So, Direct Colour . . .

[Kirk_Johnston]

First time I've seen this page, https://sneslab.net/wiki/Direct_Color, which finally makes it's much clearer to me how it actually works: Basically, you can choose between one of eight slightly different 256-colour palettes to use per tile (these palettes use direct RGB values and come from VRAM rather than the normal 4-256-colour CGRAM palettes [depending on the BG Mode] made from the full 32,768-colour master palette), and you can apparently show up to 2041 visible colours on that single background layer alone. And that's along with the additional 120 visible CGRAM colours you can use for the second layer and then the further 120 visible CGRAM colours for sprites on top of that. That's potentially a whole lot of colours on-screen on SNES there [before using any colour math or HDMA and the like], and, importantly, all totally useable during actual normal gameplay with multiple BG layers and sprites and most of the other SNES effects like colour math and window/shape masks and so on (careful management of resources will get the best results there), rather than maybe just on some static high-colour image or whatever.

 

Now, I wonder what could be done with that. . . .

Edited December 28, 2023 by Kirk_Johnston

[jeffythedragonslayer]

55 minutes ago, Kirk_Johnston said:

2041 visible colours on that single background layer alone.

I think this is a typo - dunno where 2041 came from.  In subparagraph 8.1 it says 2048:

 

https://archive.org/details/SNESDevManual/book1/page/n73

 

I suppose you got 120 by subtracting 8 from 128?  I don't think you need to subtract 8...

 

Take a look at... https://patrickjohnston.org/ASM/ROM data/snestek.htm#snesppucolorpalettememorycgramanddirectcolors fullsnes is a much more mature document than most stuff on SnesLab; you should generally trust it more.  Some people trust it even more than the official SNES developer manual.

Edited December 28, 2023 by jeffythedragonslayer

[Kirk_Johnston]

59 minutes ago, jeffythedragonslayer said:

I think this is a typo - dunno where 2041 came from.  In subparagraph 8.1 it says 2048:

 

https://archive.org/details/SNESDevManual/book1/page/n73

 

I suppose you got 120 by subtracting 8 from 128?  I don't think you need to subtract 8...

 

Take a look at... https://patrickjohnston.org/ASM/ROM data/snestek.htm#snesppucolorpalettememorycgramanddirectcolors fullsnes is a much more mature document than most stuff on SnesLab; you should generally trust it more.  Some people trust it even more than the official SNES developer manual.

I was thinking it would be because you normally have to have one transparent colour in every palette, so 2048-8 would be 2040, and then the backdrop colour would take it to 2041. No?

 

But, hey, if 2048 is actually the correct answer, I'll take it.

 

And, I'll be honest, that SnesLab page is at least far more intuitively readable for me personally, which is helpful.

Edited December 28, 2023 by Kirk_Johnston

[Wayler]

Mhhm, yep. You finally stumbled on to something. This will finally blow the door wide open to.... .an endless quagmire bullshit topics about fucking nothing. Whee SNES. 

[Cris1997XX]

Kirk sure likes his nothingburgers, huh?

[jeffythedragonslayer]

26 minutes ago, Kirk_Johnston said:

you normally have to have one transparent colour in every palette

This I want to find more sources that confirm.  Where did you read it?  That explanation doesn't sound completely nonsensical but I still think 2048 is the correct answer because the number of palettes CGRAM is divided into depends on the BG mode.  And, direct select doesn't even use CGRAM.

 

The manual very clearly states that BG1 in modes 3 & 4 supports 2048 colors, so it's going to take a lot of evidence to convince me it's only 2041.

 

Now I think there's one caveat to the "direct color doesn't use CGRAM" mantra, and that is I think the note at the very bottom of Appendix A-17 about CGRAM address $00 being used as the backdrop when all the bits in the direct select color datum are zero.  I'd argue the manual has mistake there in saying the color 0 will be used as the "background" as I think "backdrop" would be clearer there.  BG1 is a background layer, and any piece of that will allow the backdrop (color 0) is shine through.

Edited December 28, 2023 by jeffythedragonslayer

[Kirk_Johnston]

24 minutes ago, jeffythedragonslayer said:

This I want to find more sources that confirm.  Where did you read it?  That explanation doesn't sound completely nonsensical but I still think 2048 is the correct answer because the number of palettes CGRAM is divided into depends on the BG mode.  And, direct select doesn't even use CGRAM.

