmark has included an asm.js benchmark for over two years (see [2][3][4]). Game developers have been commenting on Chrome's asm.js performance for over two years (see [5]). Chrome is just slower. Microsoft has done a good job with Edge's JavaScript engine. [1] http://blog.chromium.org/2013/05/chrome-28-beta-more-immersive-web.html http://blog.chromium.org/2013/05/chrome-28-beta-more-immersi... [2] http://blog.chromium.org/2013/11/announcing-octane-20.html http://blog.chromium.org/2013/11/announcing-octane-20.html [3] http://techcrunch.com/2013/11/06/google-updates-its-octane-javascript-benchmark-adds-asm-js-and-typescript-tests/ http://techcrunch.com/2013/11/06/google-updates-its-octane-j... [4] https://developers.google.com/octane/ https://developers.google.com/octane/ [5] https://www.scirra.com/blog/132/more-about-asmjs-powered-physics https://www.scirra.com/blog/132/more-about-asmjs-powered-phy...
    - FeepingCreature 11y ago
      Let me clarify: Chrome does not to my knowledge contain any special optimizations that are only active in asm.js mode. Firefox does, to the extent of having an entire separate codepath for asm.js AOT compilation.
      - clouddrover 11y ago
        Chrome contains special optimizations that are only active in asm.js mode: http://blog.chromium.org/2015/07/revving-up-javascript-performance-with.html http://blog.chromium.org/2015/07/revving-up-javascript-perfo... I don't know what you're arguing here. It doesn't matter what approach is taken to optimizing performance, all that matters is the performance achieved. Chrome is simply slower. Microsoft is to open source Edge's JavaScript engine (https://blogs.windows.com/msedgedev/2015/12/05/open-source-chakra-core/ https://blogs.windows.com/msedgedev/2015/12/05/open-source-c...), so perhaps the Chrome developers can study that for ideas on speeding up Chrome.
        FeepingCreature 11y ago
        Huh. Fair enough, cool I guess. It was my impression that for a long time, Chrome's stance was basically "We aren't gonna bother optimizing specifically for asm.js; instead, we're just gonna make JS faster in general." Which would seem to explain why > Chrome's head start hasn't helped it beat Edge in this benchmark.

vegabook 11y ago

Amazing to see Firefox Nightly more than double the speed of Safari on my late 2013 i7 mac mini.

tgb 11y ago

Unitless scores with no "higher is better" label meant I spent far too long trying to guess what the graphs meant.

throwaway13337 11y ago

Unfortunately, the packed WebGL file size for a scene with nothing in it comes at ~80mbs which is still way too big to be a reasonable alternative to three.js-based libs. I hope they'll be able to trim it down in the future because I'd really like to use unity to deploy to WebGL. It just doesn't seem to be something that they can do, for whatever reason. Edit: The benchmark seems to have a minimalist packed size of just under 10mbs. Cool that they achieved that. I haven't been able to with my project.

golergka 11y ago
Right now making a mature Unity3d project that was developed for mobile platforms run on WebGL takes 2-4 weeks of developer time to reach feature-complete, semi-stable state. And still, it runs abysmally slow, and hogging at least 1Gb of memory. After Unity browser plugin was retired a year ago with the whole Netscape plugin architecture, teams that use Unity just better accept the fact that they can't realistically deploy their projects in the browser anymore.
- throwaway13337 11y ago
  I just spent the last 20 minutes compiling my unity project with 5.3 webgl. It's a voxel game so it's quite small, asset-wise and fits into under 20mbs! This is a huge leap. The upstart performance feels a bit like a java applet (freezes for a second... not great) but once it gets started, it's performant for my use case. In my brief test, I only notice one weird thing (the input sometimes got screwy). This is a huge improvement. Give it another try?
  - golergka 11y ago
    Uhm, is your game a commercial project, with a lot of art, complicated requirements involving SDKs from social networks as well as marketing, analytical and bug report tools, commits and random code from incompetent developers who were fired a year ago, urgent gufixes that were put in as "temporary" and stayed forever? Because of course it would work nicely with a small hobby project ran by a single programmer of adequate level.