wrong, wrong, wrong

| 22 Comments

Microsoft misleads PCMag.com's Michael Muchmore into writing fiction about Firefox in the article Microsoft: IE9's Graphics Acceleration Better Than Chrome and Firefox.

Show did this happen? It started when the the IE blog said "Based on their blog posts, the hardware-accelerated implementations of other browsers generally accelerate [rendering or compositing] but not yet both." and "Today, IE9 is the first and only browser to deliver full hardware acceleration of all HTML5 content."

(Did that MS guy somehow miss our blog post that said pretty much exactly opposite?)

From those misleading and simply wrong statements, one could see where Michael Muchmore was led astray, though his headline is a bit of a stretch given that MS never explicitly calls out Firefox.

But what's really strange is that Michael's own testing pretty much disproves Microsoft's claim (or proves it wasn't Firefox that Microsoft was referring to) but Michael is unwilling to say that explicitly in his article, He does note that Firefox is neck and neck with IE, even beating it on some of Microsoft's own acceleration tests, but nowhere does he challenge the premise of the article, that Firefox hardware acceleration isn't complete. Didn't he just not realize as he typed it that Firefox and IE were at near parity and in a completely different league than Chrome?

The Microsoft blog post is misleading and wrong and the PCMag article is also wrong along with the added benefit of being somewhat confused.

The facts are that Firefox takes advantage of the same Windows 7 APIs that Microsoft does to accelerate both the compositing and the rendering of Web content and that Mozilla provided test builds of Firefox, for folks like you, me, and Michael Muchmore, with this hardware acceleration well before Microsoft did.

We are faster and we were first.

But here's the even cooler part. Firefox goes one step further and accelerates for Windows XP users too -- something Microsoft says they can't do. So, for all their hand-waiving about the difficulties of multi-platform acceleration that, according to them the other browser vendors face, it seems Microsoft are the ones struggling to support even their most popular Windows version. If Mozilla can accelerate browsing for the hundreds of millions of PC users on Microsoft's Windows XP, why can't Microsoft? And why are they spreading FUD about Firefox when we beat them not just on the latest and greatest Windows 7 but on Windows XP where a majority of their users are today?

As for the Chrome supporters claiming that Chrome 6 wasn't the right version to test, that it's Chrome 7 that's got acceleration, yes, Chrome 7 preview builds do accelerate compositing, (6 did with a pref switch) but Microsoft is mostly right about Chrome in their blog post. Compositing is only part of the puzzle and with just that part it won't be as fast across the board as IE and Firefox -- even Chrome 7.

update: Go read what roc wrote.

22 Comments

I am a bit sad with all the talk of acceleration. I have a older computer with a AGP slot and no PCI-E slots.. so I am sorta out of luck with any sort of hardware acceleration... I am almost considering getting a NVidia 8400 or 9400.. as those have CUDA... and are supported on regular PCI.. which probably wouldn't be enough to speed up a browser.

I have a AGP Geforce 6800.. which is pretty much as fast as one can go on a AGP slot.

It would be nice if there was Open GL/CL hardware acceleration that was usable.. so that it could be applied to older video cards.

No, the blog post *does not* say exactly the opposite. Here's what the blog post says:

"Content acceleration [...] has been activated in this new beta."
"Compositing acceleration [...] is not activated by default yet."

Now, this certainly implies Firefox is going to have both types of acceleration if you are following Gecko's development closely. This is not explicit at all in the blog post, though; the "yet" in the blog post could be referring to future versions of Firefox. If you read only the blog post and nothing else, Microsoft's conclusion is the only conclusion actually reachable.

In fact, the IE blog post says:

"Based on their blog posts, the hardware-accelerated implementations of other browsers generally accelerate one phase or the other, but not yet both."

The "yet" in the IE blog post matches the "yet" in the hacks.m.o blog post perfectly.

Yeah, the MS guy responded in the comments with basically the same exact point that MauricioC made. It's disingenuous and wrong-in-a-week when the next beta comes out, but he can claim to be technically correct. Very sneaky in getting media to report that; about what I'd expect from MS.

I don't understand what all the fuss is about with this graphics acceleration. Are there ANY sites that could see a speed boost from it, beside IE's own test suite?

Microsoft has a very good motive for claiming they can't use graphics acceleration on XP: they want people to upgrade to Win7. They said from the beginning IE9 will not be available for XP users.

"not activated by default yet" means that the feature exists in the product but the pref isn't flipped to default it to on. Just like Chrome 6 where they had compositing acceleration included but not on by default. Asserting that something doesn't exist yet when it in fact does is wrong. Firefox can clear your private data on every exist, but that's not activated by default. That feature exists and can be turned on if the user wants it. See how that works? If you finish reading the blog post, this becomes obvious.

As of right now, if I download the latest Firefox beta and the latest IE9 preview then run the Psychedelic Browsing test on the IE test site, IE9 manages 1449 revolutions per minute and Firefox just 4.

That may change if I can dig around deep enough in the Firefox settings and enable something that clearly hasn't been considered stable enough to have on by default, even in a beta build. It's just disingenuous to suggest that constitutes a working implementation before the IE9 preview.

Now, the interesting part will really be once both browsers are final, then we'll see which is indeed the faster. Before then, it's just pointless posturing.

