I ordered my iBook right on October 30, when it had just been released to the public. I always wanted to buy an Apple notebook to complement my various desktop systems, because they, quite frankly, all lacked protability.
I had been eyeing at the 12" Powerbook for a while, because I wanted a small, yet powerful notebook that would both fit into a small backpack as well as play some games decently. But the poor graphics card in the 867MHz model (the GeForce MX 420 Go) and the high price had kept me from buying one right away.
When Apple released the 1GHz model with the somewhat improved GeForce FX 5200 Go, I was tempted again, but I was still unhappy with the graphics card, as it was again the slowest card available (in the PB G4 new line) and the pricetag hadn't improved much either.
So the announcement of the iBook G4 with a decent 800MHz G4 and a very decent Radeon 9200 Mobility came in perfectly for me. In addition there's a programme for University students here in Switzerland, offering massive rebates on newly released Apple products. So I was able to get the iBook at 22% less than usual. I couldn't resist. - And I don't regret it.

Shipping took an extremely long time, by all standards, not just the "anxious Apple user" one. The package took two 4 day breaks between Taipei and Switzerland, and eventually arrived on October 12 in the afternoon. Almost 2 weeks from ordering. But I hear other customers are facing even worse situations. Apple has to work on their shipment(-partners) in Europe...

Other than that I was very positively surprised by the general appearance of the new iBook. My mother had owned an iBook G3 600 for some time and I had grown quite fond of the little thing. The new iBook G4 is, compared to the old one, more gray/white and doesn't have that metal finish. However it also does have the same color for the keyboard and the rest of the surface, which the old one did not have. The new one looks better in that respect. However it's a bit more "plasticky" than the old one. The shell is now made of a white plastic, instead of transparent materials backpainted with white color. I like both designs, but the G3 iBook scratches easier, due to the transparent material resulting in some kind of mirror/prism effect, making scratches very visible.

Size and weight are almost identical, the new one being a tad lighter.
The biggest difference between the two is really the keyboard, which is far better on the iBook G4, thanks to a metal backplate and a more solid, powerbooky feel. A definite improvement over the old one.
Battery life, while already excellent on the iBook G3, has been further improved and easily reaches between 4 and 5 hours of actual use, with Airport running. Far beyond any PC laptop I have ever seen.
Talking about Airport: Reception is again, excellent. I shared an internet connection from my MDD G4 over Airport and went for a walk in the garden with the iBook. Despite the PowerMac being in the house and behind aluminium window covers, I easily reached an approximate 50-60 meters before losing the signal completely. I don't have any means of testing Airport Extreme range (802.11g), but I expect it also to be great.

Overall, the iBook feels very fast compared to the older model. Panther is a big improvement for the G3 too, but no match for the iBook G4 with it's faster frontside bus, processor and RAM, not forgetting the graphics chip, which is about a zillion times faster.

So, being a long time Apple user and digging into benchmarks, I had to compare to all modern Macs within my reach. All Macs were equipped with Mac OS X 10.3.1 and the latest updates.

That list consists of:

• PowerMac G4 MDD Dual 1250MHz, 2MB L3 cache per chip, 1GB RAM, Radeon 9000 Pro 64MB
  
• same machine as above, but with one CPU disabled with Apple's CHUD tools.
  
• PowerBook 15.2" Titanium G4 550MHz, 256MB RAM, Radeon Mobility 16MB
  
• iBook 12" G3 600MHz, 640MB RAM, Rage 128 Mobility 8MB
  
• iBook 12" G4 800MHz, 384MB RAM, Radeon 9200 Mobility 32MB

I tested with synthetic benchmarks, productivity applications and games.
Some tests can be directly compared to previous tests Mike has published on the site, while other's can't. Configuration files can be found here, if you're interested on all the settings in the custom configs (would be way too much to post here).
Comment's to each test are noted below each diagram. Keep in mind that benchmarking is rarely a 100% accurate way of measuring a machine's performance. Especially not when comparing different configurations of RAM, GPU, etc.

Synthetic benchmarks:

Xbench 1.1.3
This application tests a broad range of system features and performance. It's scores can vary quite a bit from run to run, so it's not really the most accurate benchmark on the planet. Still it's a usable general indicator for overall performance. Almost full or fragmented harddisk adversely affect the overall score, therefore the PowerMac scores a bit below expectations, because the disk takes about a 50% speed penalty compared to a freshly formatted one. Same goes for the iBook G3 and PowerBook G4, but not to that degree.

This graph shows several things that might be of interest:
The PowerBook scores faster than the iBook G3 600, despite being 50MHz slower. This is mainly due to a much better GPU, Altivec enhancements and better RAM and harddisk performance. It also shows that Xbench makes little use of Dual CPUs.

