Windows 8 CP's Internet Explorer 10 benchmarked

Wondering which browser out of those available for Windows 8 is the fastest? Look no further than our benchmark comparison betwewn Internet Explorer 10 (installed by default) and Chrome 17, Firefox 10 and Opera 11.60. All benchmarks were conducted on a freshly installed browser (all settings default) with just one tab running; all were run on the Windows 8 Consumer Preview build 8250 on the same system with no background applications open.

Sunspider (in ms, lower is better)

As most people know, Sunspider tests the JavaScript performance of a browser

HTML5 Test (out of 475)

The HTML5 test awards the browser a score out of 475 that reflects how much support for HTML5 the browser has

Peacekeeper (higher is better)

This benchmark shows the all-round capabilities of the browser

BetaFishIE (in FPS, higher is better)

This benchmark is made by Microsoft to show the CSS3 animation capabilities of Internet Explorer 10. It worked with all browsers except Opera 11.60

Fishbowl at 2000 fish (in FPS, higher is better)

This benchmark is made by Microsoft to show the HTML5 animation capabilities of Internet Explorer 10. Take the results with a grain of salt though because Microsoft clearly optimzed this benchmark to work well on IE10.

We would also have run a WebGL benchmark but WebGL is not supported in Internet Explorer 10 natively. If we do more benchmarks of these browsers we'll post them here, so check back occasionally for updates.

Report a problem with article
Previous Story

Windows 8 woes got you down? We have your support forum ready

Next Story

Windows 8 integrates Xbox games like Pinball FX2

58 Comments

Commenting is disabled on this article.

BumbleBritches57 said,
NO WAY IN HELL IE is faster than Chrome.
As much as your protestations are valued, could you point out how? IE9 is well known for having pretty excellent hardware acceleration.

Kirkburn said,
As much as your protestations are valued, could you point out how? IE9 is well known for having pretty excellent hardware acceleration.

IE haters, you know.

Kirkburn said,
As much as your protestations are valued, could you point out how? IE9 is well known for having pretty excellent hardware acceleration.

IE haters, you know.

How hard do security professionals need to scream:
"WebGL is NOT a standard, has NOTHING to do with HTML5, and is highly dangerous!"

At the end of the article you 'specifically' say you didn't run WebGL tests; however, this is a lie.

The HTML5 Test DOES test for WebGL features, and marks down browsers for not supporting them.

(It also rates on several other things that have NOTHING to do with HTML5. Maybe you should consider NOT using that site as a reference, or at least ask them to remove the non-HTMl5 items from their 'so-called' HTML5 Test!


Another thing you fail to mention...
Sunspider is 'javascript benchmark, for 'WEBKIT' based engines. Surprisingly IE10 does well on the test, but considering that many of the items in the Sunspider test are 'irrelevant' to how IE10 works, it is more amusing than indicative of anything important.

thenetavenger said,
Sunspider is 'javascript benchmark, for 'WEBKIT' based engines.
It's not *for* WebKit, it's *by* WebKit. It's still valid for other browsers, just might be a bit biased to WebKit's workings.

BetaFish seems doesn't work in Opera 12, yes, but I have 60 fps in Fishbown with up to 3500 fishes. Drops to 59fps above. Chrome 18 shows 41 fps at 2000 fishes. And just about 0,5-2 fps in Firefox 11 at 2000 fishes.

coth said,
BetaFish seems doesn't work in Opera 12, yes, but I have 60 fps in Fishbown with up to 3500 fishes. Drops to 59fps above. Chrome 18 shows 41 fps at 2000 fishes. And just about 0,5-2 fps in Firefox 11 at 2000 fishes.

hmm missing edit icon...
43 fps in Firefox 11, hardware acceleration was set to disable

The HTML5 graph doesn't start at 0 and instead starts at some bogus random value, therefore the graph doesn't fulfil its main purpose which is to visually and quickly assess the extent of the differences between the values. These differences are completely skewed visually, 370 looks like it's twice as much as 330. Listing the raw values in a sorted table would actually have given just as much information and been less misleading.

