NES 'character-mapped background graphics'

Hi all.

I've been reading issue 147 of Retro Gamer Magazine. For Mario's 30th anniversary it discusses Mario, his greatest moments, and games developers chime in with their memories/thoughts.

David Crane, of Activision fame and a veteran in the games industry, had this to say: "My first reaction to Mario was surprise that the game so obviously used character-mapped background graphics. Shigeru Miyamoto made a conscious decision to build Mario's world out of blocks. In doing so, he allowed for a very large world due to memory efficiencies, but it repersented a step backwards in videogame imagery. He took a big risk. As Super Mario Bros. was the flagship game franchise that everyone would associate with the NES system itself, the risk was that gamers would consider the NES to be inferior hardware to machines of the day providing much more realism."

Does anyone know what he means by 'character-mapped background graphics'? Either way, I'm posting this to incite discusison and because it's interesting to hear from a veteran games developer.

I think he's talking about tile-based graphics. The comment is actually kind of strange considering that NES simply doesn't support anything but tile-based graphics (without getting tricky). So I don't think there was much of a decision for Miyamoto to make.

NES backgrounds are made up of 8 x 8 pixel tiles, selected from a palette of 256 different tiles (known as CHR, from "character"). The full screen covers 256 x 240 (61440) pixels, so 256 tiles (16384 pixels) only provides unique coverage for 27% of the screen. There are some ways to get around these limitations, but that's the basic thrust of it. (Colour combinations are also strictly limited on the NES.)

This kind of approach was very natural for displaying text (i.e. a grid using a limited set of characters), but it was somewhat practical for games as well. It provides a good way to build backgrounds with limited RAM.

Nowadays, systems generally have enough RAM to store all the pixels of a screen at once, known as a "framebuffer". You draw the whole image, pixel by pixel, and then display it to the screen. This is much more versatile than a character grid, but it took a long time for RAM to become cheap enough for this approach to be universal. (An example from outside of games: in the early days of CG rendering, "scanline rendering" was favoured because you only needed enough RAM to hold a single line of the image at a time.)

Most early NES games have a fixed set of CHR tiles, i.e. CHR-ROM. Many games put RAM in the cartridge instead of ROM, which allowed for dynamic changing of background tiles. There was still a lot of limitations this way, but it made it possible to do pseudo-framebuffer rendering for later games like Solstice or Battletoads.

Strictly speaking, "character-mapped background graphics" means that backgrounds are composed of a grid of references to a dictionary of blocks. I find it strange that someone would be surprised by an NES game being like this, considering that this is how the system itself is designed... NES games don't really have many other options for drawing their graphics, unless they use a lot of dynamically updated CHR-RAM with lots of PRG-ROM to supply new art, but this was anything but feasible back then.

Maybe Crane was talking about how the concept was extended to another level, with the use of metatiles, which greatly reduce the memory requirements, but also make the blockiness of the world even more apparent. I don't know what he's comparing SMB to when he says this approach is a step backwards in video game imagery though... it's not like the competition had a lot of memory to work with. Maybe the comparison was against games with small levels, which were more common in the early days of gaming, and didn't need to repeat graphics as much.

I don't think the comparison is being made to other NES games. The comparison is being made to other video games in general. The same year Super Mario Bros came out you had arcade machines like Space Harrier that were creating rather spectacular visuals with a non-tiled approach to graphics, and I think this must have looked liked the leading edge to David Crane, the direction graphics were going for the future. He seems to be saying that the character tile approach was getting old and dying off. (It did, eventually, on the graphics harware side, but not for many years.)

I think something interesting that comes out of this is the realization that tile-based game design was not a foregone conclusion at that point in time. I think what David Crane is getting at is that he felt like the tile grid was an unnatural constraint, that games should not feel like they're being played on a grid, and that with graphics finally starting to escape from these kinds of limitations, SMB comes out and fully embraces a grid-based world. Super Mario Bros. was one of the big hits that made that kind of approach to game building popular. It makes tiles a blatant part of the design instead of working to hide them.

Lots of older games were grid-free, e.g. Joust (1982), or even if they're grid-based they often went to great lengths to hide the system's tile limitations; e.g. Congo Bongo (1983) using a 3/4 perspective for apparent 3D gameplay and graphics to hide the tile grid imposed by the system.

Yeah, I didn't think so.


But that was on expensive arcade hardware, he had to know that home consoles didn't use that kind of technology. Home computers and the cutting edge home consoles still relied heavily on character tiles, it would be unrealistic to expect arcade-like graphics from them.


It certainly defined the standards that platformers would follow for several years. It's funny how David Crane thought that the blocky approach was outdated, when in fact it worked so well in that context that blocks stayed around long after that.

I remember reading somewhere that Sega was originally planning to put a stripped-down version of Super Scaler hardware into the Mega Drive. In addition, SNK's Time Soldiers and P.O.W. boards would soon usher in an arcade platform (Neo Geo MVS) and a console (Neo Geo AES) without tile maps as we know them.


The ColecoVision and MSX had tiles, and we've discussed before how the high quality of the port of Donkey Kong to ColecoVision inspired Nintendo to build the Famicom PPU as a tribute of sorts to the TMS9918. So yes, Nintendo thought it could do arcade-like graphics with tiles.


Yet artists managed to develop art styles that hide the grid. I'd link to a topic about a game that hid the grid fairly well, but it got derailed into a flame war.

I was making a comparison. Space Harrier represents the bleeding edge graphical extreme in 1985.

I don't think David Crane's point was that the NES needed better hardware, but simply that SMB made no attempt to disguise its limitations; instead it emphasized them strongly and turned them into an integral part of how the game was put together, while he and many other developers worked hard to hide the grid.

David Crane made Pitfall on the 2600. I think he must have felt severely constrained, and wanted more freedom of space in game design. In 1985 he had moved on to C64 development, he put out Little Computer People that year. In 1989 he made A Boy and His Blob, and you can see how this game goes to a lot of trouble to not look like a 2D grid; an angled floor, a lot of variation in tiles and organic edges, etc. Here's an example screenshot:


This is his aesthetic. The griddyness of SMB is exactly the kind of thing he was working hard to shake loose from. The comment is really speaking about his desires as a developer, he's not really talking about the hardware, per-se, but about how he himself struggles with it. SMB looked like a cop-out to him. A regression. A step back to earlier days where you had no choice.

Case in point for a "cop-out" becoming "embracing": zzo38's Attribute Zone.

Yes, I did use the word "embrace" as well in my previous post.

I'm not making any judgement on SMB here, I'm just trying to explain the context of David Crane's statement. It makes sense to me when I consider who he was and what he tried to do in his games, and the contemporary state of other games.

What game?

The game was Haunted: Halloween '85. And the lead designer PM'd me yesterday with an offer to restart discussion civilly.

I think the confusion that sparked the original question was the "character" part. See, back in the day the term "tile" wasn't used, normally it was "character" or "cell", since tilemaps were originally used for text mode in computers. So yeah, character = tile.


To be fair, what killed it off was the fact that it didn't translate well to 3D (more specifically, memory usage exploded in comparison to 2D). Using a grid made things much simpler for actually developing the maps and in fact modern 2D games do still usually resort to tilemaps (even if now they have to be "faked" on screen).

The thing with Super Mario Bros though is that it doesn't even attempt to hide the grid look, rather it explicitly outlines the boundaries of the blocks.


They were going to include scaling hardware, but it had to be removed due to running out of die space. The Sega CD was created mainly to include the hardware that was missing from the original design (the CD drive itself was a last minute decision, in fact).