<quote>For all of the positive aspects we've mentioned above, what good are they if we go mad trying to use the application?</quote>
As a long-time windows user (haven't had an Apple computer since the IIe), I was put-off enough by the font smoothing and other oddities in Safari to uninstall it right away and stick with Firefox.
But I can see Apple's intentions here... if they're trying to get people to switch to Macintosh for their next computer, perhaps those people will prefer the look and thus be happy to get a Mac - not surprised when Safari renders significantly differently on Windows vs Mac.
I'm having a little trouble with the values you guys are getting on how much each app is using for ram. Under windows xp pro, safari on average uses about twice as much ram as firefox when multiple windows were open. In cases where firefox was using about 45MB of ram, I noticed that safari was edging onto 100MB. You guys should redo that test, and do it while having several tabs open. Also, it was never mentioned how much flash/java/etc. the test site used, so theres no way of knowing how effecient the browsers really are in real life settings.
The test was each of our test pages(AT, ./, CNN, and the WH) open in separate tabs; this gives a good mix of flash-rich pages(AT and CNN) and simpler text-rich pages(./, WH).
I'll stick with Opera as it is not only more standards-compliant than Firefox or IE (it passes the ACID2 test perfectly), but also much less prone to security breaches thanks in part to using proprietary code (rather than IE and Mozilla-related projects which re-use many modules), and also because Opera has such a small userbase (~2.5% I believe) that it just isn't worth the effort of writing totally new code to attack its users.
Opera is fast, functional and safe. I do have Firefox as well as IE7 installed as well, but I prefer Opera for day-to-day browsing.
FWIW: I have problems with IE and Java under Vista as well... The JRE isn't too happy when DEP mode is enabled. :( (I have no clue if it will help to disable DEP for Safari, but give it a spin if you have the time)
Oh, I forgot to add: Java sucks. (I have to admit reinstalling JRE many, many times before I discovered the DEP issue -- I just couldn't imagine a modern piece of software not supporting DEP, the API has been there since before 1993 for Pete's sake!)
This is problematic for Apple. Although we have other theories on Safari that we'll get to in a moment, we're not ready to be so bold as to proclaim that Apple doesn't intend for this browser to be used on Windows by the masses - if that was the case they wouldn't have made it WWDC's "one more thing" or have giving it such prominent billing on their website.
That WWDC was rather dull and boring, they needed some flashy announcement and had this in the drawer - call it plan B for rainy days. It created alot of hoopla but in the end, its nothing to call home about. On the useful side are the points you mentioned.
What exactly is the advantage of having several GB of RAM not being used by any application? A well designed program will use RAM as cache (for pages, interface elements, bookmarks, history, etc.), to speed up its operation.
One thing is checking how much RAM needs to be available for a program to run (or how well it runs when very little RAM is available), but simply checking how much RAM a program is using (when there's still a lot of free memory) tells you nothing about its efficiency.
I'd much rather use a program that's smart enough to load things I might want to use in the background (if enough RAM is available) than one that loads things only when I specifically ask for them, making it slower in exchange for not using RAM that's just sitting there.
Also, site "loading" times depend a lot more on the network conditions than on the browser, and the same browser will "load" the same site in different times, if you test it multiple times. What you might want to measure is how long it takes each browser to _render_ a complex page. Make sure everything is cached, hit "refresh" and time it. Or use the "back" and "forward" buttons. That way you take network delays out of the equation, and check how responsive the browser actually is.
As for the RAM, you're right to a certain point. There are far too many people out there who like to stare at their unused RAM and feel proud that their system is so slim, sleek and... slow, because everything has to be loaded from the slow disk. RAM is here to be used, not to be left unused. End of story.
But reality often isnt that simple. When software starts to unnecessarily hog RAM, resulting in disk trashing when its needed elsewhere, then its time to have a closer look and criticise.