 

The manual very clearly states that BG1 in modes 3 & 4 supports 2048 colors, so it's going to take a lot of evidence to convince me it's only 2041.

 

Now I think there's one caveat to the "direct color doesn't use CGRAM" mantra, and that is I think the note at the very bottom of Appendix A-17 about CGRAM address $00 being used as the backdrop when all the bits in the direct select color datum are zero.  I'd argue the manual has mistake there in saying the color 0 will be used as the "background" as I think "backdrop" would be clearer there.  BG1 is a background layer, and any piece of that will allow the backdrop (color 0) is shine through.

Ah, interesting, because this is just something almost everyone I've talked to in the SNES development community has told me since day one, that the first colour in every palette has to be transparent. And I've even asked if that's only if you want to have some parts of the tiles that use that palette be not filled in, but even then they've told me you have to have one colour be transparent no matter what. At least that's how it came across to me all those times it's been explained. So, this will be something new for me if it's not actually necessary to have the first palette entry in each palette be transparent, even if it's just in this specific direct colour mode. Learn something new every day. 

Edited December 28, 2023 by Kirk_Johnston

[jeffythedragonslayer]

4 minutes ago, Kirk_Johnston said:

Ah, interesting, because this is just something almost everyone I've talked to in the SNES development community has told me since day one, that the first colour in every palette HAS to be transparent. And I've even asked if that's only if you want to have some parts of the tiles that use that palette be not filled in, but even then they've told me you just HAVE to have one colour be transparent no matter what. At least that's how it came across to me all those times it's been explained. So, this will be something new for me if it's not actually necessary to have the first palette entry in each palette be transparent, even if it's just in this specific direct colour mode. Learn something new every day. 

Yeah, me too, but it would be nice to see where Nintendo themselves say that.

[Tanooki]

1 hour ago, Cris1997XX said:

Kirk sure likes his nothingburgers, huh?

Maybe we should all chip in and get him a nice box of Nothingburger Helper?

[Kirk_Johnston]

It's interesting that this SNES Direct Colour mode has been almost completely ignored historically and indeed even very much downplayed by some people in recent times too (much like a bunch of other cool SNES modes and features as well actually). I wonder why, when it's pretty clear to me that it's totally useable if you just design whatever SNES games/levels/demos around its particular details, which is kinda the rule for all game development. I mean, really, it's not that hard to at least think of potentially interesting and cool ways to use these various SNES modes and features I'm learning more about every day. Even as just some "back of the box" USP it's compelling to some degree imo.

 

And, I'll say right now that the idea of having a gorgeous 8bpp scrolling background in one fully playable level of a SNES game or whatever, or having a level with a load of row/line scrolled parallax plus so many fully overlapping parallax layers with layer priority shifting on SNES that it goes well beyond anything we've seen on these 16-bit era systems to date, or seeing a bunch of lovely high-colour SNES images that look nicer than pretty much everything we've seen to date in any commercial games on these old 16-bit era systems, or now thinking of some interesting ways to use this Direct Colour mode on SNES, etc, is certainly nothing for any genuine SNES fan and/or genuine SNES developer to scoff at in the slightest.

 

So, I'll just be real honest and say it's actually pretty dang disappointing on a personal level for me repeatedly discovering all these cool modes and capabilities of SNES have basically been largely ignored/overlooked/dismissed all this time, especially if that's even remotely coming from someone inside the SNES development community, never mind anyone on the outside who maybe has their own motivations and agendas to not want to see all these oft-ignored/overlooked SNES capabilities highlighted, explored, and even fully exploited ultimately.

 

Maybe/hopefully that's going to change going forward though as the modern SNES indie/homebrew scene continues to grow and mature and more people move from just learning about SNES' inner workings, maybe trying a few standard tests to see what they can do with it, to actually wanting to make it shine. . . .

Edited December 28, 2023 by Kirk_Johnston

[jeffythedragonslayer]

30 minutes ago, Kirk_Johnston said:

It's interesting that this SNES Direct Colour mode has been almost completely ignored historically and indeed even very much downplayed by some people in recent times too (much like a bunch of other cool SNES modes and features as well actually). I wonder why, when it's pretty clear to me that it's totally useable if you just design whatever SNES games/levels/demos around its particular details, which is kinda the rule for all game development. I mean, really, it's not that hard to at least think of potentially interesting and cool ways to use these various SNES modes and features I'm learning more about every day. Even as just some "back of the box" USP it's compelling to some degree imo.