"Asserting that something doesn't exist yet when it in fact does is wrong. Firefox can clear your private data on every exist, but that's not activated by default. That feature exists and can be turned on if the user wants it. See how that works? If you finish reading the blog post, this becomes obvious."

it depends, "clear private data on exit" is in the options menu, clearly presented to the users, while compositing acceleration is a feature in development, and not activated by default in the test build because the developers think it's still not stable enough even for the regular testers. So I think it's okay to say the feature is still not there if the developers think it's still not ready as a default even in the test build. A feature that's not even ready for the testers, and even a tester needs to go into the about:config to enable it and experience how unstable/problematic it currently is, is as good as non-existent for the general public.

This is an odd one -- I also wrote a story based on the IE9 Blog, because it echoed my findings (as you know, I've done quite a lot of hardware acceleration testing).

As you know, HW acceleration is broken in the latest FF build, but in the previous build, IE9 was faster.

I spoke to Vladimir about it, and last I heard you guys were putting some final touches on D3D stuff. I'll be interested to see whether you are faster than IE9 once you get the 'on by default' thing to work.

>I don't understand what all the fuss is about with this graphics
>acceleration. Are there ANY sites that could see a speed boost
>from it, beside IE's own test suite?

Most sites which would benefit from it weren't built in first place, because initial testing revealed that it's currently way too slow. For example a scaled background image with a fixed position and some foreground image with a drop shadow (which acts as the content's background) is currently not feasible (unless everyone switches to Chrome).

There are also quite a lot of sites out there where scrolling feels very sluggish and slow. Retained layers do help a lot there.

Well, even if you don't need the performance, you still might like the lower energy consumption.

And at the other end of the spectrum is animated stuff like games. Hardware acceleration, retained layers, and the vastly improved JavaScript engine do wonders there. With Firefox 4 a huge selection of genres will become possible.

"We are faster and we were first."

Wasn't Opera first?

Just to note on acceleration on Windows XP. Microsoft decided to use WPF as base for IE9, which is from several points wise decision - my guess is that Microsoft will use XAML + .net as a basis for IE9 addons, it seems comparable to Firefox system, but it will have advantage that it is also language for Windows 7 Phone, and I guess Microsoft will try to make it also Windows "default" platform.

But somehow (I have no idea why), WPF on Windows XP sucks. So, from perspective of IE team, I guess it seemed reasonable to drop XP support (and probably their chiefs also liked idea to improve Windows 7 sales)

So just to note that IMHO it is not up to IE9 team that Windows XP is not supported, MS already made such decision before by making WPF on XP not fully functional, and specially not performing well. And also to note that it doesn't change anything global in what you written, though it might put another light on some comments on your post.

Microsoft missed Firefox's blog post and, in turn, Firefox missed Chromium's blog post.

http://blog.chromium.org/2010/08/chromium-graphics-overhaul.html

Travis, in fact, no one missed Chrome's blog post. It's the basis for Microsoft's claims that other browsers (turns out it's only chrome they're following) only accelerate _compositing_ which is only one piece of the puzzle. According to that post, "most of the common layer contents, including text and images, are still rendered on the CPU and are simply handed off to the compositor for the final display... Other layers use the GPU to accelerate needed operations that touch a lot of pixels. Video layers, for example, can now do color conversion and scaling in a shader on the GPU. Finally, there are some layers that can be fully rendered on the GPU, such as those containing WebGL elements." Most of their rendering, with the exception of parts of their Video and WebGL content is still done on the CPU. They've accelerated _compositing_ and a few bits of rendering. That's microsoft's point. Only it's not accurate for Firefox which accelerates all of the usefully acellerated layers just like Microsoft.

Really guys...who cares?

Let the standard mature and become finalized before you hold the "I can do this better than you!" contest.

God forbid a MS employee didn't read a blog that I never knew existed before writing down his own opinions and assumptions.

Get over it and release the browser already. Not like it will make a difference anyways.

I don't think it's fair to say Chrome only does accelerated compositing. True, it does not accelerate ALL rendering, but it certainly is accelerating the more processor intensive rendering operations. And given the Chromium blog post, full rendering acceleration is on the way.

My primary objection is that Chrome's status in this area was not accurately portrayed. Given this is a Firefox blog, I likely would not have complained except you made it on Slashdot. Really, it's just nitpicking, but given that we're all talking about unreleased browsers it's only fair to point out that Chrome does do some accelerated rendering and is working on more.

Michael Muchmore's article is more talked about nowadays. Microsoft made IE9 not available for XP users. Hence they can't use graphics acceleration on windows XP, Its accessible on Windows7 though.

Chrome is safer.

Chrome is safer.

The Radeon HD 4670 AGP is what you need!

Whichever the first one or the fastest, what's more important for the community is that all browsers get more or less equally fast. And that we do not have to write hacks to have pages work across browsers. It looks like from IE9 on we are not going to have one vendor to block the web any longer, and this is good, at last.

While I'm no expert on neither who was fastest nor first, I do know that th IE9 beta delivers full GPU acceleration on my integrated Intel card, while Minefield is currently struggling at 40% CPU to handle a simple animated GIF.

Monthly Archives