Google Chrome - first benchmarks. Summary: wow.

We've got the first comparative benchmarks of Google Chrome, which we downloaded around half an hour ago as the official launch podcast came to an end.

We'll be looking at Chrome very closely over the next few days: modern browsers are complex beasts, and there are a lot of things to test. But JavaScript performance benchmarks we can run right here, right now - so, we have. We took SunSpider 0.9 and ran it twice. Earlier today, in expectation of getting Chrome, we ran the same benchmarks on the same machine for a representative set of other current browsers.

We compared Google Chrome beta 1 running on Vista SP1, on a Lenovo T61 with 1GB of RAM, a 2GHz Core 2 Duo processor and GMA x3000 graphics, comparing it against IE 7, IE 8 beta 2, Opera 9.52, Safari 3.1.2, Firefox 3.0.1 and Firefox 3.1b1. We couldn't run it on Firefox 3.1 with TraceMonkey, because the nightly build failed to complete the benchmarks.

And it is fast, easily topping the chart. In particular, it's more than ten times faster than IE 7, and three times faster than IE8 b2. FF 3.1 comes closest.

Here are the figures.

There is no doubt that Chrome more than lives up to its performance claims: it is the fastest browser we've seen on standard platforms. That may not be too surprising - one of the people behind the V8 JavaScript compiler is Lars Bak, who was the lead developer on the high-performance Sun Java VM that became HotSpot.

We'll flesh out these tests over the coming days, and go into more detail about what they show and what they might mean for Google - and its competitors. Meanwhile, Chrome looks set to more than justify the initial excitement raised in the day between it being announced and becoming available.

[UPDATE]
The latest TraceMonkey vs V8 benchmarks show the Monkey ahead on points. Be sure to read the comments - for their tone, as much as for their technical content.

[UPDATE 2]

We've now got much more comprehensive benchmarks, courtesy of ZDNet Germany. Chrome is maintaining its reputation as an interesting, speedy browser that gives the competiton a serious challenge

Talkback

ekk it's a beast.

Good to see some figures though. I look forward to seeing some more of the results in the next few days.

welshtroll 2 Sep 08 21:34 Reply

Was there any concrete date for the Mac version Rupert ?

utzy 2 Sep 08 23:40 Reply

There was no date given for the Mac and Linux versions. They said they were "working hard" on them, and I believe that - Google does not want to be dependent on Windows.

One of the Mac-using Google people said he'd been running VMware Fusion a lot, so he could run Chrome.

PeterJudge 3 Sep 08 05:15 Reply

You forgot to test it against the recent WebKit Nightly, which has a new JavaScript VM, and is faster than Chrome's V8, according to my tests.

dchest 3 Sep 08 06:39 Reply

The Mac and Linux versions aren't here yet. It's apparently possible to build them yourselves if you like, but there's only so much Charles and I (OK, Charles!) could do in the day we had to prepare for the Chrome launch. It's not quite as simple as you might think to load a lot of browsers into one machine and harvest all the benchmark data. I'll see if Charles can find the time to blog his experiences doing that, but he's got a nervous twitch in one eye now and a murderous gleam in the other.

The same is true of the Webkit nightly. We did spend some time trying to get the FF 3.1 nightly with TraceMonkey going, but it borked out on SunSpider 0.9 - a problem that didn't seem to be reflected in the regression bug reports, so we weren't quite sure what was going on there.

Very interested to hear other people's experiences with benchmarking browsers. It's the glory days of MS-DOS PCs all over again...

Rupert Goodwins 3 Sep 08 08:09 Reply

I downloaded and installed Chrome yesterday evening, shortly after Petersaid that he had it. Everything went very smoothly, but...

I have AVG Internet Security installed, and I don't give blanket internet permission to anything new, I let the firewall ask me for a while so I can get a feel for what is going on. Ever since installed Chrome, I have periodically gotten a request from the "Google Installer" to connect to some IP address, even when I don't have Chrome running! Why in the world is it running something like that in the background all the time, independent of Chrome, what is it trying to do, and how do I stop it?

I don't want to rain on anyone's parade, and I agree that Chrome looks like an interesting new browser, at least in the few minutes that I tried it last night. But I'm not the least bit impressed if they are putting something on my laptop that is poking around, passing on information or whatever, especially when I don't even have Chrome running. I recall that there was a question about whether I was willing to participate in a feedback program, but I assumed that had to do with feedback about Chrome use, and I would NOT expect that it would be running, gathering information and passing it along when I am not using Chrome.

jw 3/9/2008

J.A. Watson 3 Sep 08 12:13 Reply

P.S. As I was writing the above, I realized that I was really not at all comfortable with what Chrome (or Google itself) was doing on my laptop, so I uninstalled it. As is typical when removing something from Google, it popped up a browser window asking for feedback about why I was doing such a thing. I told them, and pressed send. It replied that they would investigate, and they would contact me if they needed more information. Perhaps it was a poor choice of words, or perhaps I am just being paranoid... but there was nothing on that page which told them who I am or how to contact me, so how are they going to manage that feat? Have they already picked up more details about me than I ever wanted them to have? I'm even less pleased now than I was before...

In any case, I will leave it to braver souls than I to continue testing and reporting on Chrome.

jw 3/9/2008

J.A. Watson 3 Sep 08 12:18 Reply

I read Mr. Watson's post and I also had noticed HD activity when the desktop was up. I checked the scheduled tasks and found two google updaters. One was scheduled to run at logon, and the other when the machine was idle for 10 minutes.

ator1940 3 Sep 08 13:55 Reply

Google has said that Chrome regularly (I think every few minutes) downloads the latest list of known hostile websites from the Googleplex.

The best thing about Chrome, and something I find very reassuring, is that because it's entirely open source it is impossible for Google to do something with it that won't be found out in short order. Either it puts it in the source code it releases to the world, in which case anyone who suspects something's up will find out quickly, or it issues a different executable than that which can be built from the public source - equally obvious to those looking.

It's quite in order to be paranoid about what Google gets up to with its servers. Where Chrome is concerned, it feels as if Google is going for maximum openness to attack Microsoft where Microsoft is most vulnerable - its business model. But I do feel that means that Chrome is inherently more trustworthy.

Rupert Goodwins 3 Sep 08 14:05 Reply

Oh, and latest news from those wot know - seems that TraceMonkey's latest nightly has fixed the iloop problem which broke Sunspider. It now runs - and is faster than V8 overall (although V8 still beats it on some individual tests).

Charles has been doing more testing the while, checking how Chrome compares to IE 7 on the smaller platform. He reports that "Numbers for Eee PC 900: SunSpider IE7 = 103775ms, Chrome = 5093ms; Memory footprint IE7 = 61132KB, Chrome = 25304."

Rupert Goodwins 3 Sep 08 14:09 Reply