 

And, I'll say right now that the idea of having a gorgeous 8bpp scrolling background in one fully playable level of a SNES game or whatever, or having a level with a load of row/line scrolled parallax plus so many fully overlapping parallax layers with layer priority shifting on SNES that it goes well beyond anything we've seen on these 16-bit era systems to date, or seeing a bunch of lovely high-colour SNES images that look nicer than pretty much everything we've seen to date in any commercial games on these old 16-bit era systems, or now thinking of some interesting ways to use this Direct Colour mode on SNES, etc, is certainly nothing for any genuine SNES fan and/or genuine SNES developer to scoff at in the slightest.

 

So, I'll just be real honest and say it's actually pretty dang disappointing on a personal level for me repeatedly discovering all these cool modes and capabilities of SNES have basically been largely ignored/overlooked/dismissed all this time, especially if that's even remotely coming from someone inside the SNES development community, never mind anyone on the outside who maybe has their own motivations and agendas to not want to see all these oft-ignored/overlooked SNES capabilities highlighted, explored, and even fully exploited ultimately.

 

Maybe/hopefully that's going to change going forward though as the modern SNES indie/homebrew scene continues to grow and mature and more people move from just learning about SNES' inner workings, maybe trying a few standard tests to see what they can do with it, to actually wanting to make it shine. . . .

Direct Color isn't as obscure as Mode 6:

 

https://snes.nesdev.org/wiki/Uncommon_graphics_mode_games#Direct_Color

 

It sounds easier for people familiar with PC game dev to use than indirect color to me.

Edited December 28, 2023 by jeffythedragonslayer

[jeffythedragonslayer]

2 hours ago, Tanooki said:

Maybe we should all chip in and get him a nice box of Nothingburger Helper?

Now you're just making me hungry...

[Kirk_Johnston]

22 minutes ago, jeffythedragonslayer said:

Direct Color isn't as obscure as Mode 6:

 

https://snes.nesdev.org/wiki/Uncommon_graphics_mode_games#Direct_Color

 

It sounds easier for people familiar with PC game dev to use than indirect color to me.

Mode 6--now that's that rarest of them all. I have a couple of ideas to at least try with this, but it is still one I've yet to think of some genuinely cool way to show it off. I do think it could be used in some way(s) though, even if for nothing other than a basically a rippling water wave effect or something like that [on part of the screen], but one that's at least smoother than can be achieved in Modes 2 and 4, and certainly much smoother than if you were doing it with even wider columns than SNES can vertically scroll normally. So, yeah, some potential there.

[jeffythedragonslayer]

9 minutes ago, Kirk_Johnston said:

Mode 6--now that's that rarest of them all. I have a couple of ideas to at least try with this, but it is still one I've yet to think of some genuinely cool way to show it off. I do think it could be used in some way(s) though, even if for nothing other than a basically a rippling water wave effect or something like that [on part of the screen], but one that's at least smoother than can be achieved in Modes 2 and 4, and certainly much smoother than if you were doing it with even wider columns than SNES can vertically scroll normally. So, yeah, some potential there.

That reminds me, would anyone be interested in getting the answer to how taxing each of the PPUs features are?  Might lead to some insights about why which background modes support what.

 

I think the key to solving that is system of linear equations.  You get the same time budget every scanline.

[KulorXL]

1 hour ago, Kirk_Johnston said:

So, I'll just be real honest and say it's actually pretty dang disappointing on a personal level for me repeatedly discovering all these cool modes and capabilities of SNES have basically been largely ignored/overlooked/dismissed all this time, especially if that's even remotely coming from someone inside the SNES development community

As an artist, you could actually do something to help shift the perspective on direct color mode. Draw up some cool stuff using direct color mode limitations that shows off what it can do. The only thing to keep in mind is, whatever you come up with has to look better than it would in either 4bpp or in indexed 8bpp modes. That should be completely doable for someone without any need for programming knowledge.

[jeffythedragonslayer]

A direct color screenshot from the maker of those palette diagrams:

 

https://www.smwcentral.net/?p=viewthread&t=125133&page=1&pid=1625490#p1625490

[Wayler]

There's a familiar pattern to all these threads. 

