WEBVTT FILE 00:00:05.142 --> 00:00:08.501 My name is Jason Weber, and my job is to make Internet Explorer fast. 00:00:08.900 --> 00:00:14.305 We care about Windows users, and Windows users spend over half their time inside the web browser. 00:00:15.000 --> 00:00:17.199 We want those users to have a great experience 00:00:17.368 --> 00:00:18.390 and that means performance. 00:00:18.921 --> 00:00:22.593 Regularly on any given day we will measure the performance of IE hundreds of times 00:00:22.853 --> 00:00:25.504 typically using thousands of different tests 00:00:25.642 --> 00:00:31.291 which results in anywhere from five to six million plus different metrics that we end up analysing. 00:00:31.484 --> 00:00:33.567 We think about it in three different ways 00:00:33.882 --> 00:00:38.576 the first is we wanna make the web sites that customers use today extremely fast 00:00:38.712 --> 00:00:40.511 so your traditional browsing scenarios 00:00:40.616 --> 00:00:45.334 the second class is that we want to make the interactive AJAX applications 00:00:45.465 --> 00:00:48.958 that the customers use every day extremely responsive 00:00:49.102 --> 00:00:52.892 and the third class is really the HTML5 applications of tomorrow 00:00:53.022 --> 00:00:55.391 and enabling the next generation of the web. 00:00:55.547 --> 00:00:58.207 In order to measure the performance of the browser 00:00:58.340 --> 00:01:01.048 we sorta had to build our own version of the internet 00:01:01.191 --> 00:01:05.000 Both replicating different network topologies, different server configurations, 00:01:05.100 --> 00:01:12.000 as well as different client configurations and measuring anything from memory usage to CPU utilization, elapsed time, 00:01:12.191 --> 00:01:18.100 that end users see and really try to get a representative sample 00:01:18.191 --> 00:01:21.770 of what customers will see in their real world experience replicated in this lab. 00:01:21.905 --> 00:01:24.468 Benchmarks can be a very good way to measure performance. 00:01:24.611 --> 00:01:28.201 However, if they are are representative of what actual users are doing 00:01:28.300 --> 00:01:32.663 so if you have a small isolated benchmark it can be great for a subsystem 00:01:32.850 --> 00:01:37.523 but it is not so great to represent what the real world users are doing. 00:01:37.657 --> 00:01:40.023 Performance is a multidimensional problem 00:01:40.257 --> 00:01:44.023 and Javascript is only one of the eleven subsystems in the browser 00:01:44.158 --> 00:01:45.780 that impact performance 00:01:45.876 --> 00:01:50.384 With Internet Explorer 9 though we completely rearchitected our Javascript engine 00:01:50.521 --> 00:01:55.661 so that applications that make heavy use of Javascript are extremely fast. 00:01:55.790 --> 00:02:00.269 So we're optimizing for a wide range of machines. From things like the netbook over here 00:02:00.404 --> 00:02:05.077 to high end desktops with eight cores, lots of memory 00:02:05.214 --> 00:02:09.131 and we really wanna have that be a great experience across that range. 00:02:09.262 --> 00:02:14.829 Today's web browsers really haven't taken advantage of the full modern PC hardware 00:02:14.965 --> 00:02:18.918 and you're really limited to what developers can do inside their web applications. 00:02:19.089 --> 00:02:23.783 With Internet Explorer 9 we're unlocking the full power of the PC 00:02:23.957 --> 00:02:29.647 and enabling web developers the potential that native application developers have had for the past two decades. 00:02:29.786 --> 00:02:34.861 Our approach for graphics in IE9 has been to accelerate everything. 00:02:34.977 --> 00:02:40.005 Text rendering, graphics rendering, colors on background images, I mean 00:02:40.146 --> 00:02:43.786 really everything in IE9 that you see on the screen 00:02:43.902 --> 00:02:46.123 is powered and displayed through the GPU 00:02:46.271 --> 00:02:51.978 As we move into the future with vector graphics technology, whether it's canvas or SVG 00:02:52.083 --> 00:02:55.680 In IE9 those are also all hardware accelerated 00:02:55.772 --> 00:03:00.635 and as people employ those you should see much better performance than you would have 00:03:00.765 --> 00:03:03.101 had we not used the GPU. 00:03:03.175 --> 00:03:08.784 With Internet Explorer 9, the web pages that you browse to every day just got a whole lot faster 00:03:08.975 --> 00:03:11.214 and the HTML5 applications of tomorrow 00:03:11.385 --> 00:03:17.028 are now possible. 00:03:17.227 --> 00:03:19.164 I think long term what GPU accelerated graphics 00:03:19.330 --> 00:03:21.322 and coupled with the other performance improvements 00:03:21.431 --> 00:03:25.018 it changes what you can do 00:03:25.174 --> 00:03:28.341 Things don't change overnight in the world anymore 00:03:28.476 --> 00:03:30.104 Not even on the web 00:03:30.214 --> 00:03:35.957 But you look two years out and I think people will wonder what a browser was like without GPU acceleration.