Windows XP bests OS X in RIA test on Intel A benchmark test for rich internet application (RIA) frameworks claims Apple's OS X lags Microsoft's Windows XP on Intel when rendering HTML, being just over half as fast. Sean Christmann, an experience architect at user interface specialist EffectiveUI, released the GUIMark benchmark following concerns over the lack of a …
Occasionally it's necessary, mostly it's just stupid designers determined to dump as much 'user experience' on the end user as possible, and damn the security, sloth and brittleness implications of the end result.
Not to mention these damn sites never degrade gracefully if you turn scripting/flash/whatever else off.
So, uh, whoever wrote that line knows very little about acid3.
That aside, in those numbers you have a very clear demonstration of why Mac users hate Adobe - 46fps under Flash in Windows vs 8 in OS X. Flash on my MacBook is not demonstrably faster than it was on my ten-year-old tower, which is just embarrassing. No wonder Apple won't allow them to put Flash on the iPhone, eh?
I'm not even going to get started on html rendering benchmarks. Suffice to say every browser manufacturer can tell you that their browser is fastest, measured in their testing suite of choice. That's the suite they make their optimisations for speed against, so it's kind of a stupid game. This test favours IE, apparently; big deal. But just try to tell people IE is the best browser to develop web applications for, and they'll laugh you out of town.
What's kind of cool is that Silverlight has comparable performance cross-platform; and this is the company that took years to make Office Intel-native. Kinda showing Adobe up, methinks.
"tested a range of RIA frameworks on an Intel-based MacBook Pro under Windows XP and Mac OS X 10.5 and found that XP consistently outperformed OS X."
Well that says it all. If they ran XP on an Intel Mac they were using BootCamp, which means XP was using Apple drivers for the hardware. Of course it ran well. Drivers make all the difference and those drivers came from Apple. I also wonder if the graphics card has some native support for DirectX.
Also, XP is an older OS with less overhead than 10.5 Leopard. They should have compared it to 10.4 Tiger.
How about testing Leopard against Vista, instead of a decade old XP. I bet I know the winner.
Vista ME is a dog. Wait for Windows 7. By then Apple will be on 10.8. By the way, a point update or Service Pack is 10.x.x (e.g. 10.5.1) not 10.x. (10.4 or 10.5).
Did they test the responses say a year after the system has been in use? I would hazard a guess that a new install of XP with no additional overheads (antivirus, spyware) would be much faster than an old install with the necessary extras running. We've all seen how all MS OS's need to be reinstalled regularly to maintain any performance.
Although I don't have anything against Apple hardware (or HP, or IBM hardware for that matter... Dell and Compaq pre HP is another story altogether) one thing i DO disslike is a strong notion amongst a certain subset of apple hardware users that they have to defend their choice in hardware with a zest that can only be described as religious.
A 1.6:1 difference in rendering HTML mean feck all to an end user. 60% faster than (or slower than, depending which way you compare) instantanious (to the users perception) is still instantanious.
If the fact that HTML renders inperceptionaly slower on your OS of choice keeps you awake at night and (or causes you to post reasons why this kind of testing is flawed on the register) then you might have some issues that you need to deal with.
And as for comparing XP to OSX 10.5, if you really wanted to, you could always compare XP, XP-SP1, XP-SP2, XP-SP3, VISTA, VISTA-SP1, OS9, UBUNTU8.04 and why not the last version they made of BEOS, but somehow I don't think that that would make you sleep any better at night.
Caveat Emptor: I am also a bit vary of the (smaller but equally fanatical) equailant subset of the microsoft 'fanbase'.
It's considered a known fact by webbies that Opera will outperform all other browsers for rendering HTML/CSS and processing Javascript. Firefox and Safari (on Mac) will also outperform IE7, although only by a fairly minor margin.
Using things like Sliverlight and Flash isn't really up my street but I would assume it largely depends on who wrote the plug-in rather than the browser or platform.
Either way, poorly written pages and scripts cause far more issues with speed than anything else. (I know of a 400KB HTML table on one websites front page and have seen MB's of Javascript source files being loaded as part of a standard template on another.)
There isn't a version of OS X that really matches XP. A hugely simplified but probably not too unfair comparison is that while XP and OS X v10.0 came out at about the same time, XP represented Microsoft pushing the old Windows model as far as it would comfortably go and OS X represented Apple trying to come up with someone more 21st century (and falling over badly until at least v10.2).
There isn't a version of the Mac OS that is similar in features to XP. OS X onwards are already aiming for Vista territory technology. Conversely, OS 9 and earlier are co-operative multitasking, non-memory protected OSs that are probably most similar to something between Windows 3.1 and 95/98.
I've been using OSX for about two years now. I manage a network of 20+ macs with a combination of Mac and Linux servers. At work I have a Quad core Mac Pro with 8 gigs of ram running Leopard. My home machine is an amd dual core with 4 gigs of ram running Vista.
My home machine feels SIGNIFICANTLY faster than my Mac Pro at work. From rendering HTML to launching programs to building software in Eclipse. Everything is faster on the windows box.