 

1. Read about something-something related to SNES capabilities

2. Understand about 20% of the concept but still go over various use-cases how this feature should be implemented. 

3. Throw the original developers and the current dev scene under the bus for not using said feature in anything (although you have no idea if it has been used). 

4. Ignore any further information about the subject on the comments. jeffytheerrandboy will most likely quote them at some point if you need reality for some reason. 

5. Remember to mention "your" shoot 'em up demo and how this newly discovered ability will totally be used in it when its finished for real. 

6. Don't forget that untapped potential and give hope to the future as now the information is out there for all to see as you made it possible. 

7.

8. Repeat

[jeffythedragonslayer]

31 minutes ago, Wayler said:

There's a familiar pattern to all these threads. 

 

1. Read about something-something related to SNES capabilities

2. Understand about 20% of the concept but still go over various use-cases how this feature should be implemented. 

3. Throw the original developers and the current dev scene under the bus for not using said feature in anything (although you have no idea if it has been used). 

4. Ignore any further information about the subject on the comments. jeffytheerrandboy will most likely quote them at some point if you need reality for some reason. 

5. Remember to mention "your" shoot 'em up demo and how this newly discovered ability will totally be used in it when its finished for real. 

6. Don't forget that untapped potential and give hope to the future as now the information is out there for all to see as you made it possible. 

7.

8. Repeat

and where exactly do your posts fit into that sequence?

[jeffythedragonslayer]

Dropping this here from Appendix A-17, that is where the subtracting 8 for the number of visible CGRAM colors came from (sorry for that seed of doubt; finding this in the bsnes source would be good enough for me tho to accept not having Nintendo just word-of-god the answer)

[KulorXL]

Since there seems to be a lot of confusion about those extra 8 colors...

https://sneslab.net/wiki/Direct_Color

^If you actually take a look at the palettes here...

https://sneslab.net/wiki/File:Direct_Color_Mode_Palette_0.png

...you'll see that they each have 255 unique colors, and where pure black should be is transparent. This is because a value of 0 signals to the PPU to bleed through to layers, sprites, or the background color behind the direct color layer. So, if you put every single color possible in direct color mode on the screen, it would be 255 * 8 = 2040 unique colors for that layer. In other words, if you want pure black on a direct color layer, it has to be the background color, which would be the 2041st color. Otherwise without the background color, the closest you get is a very dark red, blue, or gray.

[jeffythedragonslayer]

3 minutes ago, KulorXL said:

Since there seems to be a lot of confusion about those extra 8 colors...

https://sneslab.net/wiki/Direct_Color

^If you actually take a look at the palettes here...

https://sneslab.net/wiki/File:Direct_Color_Mode_Palette_0.png

...you'll see that they each have 255 unique colors, and where pure black should be is transparent. This is because a value of 0 signals to the PPU to bleed through to layers, sprites, or the background color behind the direct color layer. So, if you put every single color possible in direct color mode on the screen, it would be 255 * 8 = 2040 unique colors for that layer. In other words, if you want pure black on a direct color layer, it has to be the background color, which would be the 2041st color. Otherwise without the background color, the closest you get is a very dark red, blue, or gray.

That explanation makes so much sense it has to be where the number 2041 came from.  That upper left cell in the PNG even seems to have transparency enabled on it.

[turboxray]

4 hours ago, Kirk_Johnston said:

First time I've seen this page, https://sneslab.net/wiki/Direct_Color, which finally makes it's much clearer to me how it actually works: Basically, you can choose between one of eight slightly different 256-colour palettes to use per tile (these palettes use direct RGB values and come from VRAM rather than the normal 4-256-colour CGRAM palettes [depending on the BG Mode] made from the full 32,768-colour master palette), and you can apparently show up to 2041 visible colours on that single background layer alone. And that's along with the additional 120 visible CGRAM colours you can use for the second layer and then the further 120 visible CGRAM colours for sprites on top of that. That's potentially a whole lot of colours on-screen on SNES there [before using any colour math or HDMA and the like], and, importantly, all totally useable during actual normal gameplay with multiple BG layers and sprites and most of the other SNES effects like colour math and window/shape masks and so on (careful management of resources will get the best results there), rather than maybe just on some static high-colour image or whatever.

And there it is! I've been wondering when you were going to make this post. For the past two years.. you've been on forums, YT comments, twitter, etc.. blasting all these specs the SNES has.. always listing the "2048 color mode". Even on spritesmind when I told you this is not a real 2048 color, explained to you why, and even asked you to provide art proving otherwise... nothing from you.

