How Far Is Google Chrome Pushing the Browser Speed Envelope?
2008-09-04 23:52 - Web
So, the big news this week is that Google has released their own browser, dubbed Chrome. They put together a comic book, of all things, to explain what, and why, it is. This was actually a very good idea; it was well made and quite nifty.
There's a variety of things going on with Chrome. It's pushing the browser speed barrier, at a point in time that the barrier is being rather well smashed from all sides. The point is clear: Google is all about web apps. Faster browsers mean Google's apps are more usable, and people can depend on them rather than the more traditional alternatives.
They built a javascript engine from scratch, with some awesome speed tweaks baked into the design. But is it really that fast? I've put together the most comprehensive and fair evaluation that I could come up with to examine just that point. Let's start with the graph:
I've cut off the top end of the scale; the numbers for IE smashed all the rest down to be unreadable. I also broke a decision I made starting off. I wanted to compare shipping browsers only, not betas. Neither IE6 nor 7 could run all the tests, however, so I budged and included the latest IE8 beta. As you can see, 8 is generally an improvement, but generally just to crappy from crappier.
The results are surprising, to say the least. It appears that, yes, Google Chrome really is that fast. Here's the numbers:
| Browser | Sunspider | V8 | Dromaeo | Slickspeed (Avg) |
| Google Chrome 0.2 | 2185 | 2100 | 347.4 | 233.8 |
| Mozilla Firefox 3.0.1 | 4567.6 | 193 | 1598.6 | 225.2 |
| Apple Safari 3.1.2 | 5448.8 | 165 | 1972.4 | 141.6 |
| Opera 9.52 | 5884.8 | 242 | 1484.8 | 166.8 |
| Mozilla Firefox 2.0.0.16 | 16989.8 | 87 | 7681 | 520.6 |
| Microsoft IE 8.0b2 | 10199 | 45 | 7291.6 | 582.8 |
| Microsoft IE 7.0 | 38608.2 | 793.2 | ||
| Microsoft IE 6.0 | 38525.2 | 32 | 1471 |
Note that the V8 test was the only one to return a "score" (high good) rather than a time-to-execute (low good), so the graph displays max-score, rather than just the raw score. And makes it only more obvious that something weird is going on with the results. The V8 page describes itself as "a suite of pure JavaScript benchmarks that we have used to tune V8". Either they used it too specifically, or they really created/chose the tests specifically to highlight the narrow area that V8 is specifically capable with. Every browser is within a plus-or-minus range of 100 from the average, discarding Chrome, which scores more than 1800 over second place. That's a giant margin.
There's a thousand and one other, probably better, reasons to choose whether or not to use Chrome, especially from among Firefox, Safari, and Opera which are all very close seconds. It's certainly not for me (yet) as there's so many Firefox extensions I'd feel lost without. But I can certainly see it being a wonderful browser for a more casual user.
