After being tired of high OmniWeb memory usage I decided to see if this is a necessary evil in today's heavy-web age. I'm running Tiger and tried to simulate a typical browsing session on several browsers: open 10 tabs to various pages, then open /close several more to NY times, wikipedia, google, etc., ending with the initial 10 tabs open. Here's what I found ("Real memory" in Activity Monitor a couple of minutes after Hiding the app, on a system with ample free memory and no swapping going on):
Immediately after loading first 10 tabs
-------------------------
Code:
Camino 2.0b 66 MB
Safari 3.1 69 MB
Opera 9.6 74 MB
Firefox 3.0 78 MB
Firefox 2.0 85 MB
Camino 1.6 86 MB
OmniWeb 5.8 110 MB
After additionally opening/closing some NYTimes/wikipedia/google pages
--------------------------------------------
Code:
Opera 9.6 78
Camino 2.0b 82
Safari 3.1 128
OmniWeb 5 132
I gave up the test for some of the browsers after the first 10-tab opening. However, even the second test is not sufficiently realistic.
In my experience OmniWeb and Safari both tend to grow to 500+ MB after a day or two of continuous use, and I haven't tested the others yet under these conditions. However, if we assume longer term performance can be extrapolated, what I THINK can be concluded here is:
- WebKit-based browsers have terrible memory mgmt
- Gecko 1.9 made a major advance over 1.8 in this area
- Opera's superiority demonstrates proprietary technology underperforms open source in this area
I'd like to keep using OmniWeb for its best-in-class interface, but with these memory numbers it doesn't cut it on a laptop with 2GB that needs to run other things such as Parallels. Taking up 25-50% of available RAM is not acceptable for any application, not even a web browser.
I don't know if OmniWeb is able to make any improvements due to what looks like a poor base in WebKit. Perhaps it can make explicit efforts to policing the latter's memory usage and get closer to Opera and the new Gecko-based browsers. I hope so.