I don't believe that it is a question of hardware. I think the article has it right when it speaks of slow Apple API's. Windows XP running in Vmware on a Mac Pro is "snappier" than the native OS.
I'm far from being a "windows fanboi". I work with many operating systems everyday, and tend to prefer Linux for serious work.
Well, I lost all respect for EffectiveUI once I saw they listed eBay as a client. eBay's pages suffer from some of the worst bloat, use of unnecessary script, and just generally poor architecture I've encountered so I wouldn't be proud of that work. I remember the days when I thought their pages were bloated at an average around 35K, now they average more like 100K with NO ADDED USER FACING FUNCTIONALITY! And yes, I'm talking only about the eBay provided framework I realize lusers are responsible for filling *some* of the page in.
Youse guys SUCK in other words, learn to fight the code bloat! Is eBay too stupid to realize their moronic page design has doubled or tripled their bandwidth? And anyone who allows script to run on eBay pages is a bigger maroon than the designers. Their site is totally infested with malicious code.
Let us know when you get past speculating to knowing. If I took everything my CEO and CTO speculated on as gospel I'd be living in a cardboard box under a bridge. : ) No offense meant but . . most of my customers care about not only speed if data back to them but the speed of data to the user. If there's this huge of a difference with rendering in Safari verses IE on these rich internet application FWs than IE wouldn't have the reputation as a speed pig. Not that it couldn't be somewhere in Webkit, I guess I'm just a touch dubious :D
If, as is generally agreed most of Rich Internet Applications (RIA) are being developed with cross-platform tools and indeed on UNIX/LINUX, why then would not the team include GNU/Linux in their comparison?
It is unfortunate that because of the limited knowledge of and experience with GNU/Linux of most technology "writers", a fair anf worthwhile comparison in technologies is missed.
W. Anderson
wanderson@nac.net
On all platforms, Opera's best. I'm often surprised how slow other browsers are, and it's not just because of popup and other advertising blocking, much though those are reason enough to adopt the Norwegian browser. Microsoft should buy the company, though I sincerely hope they don't.
-A.
Personally I couldn't give a rat's arse about the FPS that Flash can be rendered at. By the time it would be there to be rendered, I've useually already buggered off somewhere else. It's the Intertube in the middle that the meaningless crap that these fsckwit muppets vomit up is already too crufty for.
I do find the fact that there's a company specialising in this arty-farty cobblers calling itself "EffectiveUI" funny in an ironic sort of way though.
The horned one, 'cos I don't want anyone to think that this rant is due to any Apple sympathies that I might be harbouring.
...if they had compared MacOS X 10.5 against Vista, both running Firefox 3 beta as well as the same versions of the plug-ins used, I'd bite.
As it were, they were running 10.5 vs. XP, and Safari vs. MS Internet Exploder and in part different releases of the plug-ins.
No bite from me. I want to know how Debian did.
presumably one of the reasons IE is so much faster is that - at least as regards HTML - it disnae have to waste time with such trivialities as actually rendering the stuff in line with any web standards out there.
as for the other stuff - flash has always run like shit on OSX and java on OSX is a joke. so the final scores dinnae exactly surprise me.
oh yeah - and i wouldn't trust leopard to find its own arse in the dark. i suspect the results might have been slightly different if the benchmarks had used a version of OSX that actually works, like tiger.
what's that webster?... your round? thanks. i'll have a pint of kool-aid please!
It's an interesting test, certainly. It does a number of things which nobody sane would do in an actual on-page script (like updating the DOM to show the current framerate every 17 microseconds, which is a fairly expensive operation in itself, and pushing an element into an array when it does).
It's the least useful kind of test, because it doesn't actually show where the bottleneck is: is it that DOM updates in Safari are synchronous but async in IE (which is why Safari can often seem "snappier" than IE)? Is it that array manipulation is badly optimised in Safari's JS engine? Is it that creating and destroying a new instance of the Date object hundreds of times a second is slow in Safari's JS engine but quick in IE?
Sean Christmann from EffectiveUI created this test to help RIA developers make decisions on RIA frameworks -
I am frequently asked about performance of the flash player, cross OS. These tests give me but a small piece of a much bigger picture. DATA = IQ - Any executive that makes a technology decision based solely on this single benchmark has been promoted way too quickly.
One last thing that I can't let go - Eddie Johnson, EffectiveUI did not have anything to do with the eBay website, they built eBay Desktop (desktop.ebay.com) - you should do your homework before flaming a team of great people...
A test of internet (never reliably consistent speeds) applications (by definition not the OS or the Hardware) comparison... It's about as useful as comparing Super Unleaded petrol in a Yugo versus Regular unleaded in a Ferrari... Fun to watch, but the results aren't necessarily what they seem. This study was an absolutely worthless expenditure of money.
Knows instinctively that Mac OS X APIs are bloated...
I use Winbloze and Mac OS X and have for the past 7 years been working routinely on both platforms. I won't touch Vista and I won't touch Leopard. Neither are yet ready for prime time.
Now, when I can be rendering a file (say an h.264 version of a movie just edited in iMovie or Final Cut Pro) and have the render time increase by many minutes simply by daring to just open Safari (without actually rendering a web page, mind) there's a problem!