 

Quote

It's interesting that this SNES Direct Colour mode has been almost completely ignored historically and indeed even very much downplayed by some people in recent times too (much like a bunch of other cool SNES modes and features as well actually). I wonder why, when it's pretty clear to me that it's totally useable if you just design whatever SNES games/levels/demos around its particular details, which is kinda the rule for all game development.

 That's because people that know a lot more than you, understand the limitations of this mode. For some reason though, you just can't get it through your head why this presents such a huge limitation. And this, especially in the context that you can do soo much more in other modes (including indirect color mode).. makes this about 99% useless. 

 

Quote

I mean, really, it's not that hard to at least think of potentially interesting and cool ways to use these various SNES modes and features I'm learning more about every day. Even as just some "back of the box" USP it's compelling to some degree imo.

 

Then do it! Put your money where your mouth is; prove this mode is worth something. Show the example that everyone else is missing. Go ahead. Mock something up. You have photoshop. Come up with something.

 

4 hours ago, Kirk_Johnston said:

Now, I wonder what could be done with that. . . .

Pretty much nothing. Congrats on NOT understanding this mode or even the link you posted. 

 

Let's put this into perspective:

 

 1st: Direct color mode means there is no palettes.. no expanded colors to pick from (can't pick from that 32k palette)... you get a "fixed" 256 colors out of.. 256 colors. You don't get more.. and you don't get 2048. Let's get that out of the way. If doesn't not matter how you "tint" those 256 colors.. it's still a fixed 256 colors. Period. Put that into perspective of the PCE and MD.. they have 512 colors as a base palette.. this is half of that. In this direct mode, as-is, you have LESS than both those systems to choose from.

 

 2nd: 2048 is a farse. It's fake. It's just a number. It is THEE most misleading thing ever mentioned by you, in regards to the SNES. Because the bits needed to expand the direct color (3:3:2) colors of 256... is limited to a single tile of 8x8 area. If you set the one blue bit for the tile.. then the whole tile gets the blue bit. Congrats.. you managed to "tint" a single tile.. and ALL colors in the tile. Okay, so you managed to tint a tile with a little more blue.. can you imagine the attribute color clash for such art? Hahahaha. Let's not forget now, not only did you manage to affect ALL colors in the tile.. but still have that same 256 color set (not the beautiful color set coming from 32k master palette)

 

 This mode does NOTHING for color fidelity.. all the while THROWING away the 32k master color palette for a fixed 256 color palette potentially "tint" on a seriously course basis. Let me repeat that in words you can understand - you literally threw away EVERYTHING that gives the SNES its strength in the color department, even to the point where it has less colors than the PCE/MD overall.. in a trade off for color density. Not worth it, when you have the option for 256 color mode out of 32k colors... which is orders of magnitude more capable.

 

3rd: So direct color format 3:3:2.. or 8colors : 8colors : 4colors.... in RGB format. Blue.. is two 2bits. That means out of the four possible combinations of 2 bits, with one being black (absent of blue).. you only have three shades of blue to mix with. THREE shades. And it's not evenly divided either.. it's like a typical 2bit ramp which would be 0,85, 170, 255. Noooo. It's 0, 64, 128, 192. See that? 192.. not even in the upper range close to 255. Let's put that into perspective, again, against the normal master palette range of 5:5:5 or 32colors:32colors:32colors. I feel like words don't reach you, so let's get a visual comparison:

The top half is the 3:3:2 RGB (in GRB order) and the lower half in 5:5:5 RGB (in GRB order). See that column on the far right side? That's your pure grayscale range. It's 32 gray scale colors for native palette... and 4 for the direct color mode palette (that's Master System level right there). Here's the thing, when you tint (like I've said before) with the palette bits in the tilemap entry.. you tint everything. You're either going to get some level of attribute clash from tile to tile, or the offset bits will be so subtle that they won't really be perceivable.. because again.. the entire set of colors in a tile will be affected by it. So unless your single pixels are the size of an 8x8 tile, parroting around the "2048 color mode" as if it was legit 2048 usable colors is the most egregious misleading thing I've ever seen anyone ever say when it comes to specs in retro systems. And I've seen a lot of contorted logic in console war arguments hahah. I really want to drive this home because "256 indirect color mode out of 32k palette" gives you everything "256 direct color mode" does and sooooo much more. Even when you limit the colors to 128 for indirect mode, to make room for max sprite palettes.. it still gives you so much more.

