Browser speed results
I watched the Apple keynote yesterday and I was not impressed. That is not the point of this post though. I wanted to do my own tests on browsers and see for myself which would be the fastest (since Steve claimed Safari was the fastest). I tested Opera 9.21, Safari 3.0, Internet Explorer 7, Firefox 2.0.0.4 (no Add-ons), Firefox with fasterfox 2.0.0 enabled, and Firefox 3 nightly build (alpha 6) and I used WebWait to conduct my tests. All results were the average of 5 loads with 60 seconds wait time between refreshes. I tested by loading Microsoft.com, Apple.com, GetFireFox.com, Digg.com, and Slashdot.org. Before each test I made sure that all cookies and cache was erased and that the browsers were as freshly installed as possible. (with a few of the browsers they were fresh installs). Here were the results I found.
| Microsoft | GetFirefox | Apple | Digg | Slashdot | ||
| Browser | ||||||
| FireFox | 1.52 | 1.09 | 1.33 | 1.83 | 1.98 | |
| FireFox (w/ FasterFox) | 1.67 | 1.27 | 0.47 | 2.10 | 3.23 | |
| FireFox (nightly) | 0.85 | 0.50 | 0.35 | 1.72 | 1.57 | |
| Safari | 1.47 | 1.19 | 1.09 | 1.72 | 1.07 | |
| Opera | 1.58 | 2.97 | 0.66 | 2.50 | 3.12 | |
| IE7 | 1.20 | 0.29 | 0.43 | 3.07 | 1.15 |
This is the test results I found. I did not know which browser would win and I really did not care (It won’t stop me from using FireFox). If you do not believe my test results please test it yourself on your own computer. I ran all of these tests on a Windows XP machine with a P4 HT 3.2 with 1 GB of ram. The internet connection is at my work (more bandwidth than I know) and all tests were done within the same 4 hour block so all traffic on the network should be about the same. Leave a comment with how you feel about browser wars or the results I found.
How do you know which server you got sent to? A 4 hour block is too long. I would say all tests need to be done within a minute. My site for example load balances between 2 server centers and the dns changes every minute. Load times would really be best measured from a local server.
The sites you measured are huge and I bet they have similar load balancing techniques.