But similarly, when I can transfer multi-gigabyte files over a gigabit network from an iMac to a PowerMac at 70MB/s transfer rates, and yet with comparable hardware on ANY version of Winbloze I can only muster just over 30MB/s, there's a problem, there too!
If anything, these tests, however flawed, can only help point out weak areas of our beloved OS (whichever it may be) and if it helps make one aspect better, that's serving a purpose in my books.
So, um, out of interest, how much time does the average user spend waiting for HTML to render (as opposed to web servers processing requests and networks relaying the results?)
Hardware and software are moving on, and the "tit for tat" wars will continue but really, is this compelling?
In fairness -
(and avoiding all the 'tard slinging and childish "all you fanbois" stuff which never fails to make rtegister threads read like 6th form toilet walls... honestly why not just write "...is gay" after everyone's post that doesn't agree with you?)
As a UI designer who has always argued against RIAs and superfluous shite (and often lost money for doing so) I'm glad that at least one or two people in here are looking at the real picture.
I don't like high-bandwidth-for-no-reason crapola. I see RIAs as the chelsea tractors of the internet: keeping maintenance contracts alive and driving the economy, sure... but f**king things up for everyone else.
Designers have completely forgotten our place in the industry, thanks to the fact that most companies are trying to sell to 12 year olds with epilepsy inducing tat. But you know what?
Bad project managers who can't say no to a client have their share of the blame, too. Clients who don't understand the medium they are paying £50K to use can also take a tiny portion of the blame... but not really, cos it's not their job and that's why they pay £50K
I've never seen a so-called RIA that couldn't be duplicated by clever redesign of very low-level, low bandwidth technologies.
So yeah - just a point in defence of a minority of people who call ourselves "graphic designers" or "UI designers" - we are honestly not all obsessed by flashing crap. Some of us do actually believe that "design" means subtracting, not adding.
I can only imagine the disasters that would result if most web designers attempted anything real, like engineering or architecture... "sure, you'll need to move the train station, build another reservoir, and the traffic coming and going to the supermarket will submerge the entire town in three months... but look! Big lights! Shiny! Doors open-close! Swish-swish!
This is not "design", it's an excuse to save up for a stupid haircut and a plastic figurine of money mark. There are some designers left in the world.
Now, let the win v mac thinly concealed homoerotica continue, sluggers ;-)
This was a meaningless test.
When performance profiling different platforms, you have to run the same code against them. Since you can't guarantee that even both versions of Opera are the same, the appropriate course of action would have been to write a basic HTML renderer (using each platform's APIs when appropriate) and taken progressive timing measurements throughout the various stages of rendering a page.
This experiment doesn't tell you anything. For all we know the greatest performance loss on the Mac side was caused by some over-cautious thread locks in Safari. Or some carelessly unsafe threading in IE. Who knows?
This report is comparing apples and pears. (No pun intended).
I'd like to know some real figures for my own personal interest, but until someone qualified performs a real test I won't give this kind of information any credence.
Just did the HTML testy-thingy on me old iMac Core 1 Duo 2.0 GHz RadeonX1600, OS X 10.4.11
Opera 9.26 : 13 fps
FireFox 2.0.0.12 : 7-8 fps
Safari 3.1.1 : 13 fps
And on me Core 2 Duo 2.4 GHz Radeon HD2400 Pro XP sp3 (32bit OS)
IE 6 : 20 fps
Firefox 2.0.0.14 : 20 fps
Safari 3.1 : erratic between 14-24 centered around 17 fps
But on the same PC running UBUNTU 7 (64bit OS)
Firefox 2.0.0.12 : 10 fps
So that's a fact, maintaining UI consistency across different fields displaying the same Object comes at an overhead in COCOA and a F5 key stroke on Windoze ...
But is this test meaningful? Looking at Ubuntu, I'd be tempted to say no.
Anybody with set of figures on a G5 ???
>Would have been more of an ass kick, vista is much much faster >than XP at rendering to screen and takes even more advantage of >hardware than XP.
Um, NOT! - Vista on my girlfriends brand new laptop takes so long to do anything you can actually go make a coffee while you wait for it to finish almost any mundane task.
Oh the Apple front - I can report that my new Mac Pro (dual quad-core 3ghz Xeon with Nvidia 8500 512Mb card) - feels a whole lot slower in general than I thought a machine of this spec running a Unix based OS would be ... It seems that MacOS-X doing the same as wasteful bullshit as Vista - it loads heaps of crap in the background, consuming all spare RAM, making the machine do lots of unnecessary swapping.
(Oh I forgot - you're supposed to kit these things out with 32gigs of Steve'os stupidly expensive ECC memory!)
Just for the sake of a non-argument I unearthed me Clamshell G3 366, OS X 10.3.9
Safari 1.3.2 : 2.5 fps
That shows a linear scalability 13 fps / 2.0 GHz * 366 MHz = 2.379 fps (theoretical)
Complete b0ll*cks of course ... but cool 'ey!? IE users are now reasured, they are using the fastest browser.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