 

 4th: I would love to see YOU produce any art in this "2048" color mode.. and I want a lot more colors than 256 too. Better get on it haha. There's a very good reason why rilden did not bother with this in their TQ tool. I mean, you can't even have a single consistent backdrop color. I converted rilden's JS source code to python. I could easily add this mode.. but I won't. Because it's useless. Adding a 3:3:2 mode might be useful, but not this 2048 mode drivel business. Luckily I have my own tool that can do 3:3:2 if I need it.

 

 So yeah, it's not a 2048 color mode. It's just a 256 fixed color mode. 

 

 

 So here's how the mode has been used: color math. You take the direct color mode, ignore the palette "tint bits" in the tilemap, and you apply the 2bpp layer via color math. This gives you a "2048" color mode.. ala 10bit pixel data.. out of a fixed 2048 colors. This HAS been done and WILL eat the ever-living-shit out of your vram. And the added effect that you effectively have a single BG layer. Again, congrats on eating your vram up and reducing the system to a single BG layer. You could do the same thing with indirect color, and get a WAY better color combination for the same vram space.

 

For the record, I'm not saying 3:3:2 direct color mode is 100% completely useless. It does have a few tiny use-cases but they are super edge cases - literally, it would have to provide something that gives a reason to throw away ALL that the SNES offers just to this.. and this, over 8bpp indirect color mode (which is what Kirkhead loves making videos about). Literally the use case is this: I don't need more than a 256 fixed color palette (half that of PCE/MD), but I need high color density in a per tile basis, and I need it to not impact sprite palettes. That's the use case. No matter the argument, you need an incredible reason to overcome the largest limitation of this mode.

 

 Yes, I know you can use large tilemaps along with an hmda routine to sub-divide the offset bits for a tile to be pseudo 8x4 or 8x2 segment association. Still doesn't change anything that I've said or showed.

 

 If Kirk had any inclination about graphic modes and colors, and not just parrot what he sees in YT videos.. he might come to the conclusion that you could do a pseudo 6bit indirect tile mode which gives you a 6bpp tile layer and a 4bit tile layer, and it doesn't involve eating into your sprite palette area. And 6bpp eats up less vram than 8bpp. Let's see if he can figure it out.

 

 

 

[jeffythedragonslayer]

23 minutes ago, turboxray said:

And there it is! I've been wondering when you were going to make this post. For the past two years.. you've been on forums, YT comments, twitter, etc.. blasting all these specs the SNES has.. always listing the "2048 color mode". Even on spritesmind when I told you this is not a real 2048 color, explained to you why, and even asked you to provide art proving otherwise... nothing from you.

 

 That's because people that know a lot more than you, understand the limitations of this mode. For some reason though, you just can't get it through your head why this presents such a huge limitation. And this, especially in the context that you can do soo much more in other modes (including indirect color mode).. makes this about 99% useless. 

 

 

Then do it! Put your money where your mouth is; prove this mode is worth something. Show the example that everyone else is missing. Go ahead. Mock something up. You have photoshop. Come up with something.

 

Pretty much nothing. Congrats on NOT understanding this mode or even the link you posted. 

 

Let's put this into perspective:

 

 1st: Direct color mode means there is no palettes.. no expanded colors to pick from (can't pick from that 32k palette)... you get a "fixed" 256 colors out of.. 256 colors. You don't get more.. and you don't get 2048. Let's get that out of the way. If doesn't not matter how you "tint" those 256 colors.. it's still a fixed 256 colors. Period. Put that into perspective of the PCE and MD.. they have 512 colors as a base palette.. this is half of that. In this direct mode, as-is, you have LESS than both those systems to choose from.