Altivec Fractal Carbon
This application is merely a mandelbrot generator that is extremely Altivec enhanced and is SMP aware. This shows dramatically in the graph below.

Cinebench 2003
This app is not Altivec enhanced, but is SMP aware for rendering. In fact it is so aware that it even detects the second CPU when it is disabled with Apple's CHUD tools. I could therefore not run the single CPU tests on the PowerMac. Something was very weird about that app. Obviously Cinebench profits a lot from fast graphics cards and memory interfaces.

Running the dual CPU render is only possible on a dual machine, of course, therefore only the PowerMac G4 is listed there. I used the old version of Cinebench, not the G5 optimized beta version, since there is no G5 in the testbed. This way the scores are comparable to older tests around the internet. Note that the two older notebooks are faster in software OpenGL mode than in hardware OpenGL mode. Tells a lot about the graphics cards...

Quicktime Pro 6
In this test I first copied a music track over from CD, resulting in a 110.8MB AIFF file.
I then exported that file to mpeg4 using the default settings in Quicktime Pro. Since mpeg4 is quite complicated to encode, the stress on the harddisk is small enough to not make it a big factor.

Quicktime makes almost no use of the second CPU, however, Altivec seems to play a major role in mpeg encoding.

iMovie
In this test I imported a composite file into a new iMovie project. iMovie only handles DV Stream formats, so it will convert media before you can work with it. The files used were a 386MB mpeg4 movie and an 28MB music track in AIFF format.

Again Altivec is a major factor in this test.

Photoshop
I ran the whole 21 filter suite of the PSBench7 Advanced set, which applies said 21 filters to a 50MB file. Retrospectively I realize the PowerBook G4 was running mostly off the harddisk, because it didn't have enough RAM for this kind of manipulation. Then again, I doubt the few Altivec enhanced filters would have helped it much.

What I think is quite irritating is how poorly the filters in PSBench make use of a second CPU. The difference displayed is the result of a single filter, not really of the other 20. Radial Blur seems to be SMP aware, making a 30+ second difference, while other filters don't profit at all.

Preview
Preview has been sped up in Panther a lot aparently, so I wanted to test it's .pdf scrolling capabilities myself and came to the following conclusions:

It looks as if a certain amount of performance is the main factor in Previews scrolling speed. Looking at these results, I dare say it depends on the graphics card almost entirely. You can test this yourself by downloading the testfile here and then scrolling through it by using the downarrow key. Smooth scrolling was disabled for this test on all systems.

Return to Castle Wolfenstein MP
This game is, like many first person shooters, based on the Quake3 engine originally. However it has been heavily modified and now lacks proper SMP support and is much more demanding of both the CPU and the GPU, than Quake 3 itself is. I deleted the config file for the Wolfenstein MP application, forcing a new config to be built. I set graphics to high quality, but turned resolution down to 640x480. FSAA and AF were both disabled as well.

Again, a very interesting graph: The MP awareness of OS X and probably even the OpenGL library make quite a difference, even though Wolfenstein is, by itself, not SMP aware.
The iBook G3 600 with its measly 8MB Rage 128 Mobility gets totally killed by this game with these settings. The main factors here being 32bit colors and high detail textures. The lack of hardware transform and lightning capabilities does the rest. (Note: Quake3 engine games do not use hardware lighting, Carmack does that in software.-Mike) What I do not quite understand and what remains a mystery to me is why the Radeon Mobility in the PowerBook also performs so badly. Maybe a reader will be able to point me to the source of this performance impairment. (See this ATI page comparing the features of the original Radeon mobility, the 7500 Mobility and 9000 Mobility GPUs. Although they do not list specific core/memory clock speeds, the 7500 and 9000 are clocked much faster than the original Radeon mobility, as well as having 2x to 4x the number of 3D pipes and up to twice the memory bus width.-Mike)
The iBook G4 on the other hand is performing quite nicely. 30 FPS is already very playable for singleplayer games and with some extra tweaking or slightly lower graphics settings, things will only improve for the iBook here.

Jedi Knight II MP
This game is very similar to Wolfenstein in terms of technology: It's based on pretty much the same code. Here again, I started with a default config file, setting graphics to high quality.
High quality in JK2 means "Very High" texture detail, which is intended for 64MB cards with support for texture compression. The Rage 128 Mobility can't handle this, so I had to set it back to only "high" detail, The ideal setting for 32MB cards. FSAA and AF were both turned off. Since the latest version also introduced a bug with SMP support, making it crash in Panther, I had to turn it off. Too bad really. I hope this will be fixed sooner or later.

Interestingly, the PowerBook performs quite decently in this game, compared to the iBook G3 anyway, which is completely unplayable here too. I wonder what Wolfenstein does wrong, that Jedi Knight 2 does right. The iBook G4 again is perfectly playable even with these relatively high settings.

