I worked on Driv3r for GBA. I developed the cutscene and compression libraries, and I can confirm that those two guys were absolute wizards.

I can also shed some light on the Mode 5 mode used in Spider-Man. I worked on Spider-Man 2, specifically, among other things, on the FMV sequences, which also played parts of the movie in-game, as well as on the still-image compression. I wasn’t around when Spider-Man 1 was developed, but I was told the story by members of the team who had worked on both projects.
Spider-Man 1 came out in 2002, at a time when FMV in portable games was quite uncommon. Having even five seconds of video in a game would make players go “WOW!”. Mind you, the PSP, which made video on the go a commodity, wouldn’t come out until three years later.
Activision developed the video player for that game in-house, in collaboration with the developer, and it was fairly basic. It essentially consisted of an RLE image encoder. Individual frames from the video would be pre-processed with a posterisation filter to maximise flat-colour areas and make the RLE compression more effective. It was basically an image compression system applied to a sequence of image. Most of the (few) games who had FMV in that era used a variation of this methodology.
With this approach, even at very low frame rates, the payload for full-resolution video sequences could easily exceed 100 KB/sec. Cartridges commonly came in 8 and 16 MB sizes back then, with 32 MB still relatively rare. The videos were displayed in Mode 3 or, more often, Mode 4, using a 256-colour palette that could be switched at every frame or keyframe. Studios could therefore afford to play only a few tens of frames with this system, resulting in just a few seconds of video, often used to “pad” otherwise unused cartridge space.
For Spider-Man 1, however, space was rather scarce, and the simple codec they had developed couldn’t fit the entire sequence into the remaining cartridge space. So they came up with the Mode 5 trick, which essentially reduced the size of the individual frames, bringing the bitrate down to a more manageable 100 KB/sec or so.
By late 2003, when work started on Spider-Man 2, video encoding had come a long way in just those two years. At the time, I was licensing a proper video codec with motion compensation, codebooks, YUV colour space, etc., which could encode video at less than 50 KB/sec in Mode 3. Activision was my first client, with Spider-Man 2, and I was green as grass in the industry.
The team were really a class act: talented, humble, and patient. It was the most amazing “job” I ever had. Anyway, I went on to develop the cutscenes, image compression, and various other bits.
By 2004, when the game came out, 32 MB cartridges were becoming more common and other video codecs besides mine were starting to become more prominent, so Mode 5 was no longer needed.

Note: i like how that Sonic TV show teachs to kids to run by the center of the road between the cars

Holy crap, I had no idea Sonic Battle wasn't even true 3d. That's insane, and extremely clever programming.

GT Advance was NOT part of the Gran Tourismo series! Although I'm sure the "GT" letters stood for the same thing or rather the English equivalent "Grand Tour"; GT Advance was developed by MTO and published by THQ, who have no relation to Polyphony Digital or Sony Computer Entertainment Inc.

Wonderful video and really well made! Long live the GBA :)

I think that was your best video yet. These explanations are even more interesting when the different techniques shown are related thematically via their hardware or because they are different approaches to tackle the same problem.

30:28 Finally, the game I was waiting for when I clicked on the video!

My bank account constantly runs in mode 0.

Stuntman was a PS2 game initially, with the GBA version being released 1 year later. There was a new one announced recently that uses real movies for the stunts

007 Nightfire isn't true 3D and you can tell because it uses "Y-shearing" for looking up and down, it's probably using the same kind of column rendering technique that Doom and other "2.5D" shooters use.

16:00 The benefit of having separate lines from grass is decoupling the tiling / resolution of the grass from the lines. Enabling way more optimization tricks on both. Having the lines baked in the grass would stop the grass from being tilable. Which is exactly why the grass looks so much higher resolution than the other football game that don't use the mode

Bump Mapping on GBA.
I repeat, Bump Mapping on GBA.

Fun fact: there's an undocumented mode programmed into the console that was never used by any game and we're not really sure what the purpose of it was.
In fan documentation it's called green swap, and all it does is swap the green value of neighbouring pixels to create a weird chromatic aberration that you see on some 3D films.
It's not really useful for anything, as I said it was never documented and fan games don't make use of the feature.

Thanks for explaining why Mario Kart Super Circuit has a 30fps track and 60fps background elements!

Oh yea, Golden Sun! Still remember how the graphics had me marvel over them. The game looks gorgeous back then and still does today on my GBA. The console has so many great titles.

And it's worth mention that since the Nintendo DS was basically an enhanced GBA, it also had distinct graphics modes, with some extra wonkiness for how it handled the second screen. It'd be super cool if you did a deep dive on that system as well!

genuinely hilarious to see downhill jam use vertex shading only except for the giant nokia billboard

10:58 I like that when you mention it's also called soccer, you're showing a match between two countries where it is called soccer

Doom and doom 2 were absolutely mind blowing at the time. Very close ports including secrets, steps, multiple height levels and so on.