The most notable difference between Windows 8 and previous versions, in terms of fonts, is the new assymetric greyscale with subpixel positioning rendering in the Metro environment. The good news about this is that it is considerably better than I feared, and still manages to provide better stroke density than Apple's full fuzz rendering.

In windows 7, you could re-install fonts and have the new version appear after restarting Adobe apps. Now that no longer works, and you have to restart to get updated fonts to appear. I don't use any font management tools. Fontlab works just fine.

I'll be getting a new laptop - and it's hard to find a Win7 system (plus I've heard horror stories about downgrading OSes, and it would cost more). My current one (which is a dinosaur, but it's the king of dinosaurs) is running XP (like a rock) and at least I can see all the rendering problems so many users see but so many foundries look away from... But anyway I use Win7 systems often, and I like it. BTW I've heard good things about Win8 in terms of backwards compatibility; and the other day I connected a funky no-name external DVD drive into a Win8 ultrabook and it "just worked" within seconds.

Concerning what Jeremy said: I wonder if a font management tool fixes that issue; but I really don't want to resort to one.

hhp

As I understand it, Windows 8 deals primarily with touch-screen issues, so there's no need to upgrade if you're running a Grandpa Box, like me. One thing I find especially appealing about Win7 64-bit Pro is its virtual machines: you can run programs as old as Windows 95 (maybe even older) in their own space, no sweat.

@hrant
Laptopisaurus Rex, eh? Mine died some months back but I had already received a Dell with Win7 that I had ordered in anticipation of the dreaded passing of ol' reliable.

The Win7 Dell Latitude has the "XP Mode" virtual machine installed and I've found it best to just migrate everything font related to XP. Win7 gave trouble with various pieces of my font workflow from the get-go so I just avoided the problems completely by sticking with XP running in a virtual machine.

I've tried to imagine at what point this setup will fail me but you know, I really can't see with what or how it would give me trouble.

Win7 running as the parent OS gives me access to IE9 and other upgrades that won't load in XP for testing so it's the best of both worlds so far.

@Richard, that is exactly the setup I have with my Latitude. I also have the disc option of running a 64 or 32 bit Win7 platform.

Does your Latitude over heat? Mine gets really warm.

n.

Ben, here is a good general description of the change in font rendering in Win8, with examples:
http://www.extremetech.com/computing/138816-windows-8-the-desktop-review/3

Broadly speaking, there are now three main ClearType rendering modes (with some version-specific variations in the GDI rendering):

GDI : x-direction colour subpixel antialiasing, no y-direction antialiasing, full-pixel spacing.

DWrite : x- direction colour subpixel antialiasing, y-direction greyscale antialiasing, sub-pixel positioning.

Metro : x-direction greyscale subpixel antialiasing, y-direction greyscale antialiasing, sub-pixel positioning. [Note also: at small sizes the x,y antialiasing is asymmetric, at larger sizes it is symmetric.]

All three of these rendering environments may be encountered in Windows 8.

> In windows 7, you could re-install fonts and have the new version appear after restarting Adobe apps. Now that no longer works, and you have to restart to get updated fonts to appear.

I'd try removing the font first and then installing an updated version.

Subpixel may not be the best word, or maybe I need to distinguish between sub-pixel and subpixel. I would classify Metro greyscale as sub-pixel rendering by virtue of it using supersampled outlines with fractional positioning. This means that the pixel tint is determined by where the outline falls sub-pixel.

Does the 'asymmetric' antialiasing mean that horizontal stems are confined to full pixel positioning?

Yes.

In DirectWrite you can control at what pixel size range(s) asymmetric/symmetric antialiasing should be applied. You need a version 1 gasp table for that and use the GASP_SYMMETRIC_SMOOTHING flag.

If a version 0 gasp table is present (or, iirc, a v 1 gasp table without the symmetric smoothing flag), symmetric smoothing will be applied from 21 ppm upwards.

This is the second time I've heard mention of the 'greyscale' rendering of sub pixels. I'm confused as to how 'greyscale' and sub-pixel rendering work together (since sub-pixels, by definition, are not greyscale).

>The Metro greyscale rendering is sub-pixel on the grounds that the fractional positioning of the outline affects the pixel tint.

This is what I'm not quite getting. Isn't that true of all anti-aliasing?

This is what I'm not quite getting. Isn't that true of all anti-aliasing?

No, because in previous antialiasing glyphs have been positioned on full pixel widths, i.e. the 0,0 coordinate of each glyph's advance width has always been positioned on the leading edge of a pixel. This was also true of ClearType subpixel rendering in the GDI implementation. This accounts for sometimes uneven spacing in traditional greyscale and GDI ClearType.

In DirectWrite ClearType and Metro greyscale antialiasing, fractional sub-pixel positioning is used, i.e. the 0,0 coordinate of a glyph may occur anywhere within the pixel. What is means is that spacing is much more even, but that there is variation in the actual pixel pattern for the same glyph at different places in a line of text. In the Metro greyscale rendering, this variation displays as differences in the grey tint of pixels (in DWrite CT it displays as differences in the colour tinting, and is generally more prevelant and noticeable than in the Metro greyscale). This is why I classify the Metro greyscale antialiasing as 'sub-pixel', because the pixel tint is being affected by the position of the glyph according to fractional positioning.