On my installation IE10 scored 137ms total on sunspider. Again this is down to system spec, good drivers, etc

Riva said,
On my installation IE10 scored 137ms total on sunspider. Again this is down to system spec, good drivers, etc

Yep and that's exactly why I did all these benchmarks on my system

The HTML5 test awards the browser a score out of 475 that reflects how much support for HTML5 the browser has

Seriously, there are still people who rely on HTML5test to measure the HTML5 support of a web browser?

HTML5 support just asks the browser whether it does implement a feature. I does NOT verify if it is well implemented, and compliant with the W3C standards.

Chrome and Firefox are racing to implement tons of half baked HTML5 features that are NOT standard compliant, and will break websites rendering in future versions.

Look at the W3C's HTML5 test case to see the real quality of the html5 implementation in IE10 vs the other browsers.

http://samples.msdn.microsoft.com/ietestcenter/

what is the point of implementing tons of web standards is they are not reliable and have a 10% implementation fault rate? Do the webmasters miss the days of IE6 and want firefox/chrome to provide them more headaches?

read this if you want to know what you're going to face is you actually use HTML5 features (currently most self proclaimed "HTML5 based" websites only uses HTML4 features and a little CSS3):

http://blog.millermedeiros.com/ipad-is-the-new-ie6/

link8506 said,

Seriously, there are still people who rely on HTML5test to measure the HTML5 support of a web browser?

HTML5 support just asks the browser whether it does implement a feature. I does NOT verify if it is well implemented, and compliant with the W3C standards.


It's currently as good as it will get if we're interested in testing for HTML5 features using a benchmark, and the HTML5Test suite is continuously refined to make as accurate tests as it can. Yes, vendors are potentially able to "cheat" it but the only thing they'll gain from that other than a number is poor publicity.

Speaking of inaccurate tests, Peacekeeper is said to have issues where it depends on timer implementations on the tested web browser, details that are actually not impacting browser speed.

I personally think Kraken is a modern test suite that has few known flaws, which is seeing too little attention. Sunspider was developed in a time and age that doesn't match that of today, and is also quite irrelevant these days. This is evident not the least because modern browsers are fighting over milliseconds in that test, and also that the test itself doesn't test for bottlenecks in modern web technologies. Kraken was developed for this reason.

Northgrove said,

It's currently as good as it will get if we're interested in testing for HTML5 features using a benchmark, and the HTML5Test suite is continuously refined to make as accurate tests as it can.

Wrong. this test tests fetures that aren't part of the HTML5 specification. And the author doesn't want to fix that.

Northgrove said,

It's currently as good as it will get if we're interested in testing for HTML5 features using a benchmark, and the HTML5Test suite is continuously refined to make as accurate tests as it can. Yes, vendors are potentially able to "cheat" it but the only thing they'll gain from that other than a number is poor publicity.

Speaking of inaccurate tests, Peacekeeper is said to have issues where it depends on timer implementations on the tested web browser, details that are actually not impacting browser speed.

I personally think Kraken is a modern test suite that has few known flaws, which is seeing too little attention. Sunspider was developed in a time and age that doesn't match that of today, and is also quite irrelevant these days. This is evident not the least because modern browsers are fighting over milliseconds in that test, and also that the test itself doesn't test for bottlenecks in modern web technologies. Kraken was developed for this reason.

Actually you are wrong... It is purporting to be an 'HTML5' test...

Yet it scores for things that are NOT HTML5 related or even 'proposed' to the W3C standards.

One example, WebGL is NOT HTML5, and is highly dangerous, yet it gives 'points' for it.

It is a worthless arbitrary test that the sitecreators use to justify their love for Chrome.

link8506 said,

Seriously, there are still people who rely on HTML5test to measure the HTML5 support of a web browser?

HTML5 support just asks the browser whether it does implement a feature. I does NOT verify if it is well implemented, and compliant with the W3C standards.

Chrome and Firefox are racing to implement tons of half baked HTML5 features that are NOT standard compliant, and will break websites rendering in future versions.