Quake 3 1.32b
The veteran among the benchmarking games and my personal favorite still. If you don't know what this game is about, you need to read xlr8yourmac more often! I used the latest versions available, the G4 optimized version for all G4 equipped machines, and the normal version for the iBook G3.

Configs were made as follows: My match config is an evolution of over 4 years and includes so many tweaks and enhancements that I decided to simply include it in the testfiles zip, instead of posting pages full of explanations. Boli's config is tweaked to a similar degree, but a bit faster yet, due to lower geometric detail and the lack of cg_drawgun 1, mainly. It's very popular among Mac Quakers, so I wanted to include it. The "Fastest" config is a default config with the "fastest" settings applied and resolution upped from 512x384 to 640x480, since the notebooks do not support any resolution below 640x480. Additionally music was turned off and s_chunksize was set to 4096 in order to make these comparable to Mike's previous tests. The "uHQ" config is also a default config set to high quality settings, but enhanced further by changing the following values: Resolution up to 1024x768 (from 800x600), geometric detail High and Texture detail on maximum (r_picmip 0). Music was also turned off and s_chunksize set to 4096.

r_smp was set to 1 on all tests, of course it is only used when more than one CPU is actually present.

In order for you to imagine how these configs look, I took a screenshot for each of them (converted from .tga files to jpegs, since the .tgas were 900KB to 2MB+).

My config, boli's config, "Fastest" config, ultra high quality config

As you can see, the iBook G4 is a very solid performer in Quake3, distancing the PowerBook G4 by almost 50%. Of course the dual G4 is out of competition in this test. Note that the single CPU results for the Powermac are "true" single CPU results, as the CPU was deactivated, not just r_smp. Therefore also OpenGL and the OS do no longer benefit from the extra CPU, thus giving a clearer indication of true Single CPU performance for a 1250MHZ G4. I consider this important for people who have upgraded their G4s with a single CPU G4 upgrade card. What also should be noted is that the uHQ tests for the iBook G3 were run at 640x480 instead of 1024x768, since that resolution is only supported in 16bit mode in that machine (too little VRAM).
Further note that the iBook G3 is the only machine for which the "fastest" config actually IS the fastest. This is because this config is the only one of the 4 to use 16 color and textures. The Rage 128 series of graphics cards always performed better in 16bit mode, than in 32bit mode, for which they were obviously not designed.

Unreal Tournament 2003
This is the most modern game in my test series, so modern in fact, that 2 of 4 machines don't even match the minimum system requirements anymore. Still, I was amazed to find that all machines actually ran the application, although very, very slowly. The iBook G3 also had render errors, most likely due to the lack of hardware T&L and/or the lack of multitexturing support. UT was forced into 16bit mode on that card, although the settings wouldn't have indicated that. But it was obvious. The two configs used where a default config with resolution turned down to 640x480 (instead of 800x600) and my custom config, which is tweaked beyond what the menu allows you. That's mainly cache sizes, VRAM sizes, sound channels, texture details etc. That file is also included in the testfiles collection above. The asbestos flyby and botmatch benchmarks were used.

This game is a sad indicator of how quickly computers become obsolete regarding games. Both the PowerBook G4 and iBook G3 are not actually usable with this game anymore. The iBook G4 is fast enough for singleplayer games with some tweaking, and should also be ok for the casual online gamer. Consider it's extremely low price, it's e hell lot of bang for the buck!

If you really want to see this game fly however, you need a Dual G5 with a Radeon 9800 Pro Retail.
(or, don't be angry, a fast PC)

OpenSSL
Knowing some people dig into this, I spent some extra time doing a series of OpenSSL benchmarks on all systems. Since the results from this app are hard to display, I decided to simply link to the textfiles of the output. Those who are interested in this, will know how to interpret it.

In the terminal, the command "openssl speed" invokes a series of calculation on a number of opensource security protocols/encryptions. This is similar to distributed net's RC64 cracker, just that the intention isn't to crack the code, but merely to test how fast it could be done.

Here are the results for:
PowerMac G4 Dual, PowerMac G4 Single, PowerBook G4, iBook G4, iBook G3

Screen brightness
I have also made a few pictures for those who are interested in the brightness of the new iBook's screen. It's hard to photograph, even with manual camera settings, but I think you get the idea:
the new iBook G4 (to the right) is slightly brighter than the old one, especially if viewed from an angle.

Conclusion
There's no denying it, the iBook G4 is one of the best laptops Apple has ever produced. If you've been looking into buying a portable, consider this. Go into the next local Apple Store or retailer and have a look. This notebook is more than just a companion for students and fans, it's probably the most affordable yet speedy and fully featured notebook you will ever find on the market. And it's the only one that has enough battery life to watch two full length DVDs in a row.