"Metro greyscale antialiasing"

Ah! That makes sense, as antialiasing uses the entire pixel, it can truly be grayscale (or in this example, matched to the type color).

It gets confusing when when the term 'sub-pixel' and grayscale are used together...but the term 'fractional positioning' helps clear that up (my understanding is that Metro uses full-pixel anti-aliasing but allows the glyph borders to land in the middle (or some fraction thereof) a pixel).

Thank you for your patient explanation, John, I see the distinction. I presume one implication of the Metro sub-pixel positioning (and the DWrite CT) is that x-hinting instructions are ignored, since they're trying to align the stems to whole pixels so that glyphs have a consistent appearance when rasterised? Or do I need to learn more about hinting before that starts to make sense?

>In the context of Metro rendering, asymmetric refers to the difference in x- and y-direction supersampling (at text sizes). The rendering is 8x4 greyscale, meaning that the x-direction antialiasing has double the senstitivity of the y-direction antialiasing.

What's supersampling? I'm guessing it's the way the rasteriser is looking at glyph outlines and deciding how much black to put into each pixel. If that's right, then it seems that 8x4 greyscale means that each pixel is assessed by breaking it down and looking at how many of the 32 virtual sub pixels the outline covers, which results in the pixel being attributed one of 32 different levels of grey.

What's supersampling?

Think of a glyph outline being stretched to a larger number of pixels than actually requested for the target display size/resolution, and then the mapping of the outline to those pixels being downsampled (like a scaled down JPEG) relative to the requested size. So in the case of an 8x4 asymmetric supersampling, each pixel in the requested ppem size is first represented by 8 horizontal pixels and 4 vertical pixels, and then scaled down to the requested size. Obviously at larger sizes more refinement of the relationship of outline to pixels is possible and hence more antialiasing precision (samples) is available. This captured information influences the rendering at the target size, where the relationships that were relative to pixel dimensions in the supersampling become relative to sub-pixel dimensions in the final rendering.

Performance certainly may have been a priority.

But it also seems like Microsoft really wants Win8 to succeed on the tablet. I think both the screen rotation and higher PPI of that screen would justify this method (IMHO).

"Why sacrifice the quality just because you can rotate a screen?"

Well, one argument would be consistency. Having type look better in one rotation than the other might not be the best user experience. (But I'm not even sure if that's really all that important anyways, as I *think* most of the handheld Win8 devices are using high PPI screens, so sub-pixel rendering is likely moot).

I really don't think the whole rotation question played a role in the switch to the Windows 8 greyscale rendering. Remember, the alternative -- already implemented through most of the development cycle -- was DWrite ClearType, so there was already bidrectional antialiasing (unlike GDI). If tablet and phone rotation had been a concern for text rendering, I would expect this to have been something addressed quite early in the new operating system development cycle. The Metro greyscale rendering was implemented at the last minute, after initial beta versions of Win8 had shipped with ClearType, in response to performance complaints.

Yes, the illustration text -- taken from a website viewed in different versions of IE with different renderings -- was spec'd grey. For illustrating the differences between the rendering models, it could have been pink, orange, purple or, indeed, black. It just happened that the example I had to hand was grey, but I should have mentioned that in case anyone thought it was intended to be black.

This one too? (from your link above).

David, what are you getting at?

hhp

"If I upgrade to Windows 8, ..."

Don't upgrade. Buy a new computer. Read the following:

Walter Mossberg (WSJ) on upgrading: www.j.mp/w8noup

Thomas De Maesschalck: Two months with Windows 8 www.j.mp/win666

Win7 the New XP SP3?
If you’re still running Windows XP, take the time to review your options. Microsoft ends support for XP in 15 months, on April 14, 2014—the last official day the OS will get security updates. With the verdict still out on Windows 8, now might be a good time to give yourself the present of a new Windows 7 machine. Think of Win7 as the “new XP SP3.” It might be the next long-lasting, long-supported Windows operating system that Microsoft will need to pry our fingers from.
—Susan Bradley, WindowsSecrets.com, Dec. 6, 2012

@Renaissance Man:

I don't believe MSFT will discontinue security updates by that date -April 14, 2014.
That might be what they are announcing but I'll believe it when I see it.

And...... since I'm not running Office apps or anything sensitive within the XP Virtual Machine, I really don't care if they freeze security support or not.

Richard,

“Creative imitation” may be a misnomer in this case. From what I can gather, Windows 8 Metro interface is considered to be the wave of the future—less is more, if you will. The whole Apple thrust—higher res, more glitz—turns the interface into a bandwidth hog, which is great for the IPSs, but it totally lousy for the people who have to pay for that bandwidth. If I were a person given to conspiracy theories (which, I am, but I don't want to advertise it), I would aver that Apple's ultra-high-res approach to tabs/mobile/etc. has been developed in cahoots with carriers to ensure that that maximum amount of data transfer occurs (billable units) at any time. Win 7 Pro 64-bit rules. Totally. Indubitably. My story and I’m sticking to it…