Look at the W3C's HTML5 test case to see the real quality of the html5 implementation in IE10 vs the other browsers.

http://samples.msdn.microsoft.com/ietestcenter/

what is the point of implementing tons of web standards is they are not reliable and have a 10% implementation fault rate? Do the webmasters miss the days of IE6 and want firefox/chrome to provide them more headaches?

read this if you want to know what you're going to face is you actually use HTML5 features (currently most self proclaimed "HTML5 based" websites only uses HTML4 features and a little CSS3):

http://blog.millermedeiros.com/ipad-is-the-new-ie6/

I'll take those results with a grain of salt they don't pit current browsers against one another ie: IE10 vs Opera 11.61 (should have been Opera Next 12.xx)

Sub_Zero_Alchemist said,
Firefox 10 has a higher score than what is shown,that score is outdated and misleading the actually score for FF10 is actually 332 with bonus of 9.

Personally I also think a preview browser should have been compared to preview browsers but this is perhaps nitpicking (I'm not sure how much the scores would change when compared to Chrome 19 and Firefox 13).

Phalanger said,

That's a bit harsh when windows 8 is not a supported system.

Shouldn't Mozilla need to do their homework to update Firefox to fully support Windows 8?

mahara said,

Shouldn't Mozilla need to do their homework to update Firefox to fully support Windows 8?


Not when MS does not. Likewise I don't see IE10 on Windows 7/Vista yet.

Phalanger said,

Not when MS does not. Likewise I don't see IE10 on Windows 7/Vista yet.

So Mizilla's motto is "being at least as bad as MS" ?
BTW, MS updated IE to fully support Windows 8. Your comment is obviously invalid.

RealFduch said,

So Mizilla's motto is "being at least as bad as MS" ?
BTW, MS updated IE to fully support Windows 8. Your comment is obviously invalid.

You obviously don't get the fact that even Microsoft does not support Windows 8 or IE 10 yet.

Chrome 19.0.1041.0 does:
30 - 31 fps in fishbowl.
171ms in sunspider
377+13 in html5

Just for those interested.
Glad to see it's not just the number growing, but actually the browser improving with all those beta/dev updates!

Praeses said,
Chrome 19.0.1041.0 does:
30 - 31 fps in fishbowl.
171ms in sunspider
377+13 in html5

Just for those interested.
Glad to see it's not just the number growing, but actually the browser improving with all those beta/dev updates!

so, you really believe that this information is useful without providing your screen resolution (more pixels to draw = a lower framerate) and your hardware configuration?

cross browser performance need to be compared on the same configuration. Otherwise, raw data is just meaningless.

Praeses said,
Chrome 19.0.1041.0 does:
30 - 31 fps in fishbowl.
171ms in sunspider
377+13 in html5

Just for those interested.
Glad to see it's not just the number growing, but actually the browser improving with all those beta/dev updates!


Useless.
So on my machine IE10 does 1000FPS, 1ms and 400+200. It means nothing unless you compare the results on the same machine.

When clicking the tweet to see this article, I seriously knew it would be shown in a positive light. Why? ...because this is Neowin, talking about anything Microsoft. I like MS, don't get me wrong, but the articles here are nearly always showing MS in a positive light.

Jaybonaut said,
When clicking the tweet to see this article, I seriously knew it would be shown in a positive light. Why? ...because this is Neowin, talking about anything Microsoft. I like MS, don't get me wrong, but the articles here are nearly always showing MS in a positive light.

You do realize that IE10 is shown losing to Chrome in the HTML5 Test and Peacekeeper...?

Jaybonaut said,
When clicking the tweet to see this article, I seriously knew it would be shown in a positive light. Why? ...because this is Neowin, talking about anything Microsoft. I like MS, don't get me wrong, but the articles here are nearly always showing MS in a positive light.

Wrong. I nearly dropped Neowin couple of months ago since it became too much anti-MS biased (And I'm not even talking about comments). Even Engadget is much more balanced now.

Scorpus said,

You do realize that IE10 is shown losing to Chrome in the HTML5 Test and Peacekeeper...?

Ya, HTML5 test and Peacekeeper are so 'fair' in the metrics they use...

They don't use any baseline of HTML5 features, and add in other things that have NOHTING to do with Web standards. WebGL, Really?

RealFduch said,

Wrong. I nearly dropped Neowin couple of months ago since it became too much anti-MS biased (And I'm not even talking about comments). Even Engadget is much more balanced now.