Some older FF builds in the 1.5 days were like that. Have FF run for a full day and it would start to bloat up to several hundred megabytes thanks to caching too many previous visited sites in RAM. Manually reducing that number to a reasonable value trimmed down the RAM usage. Personally, I dont see that behavior in FF 2.x anymore.
Like I said, simply checking how much RAM a program is using tells you nothing about its _efficiency_. Most browsers let you choose how much RAM they can use for cache. If it doesn't respect that value, it's not "bloated"; it's buggy.
For example, something that Opera has done for a long time and other browsers have only recently copied is this: cache the _rendered_ pages in the browser history. That way when you press the "back" or "forward" button, the page is there _instantly_, regardless of complexity. This makes a huge difference in the user experience and is a good use of 4 MB of RAM. The latest versions of Firefox do this, too. Not sure about MSIE.
"This makes a huge difference in the user experience and is a good use of 4 MB of RAM."
I agree, as long as it releases ram it's not likely to use anytime soon.
If I've been browsing for a while and then load up a game to play, but keep the browser open, then I definitely don't need 400 MB of cached renderings.
Hopefully it would cache the rendered pages for the past few sites, but write to disk the render cache for older sites in my session to free up the ram.
The most interesting thing to me out of the whole article is the mrmory footprint chart. I never realized IE used so much memory as I don't check the usage.
I downloaded Safari and it works quite nicely for me. I don't see one browser being vastly superior to another under normal usage though. I'm not so picky.
I saw that and was able to read the first page before it was removed.
What were the "unusual circumstances" for not posting it?
PS: AnandTech is a favorite read for me because of the detail that is usually spent on a topic/item but I found this review to be rudimentary. I would have like to see other speed tests done with heavy JS and Flash pages, a comprehensive overview of how WebKit differs from IE and Firefox engines, a detailed report of how Safari imported frameworks for OS X and not just a mention of the visual differences, and a comparison of HTML5 and CSS3 Open Standards used between the browsers. One in particular that Safari has is the ability to resize text boxes.
I use portable firefox everywhere, and could see the usefulness of having a portable safari on my flashdrive. Is there any chance this is portable, or could be easily be made such?
Just to confirm this, even tho ff supports pipelining it is indeed disabled by default. All the other browsers have it enabled as far as I know, which at least explains part of the loading issue.
As to why it is disabled by default, the only reason I can remember were some dodgy webservers not supporting the feature properly.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
28 Comments
Back to Article
crimson117 - Thursday, July 12, 2007 - link
<quote>For all of the positive aspects we've mentioned above, what good are they if we go mad trying to use the application?</quote>As a long-time windows user (haven't had an Apple computer since the IIe), I was put-off enough by the font smoothing and other oddities in Safari to uninstall it right away and stick with Firefox.
But I can see Apple's intentions here... if they're trying to get people to switch to Macintosh for their next computer, perhaps those people will prefer the look and thus be happy to get a Mac - not surprised when Safari renders significantly differently on Windows vs Mac.
zshift - Sunday, July 8, 2007 - link
I'm having a little trouble with the values you guys are getting on how much each app is using for ram. Under windows xp pro, safari on average uses about twice as much ram as firefox when multiple windows were open. In cases where firefox was using about 45MB of ram, I noticed that safari was edging onto 100MB. You guys should redo that test, and do it while having several tabs open. Also, it was never mentioned how much flash/java/etc. the test site used, so theres no way of knowing how effecient the browsers really are in real life settings.Ryan Smith - Sunday, July 8, 2007 - link
The test was each of our test pages(AT, ./, CNN, and the WH) open in separate tabs; this gives a good mix of flash-rich pages(AT and CNN) and simpler text-rich pages(./, WH).jay401 - Saturday, July 7, 2007 - link
Between Extensions and its more correct CSS rendering compared to IE, FireFox still brings the best browser experience to Windows.PrinceGaz - Sunday, July 8, 2007 - link
I'll stick with Opera as it is not only more standards-compliant than Firefox or IE (it passes the ACID2 test perfectly), but also much less prone to security breaches thanks in part to using proprietary code (rather than IE and Mozilla-related projects which re-use many modules), and also because Opera has such a small userbase (~2.5% I believe) that it just isn't worth the effort of writing totally new code to attack its users.Opera is fast, functional and safe. I do have Firefox as well as IE7 installed as well, but I prefer Opera for day-to-day browsing.