I think I will join you.. it's becoming an Apple Fan boy, Anti-Microsoft, mis-truth site...

kavazovangel said,
Those HTML5Test results are way too strange.

The way the chart is laid out is misleading. It makes it look like Chrome's HTML5 support is at least twice as good as any browser because the baseline score is 280, when in fact its score is closer to 20% better (still a decent margin, but in my experience Chrome's "support" is often half-baked).

JonathanMarston said,

The way the chart is laid out is misleading. It makes it look like Chrome's HTML5 support is at least twice as good as any browser because the baseline score is 280, when in fact its score is closer to 20% better (still a decent margin, but in my experience Chrome's "support" is often half-baked).

Unfortunately that's the way the chart was automatically generated, and I couldn't be bothered manually fixing it

Scorpus said,

Unfortunately that's the way the chart was automatically generated, and I couldn't be bothered manually fixing it

Then you should not be bothered to be a journalist

It takes 2 seconds to adjust excel chart

Scorpus said,
Unfortunately that's the way the chart was automatically generated, and I couldn't be bothered manually fixing it

There are limits to "unprofessional" journalism, you know.

kavazovangel said,
Those HTML5Test results are way too strange.

And the test is to testing HTML5 support. It tests whatever it's autor lumped in (alongside some real HTML5 checks).

FoxieFoxie said,

Then you should not be bothered to be a journalist

It takes 2 seconds to adjust excel chart

LOL. That's just classic. Love it. they aren't "journalists" you give them WAY too much credit, they are just glorified Word Processor boys that do what they are TOLD, and no it's too much trouble to be bothered with Accurate details.. so much more interesting work to be done than to report on the TRUTH! We will leave that to the NEXT post to retract the earlier post.. when they are TOLD to fix it..

Don't you love it? Can't think for themselves.. what fun would that be? tsk tsk..

virtorio said,
I'd rather see these comparisons done with Chrome 19 than Chrome 17
they're on chrome 19 already?? sheesh

jasonon said,
they're on chrome 19 already?? sheesh

The Dev build is. They'll hit 20 this year!

But it's really just a number for them. They'll increase it by 1 every other month regardless of new features in order to get recently developed features and bug fixes to stable more quickly. In two years, we should have Google Chrome 30.

virtorio said,
I'd rather see these comparisons done with Chrome 19 than Chrome 17

Google strategy is releasing new Chrome versions so fast that people like you can always dismiss any test that Chrome loses complaining that the N+2 version was not tested.

RealFduch said,

Google strategy is releasing new Chrome versions so fast that people like you can always dismiss any test that Chrome loses complaining that the N+2 version was not tested.
Well my point was, they're using the latest non-final version of Internet Explorer. They should be using the same equivalent version of ALL the other browsers. Does this seem somehow unfair to you?

virtorio said,
Well my point was, they're using the latest non-final version of Internet Explorer. They should be using the same equivalent version of ALL the other browsers. Does this seem somehow unfair to you?

The fact that he FAILED to answer you IS an answer, they cannot give a FAIR test only report on what they are PAID or TOLD to report on... no money no food, so sad, give me the days of NEUTRAL testing.. because this is plain BS

For those wondering. By default the Metro version of IE uses the 64bit IE executables, where as the desktop version uses the 32bit IE executables, hence the difference in score. 32/64bit engine difference.

It may be a bug with the CP, but in the DP you could use the 64bit version of IE on the desktop, you can't seem to launch the 64bit desktop IE on the CP though.

If you're comparing a preview version then you really should be comparing it to other browsers previews which will arrive around the same time (aka. beta/alpha).

Also check that you have hardware rendering enabled in FF (GPU Accelerated Windows should be 1/1 in about:support else you're in software mode). Those scores seem very low compared to what I can achieve (when compared to Chrome).

Phalanger said,
Also check that you have hardware rendering enabled in FF (GPU Accelerated Windows should be 1/1 in about:support else you're in software mode). Those scores seem very low compared to what I can achieve (when compared to Chrome).

To be fair, everything here was done with the browser's default settings upon installation. If it wasn't enabled by default, it would not have been tested with it on

Scorpus said,

To be fair, everything here was done with the browser's default settings upon installation. If it wasn't enabled by default, it would not have been tested with it on


The reasons for it not be enabled can be multiple including:
- Old drivers
- Incorrect installation of graphics
- Multiple vendors of graphics cards (eg. Nvidia/Intel systems) which are not supported till recent update (and not using the latest to compare to the latest is not fair either).

Phalanger said,

The reasons for it not be enabled can be multiple including:
- Old drivers
- Incorrect installation of graphics
- Multiple vendors of graphics cards (eg. Nvidia/Intel systems) which are not supported till recent update (and not using the latest to compare to the latest is not fair either).

So one browser has old drivers while other browsers have new?
LOL

RealFduch said,

So one browser has old drivers while other browsers have new?
LOL

No, each browser has a safe list because they use different engines to do the rendering. In this case Mozilla decided they would support only drivers from a set version forwards (this is where the features were correct in the drivers for what they wanted to do). Because compatibility is more important than speed they take it safe and go to software mode if it does not match.

IE does not use the same safe list but will just fail to render if setup wrong. This is why sometimes people wondering why their browser is showing nothing (when if you manually force hardware rendering off you will find nothing is wrong). With Chrome I don't know it's back end well enough but it is related very closely to Mozilla's.

Phalanger said,

No, each browser has a safe list because they use different engines to do the rendering. In this case Mozilla decided they would support only drivers from a set version forwards (this is where the features were correct in the drivers for what they wanted to do). Because compatibility is more important than speed they take it safe and go to software mode if it does not match.

IE does not use the same safe list but will just fail to render if setup wrong. This is why sometimes people wondering why their browser is showing nothing (when if you manually force hardware rendering off you will find nothing is wrong). With Chrome I don't know it's back end well enough but it is related very closely to Mozilla's.

There is so much incorrect about everything you are saying, let alone that it has any relevance.

IE9/IE10 also have 'driver conditional' and GPU condidtional failsafe 'software rendering' considerations. And it doesn't just 'fail' to render...

For example, An Intel 945 GPU on specific Netbooks with early Win7 drivers falls back to software rendering in IE9.

IE doesn't just 'fail' to render, in fact if it finds even the hint a fault it will switch to software rendering mode just for the page it is finding issue with, without the user ever noticing.

If IE continues to find 'faults' it will turn all of IE back to software render mode and flag the issue in the driver that can be sent to Microsoft.

The glitch with the blank pages and turning 'hardware rendering' off actually has NOTHING to do with IE. This happens based on the GPU Driver settings, when users start playing with 'forcing' 3D GPU features on 'all applications', including IE.

The same 'type' of bug also occurs with other hardware rendered applications, especially WPF/.NET applications.

As an example, turn on Morphic AA on an AMD driver that is system wide in Preview driver 12.1, and Windows Live Messenger 'accidentally' gets blurred by being anti-aliased by the ATI driver.

The reason IE and other applications can't 'see' these issues and fallback to software rendering is that the driver is applying them at the final 'composition' layer in the WDDM of Windows.

There are some of these bugs that also affect how Firefox renders on Windows, and are not detectable by Firefox and are more than just driver, as they are user 3D setting conditional as well, so Firefox can't work around them either.

thenetavenger said,
IE9/IE10 also have 'driver conditional' and GPU condidtional failsafe 'software rendering' considerations. And it doesn't just 'fail' to render...

For example, An Intel 945 GPU on specific Netbooks with early Win7 drivers falls back to software rendering in IE9.

IE doesn't just 'fail' to render, in fact if it finds even the hint a fault it will switch to software rendering mode just for the page it is finding issue with, without the user ever noticing.

If IE continues to find 'faults' it will turn all of IE back to software render mode and flag the issue in the driver that can be sent to Microsoft.

The glitch with the blank pages and turning 'hardware rendering' off actually has NOTHING to do with IE. This happens based on the GPU Driver settings, when users start playing with 'forcing' 3D GPU features on 'all applications', including IE.


Actually it does fail with default installs of drivers which do not operate as expect. As IE does not check for a visual output (which it could have done by checking the buffer), it will blank screen. This is why others implemented lists. While I do find it great that IE attempts to do fall back, it was not implemented correctly.