LTG - Saturday, July 7, 2007 - link
It seems the most basic Java applet does not work:http://demo.aurigma.com/ImageUploader40/BasicDemo/...">http://demo.aurigma.com/ImageUploader40/BasicDemo/...
I'm running Vista and Safari 3.0.2.
Does this work for anyone else?
BikeDude - Sunday, July 8, 2007 - link
FWIW: I have problems with IE and Java under Vista as well... The JRE isn't too happy when DEP mode is enabled. :( (I have no clue if it will help to disable DEP for Safari, but give it a spin if you have the time)Oh, I forgot to add: Java sucks. (I have to admit reinstalling JRE many, many times before I discovered the DEP issue -- I just couldn't imagine a modern piece of software not supporting DEP, the API has been there since before 1993 for Pete's sake!)
--
Rune
Griswold - Saturday, July 7, 2007 - link
This is problematic for Apple. Although we have other theories on Safari that we'll get to in a moment, we're not ready to be so bold as to proclaim that Apple doesn't intend for this browser to be used on Windows by the masses - if that was the case they wouldn't have made it WWDC's "one more thing" or have giving it such prominent billing on their website.That WWDC was rather dull and boring, they needed some flashy announcement and had this in the drawer - call it plan B for rainy days. It created alot of hoopla but in the end, its nothing to call home about. On the useful side are the points you mentioned.
Donkeyshins - Saturday, July 7, 2007 - link
...the nine zero-day exploits against Safari. Not an auspicious beginning.Justin Case - Friday, July 6, 2007 - link
Why is using 44 MB "better" than using 50 or 200?What exactly is the advantage of having several GB of RAM not being used by any application? A well designed program will use RAM as cache (for pages, interface elements, bookmarks, history, etc.), to speed up its operation.
One thing is checking how much RAM needs to be available for a program to run (or how well it runs when very little RAM is available), but simply checking how much RAM a program is using (when there's still a lot of free memory) tells you nothing about its efficiency.
I'd much rather use a program that's smart enough to load things I might want to use in the background (if enough RAM is available) than one that loads things only when I specifically ask for them, making it slower in exchange for not using RAM that's just sitting there.
Also, site "loading" times depend a lot more on the network conditions than on the browser, and the same browser will "load" the same site in different times, if you test it multiple times. What you might want to measure is how long it takes each browser to _render_ a complex page. Make sure everything is cached, hit "refresh" and time it. Or use the "back" and "forward" buttons. That way you take network delays out of the equation, and check how responsive the browser actually is.
Griswold - Saturday, July 7, 2007 - link
As for the RAM, you're right to a certain point. There are far too many people out there who like to stare at their unused RAM and feel proud that their system is so slim, sleek and... slow, because everything has to be loaded from the slow disk. RAM is here to be used, not to be left unused. End of story.But reality often isnt that simple. When software starts to unnecessarily hog RAM, resulting in disk trashing when its needed elsewhere, then its time to have a closer look and criticise.
Some older FF builds in the 1.5 days were like that. Have FF run for a full day and it would start to bloat up to several hundred megabytes thanks to caching too many previous visited sites in RAM. Manually reducing that number to a reasonable value trimmed down the RAM usage. Personally, I dont see that behavior in FF 2.x anymore.
Justin Case - Sunday, July 8, 2007 - link
Like I said, simply checking how much RAM a program is using tells you nothing about its _efficiency_. Most browsers let you choose how much RAM they can use for cache. If it doesn't respect that value, it's not "bloated"; it's buggy.For example, something that Opera has done for a long time and other browsers have only recently copied is this: cache the _rendered_ pages in the browser history. That way when you press the "back" or "forward" button, the page is there _instantly_, regardless of complexity. This makes a huge difference in the user experience and is a good use of 4 MB of RAM. The latest versions of Firefox do this, too. Not sure about MSIE.
crimson117 - Thursday, July 12, 2007 - link
"This makes a huge difference in the user experience and is a good use of 4 MB of RAM."I agree, as long as it releases ram it's not likely to use anytime soon.
If I've been browsing for a while and then load up a game to play, but keep the browser open, then I definitely don't need 400 MB of cached renderings.
Hopefully it would cache the rendered pages for the past few sites, but write to disk the render cache for older sites in my session to free up the ram.
Fede777 - Friday, July 6, 2007 - link
In the Subjective Testing page it says but it should sayFede777 - Friday, July 6, 2007 - link
My mistake, it's just that iTunes looks like safari on XPcmdrdredd - Friday, July 6, 2007 - link
The most interesting thing to me out of the whole article is the mrmory footprint chart. I never realized IE used so much memory as I don't check the usage.I downloaded Safari and it works quite nicely for me. I don't see one browser being vastly superior to another under normal usage though. I'm not so picky.
Nemokrad - Friday, July 6, 2007 - link
Safari 3.0.2 was released two weeks ago today. How come you didn't use it?Ryan Smith - Friday, July 6, 2007 - link
This article has been complete for longer than two weeks. We had some unusual circumstances that kept us from publishing it on time.solipsism - Friday, July 6, 2007 - link
I saw that and was able to read the first page before it was removed.What were the "unusual circumstances" for not posting it?
PS: AnandTech is a favorite read for me because of the detail that is usually spent on a topic/item but I found this review to be rudimentary. I would have like to see other speed tests done with heavy JS and Flash pages, a comprehensive overview of how WebKit differs from IE and Firefox engines, a detailed report of how Safari imported frameworks for OS X and not just a mention of the visual differences, and a comparison of HTML5 and CSS3 Open Standards used between the browsers. One in particular that Safari has is the ability to resize text boxes.
AnnonymousCoward - Sunday, July 8, 2007 - link
I don't need that kind of detailed analysis. Since the interface sucks and the window behavior is like a stubborn Mac window, the deal is off.sprockkets - Friday, July 6, 2007 - link
I think I'll stick with Konqueror on Linux, oh wait...Googer - Sunday, July 8, 2007 - link
Opera and Safari are the only two that get a passing grade with ACID 2.Konqueror cannot pass the ACID 2 browser test. And that's not the only one IE, Mozilla, and Konqueror cannot pass. There are others.
http://www.webstandards.org/">http://www.webstandards.org/
Googer - Sunday, July 8, 2007 - link
Also according to Extremetech, Safari had problems with Citibank.comhttp://www.extremetech.com/article2/0,1697,2152775...">http://www.extremetech.com/article2/0,1697,2152775...
Xenoterranos - Friday, July 6, 2007 - link
I use portable firefox everywhere, and could see the usefulness of having a portable safari on my flashdrive. Is there any chance this is portable, or could be easily be made such?John Haller, I'm looking at you.
SilthDraeth - Friday, July 6, 2007 - link
I know in Firefox you can typeabout:config
and then you can turn on pipelining etc, and it greatly speeds up rendering and loading of websites by enabling more simultaneous pipes.
Spoelie - Friday, July 6, 2007 - link
Just to confirm this, even tho ff supports pipelining it is indeed disabled by default. All the other browsers have it enabled as far as I know, which at least explains part of the loading issue.As to why it is disabled by default, the only reason I can remember were some dodgy webservers not supporting the feature properly.
Ryan Smith - Friday, July 6, 2007 - link
Yes, it was all done with default settings.crimson117 - Thursday, July 12, 2007 - link
I appreciate keeping things at default settings, but I'd be very interested to see Firefox's load times with the one simple pipeline tweak enabled.