 

 2nd: 2048 is a farse. It's fake. It's just a number. It is THEE most misleading thing ever mentioned by you, in regards to the SNES. Because the bits needed to expand the direct color (3:3:2) colors of 256... is limited to a single tile of 8x8 area. If you set the one blue bit for the tile.. then the whole tile gets the blue bit. Congrats.. you managed to "tint" a single tile.. and ALL colors in the tile. Okay, so you managed to tint a tile with a little more blue.. can you imagine the attribute color clash for such art? Hahahaha. Let's not forget now, not only did you manage to affect ALL colors in the tile.. but still have that same 256 color set (not the beautiful color set coming from 32k master palette)

 

 This mode does NOTHING for color fidelity.. all the while THROWING away the 32k master color palette for a fixed 256 color palette potentially "tint" on a seriously course basis. Let me repeat that in words you can understand - you literally threw away EVERYTHING that gives the SNES its strength in the color department, even to the point where it has less colors than the PCE/MD overall.. in a trade off for color density. Not worth it, when you have the option for 256 color mode out of 32k colors... which is orders of magnitude more capable.

 

3rd: So direct color format 3:3:2.. or 8colors : 8colors : 4colors.... in RGB format. Blue.. is two 2bits. That means out of the four possible combinations of 2 bits, with one being black (absent of blue).. you only have three shades of blue to mix with. THREE shades. And it's not evenly divided either.. it's like a typical 2bit ramp which would be 0,85, 170, 255. Noooo. It's 0, 64, 128, 192. See that? 192.. not even in the upper range close to 255. Let's put that into perspective, again, against the normal master palette range of 5:5:5 or 32colors:32colors:32colors. I feel like words don't reach you, so let's get a visual comparison:

The top half is the 3:3:2 RGB (in GRB order) and the lower half in 5:5:5 RGB (in GRB order). See that column on the far right side? That's your pure grayscale range. It's 32 gray scale colors for native palette... and 4 for the direct color mode palette (that's Master System level right there). Here's the thing, when you tint (like I've said before) with the palette bits in the tilemap entry.. you tint everything. You're either going to get some level of attribute clash from tile to tile, or the offset bits will be so subtle that they won't really be perceivable.. because again.. the entire set of colors in a tile will be affected by it. So unless your single pixels are the size of an 8x8 tile, parroting around the "2048 color mode" as if it was legit 2048 usable colors is the most egregious misleading thing I've ever seen anyone ever say when it comes to specs in retro systems. And I've seen a lot of contorted logic in console war arguments hahah. I really want to drive this home because "256 indirect color mode out of 32k palette" gives you everything "256 direct color mode" does and sooooo much more. Even when you limit the colors to 128 for indirect mode, to make room for max sprite palettes.. it still gives you so much more.

 

 4th: I would love to see YOU produce any art in this "2048" color mode.. and I want a lot more colors than 256 too. Better get on it haha. There's a very good reason why rilden did not bother with this in their TQ tool. I mean, you can't even have a single consistent backdrop color. I converted rilden's JS source code to python. I could easily add this mode.. but I won't. Because it's useless. Adding a 3:3:2 mode might be useful, but not this 2048 mode drivel business. Luckily I have my own tool that can do 3:3:2 if I need it.

 

 So yeah, it's not a 2048 color mode. It's just a 256 fixed color mode. 

 

 

 So here's how the mode has been used: color math. You take the direct color mode, ignore the palette "tint bits" in the tilemap, and you apply the 2bpp layer via color math. This gives you a "2048" color mode.. ala 10bit pixel data.. out of a fixed 2048 colors. This HAS been done and WILL eat the ever-living-shit out of your vram. And the added effect that you effectively have a single BG layer. Again, congrats on eating your vram up and reducing the system to a single BG layer. You could do the same thing with indirect color, and get a WAY better color combination for the same vram space.

 

For the record, I'm not saying 3:3:2 direct color mode is 100% completely useless. It does have a few tiny use-cases but they are super edge cases - literally, it would have to provide something that gives a reason to throw away ALL that the SNES offers just to this.. and this, over 8bpp indirect color mode (which is what Kirkhead loves making videos about). Literally the use case is this: I don't need more than a 256 fixed color palette (half that of PCE/MD), but I need high color density in a per tile basis, and I need it to not impact sprite palettes. That's the use case. No matter the argument, you need an incredible reason to overcome the largest limitation of this mode.

 

 Yes, I know you can use large tilemaps along with an hmda routine to sub-divide the offset bits for a tile to be pseudo 8x4 or 8x2 segment association. Still doesn't change anything that I've said or showed.

 

 If Kirk had any inclination about graphic modes and colors, and not just parrot what he sees in YT videos.. he might come to the conclusion that you could do a pseudo 6bit indirect tile mode which gives you a 6bpp tile layer and a 4bit tile layer, and it doesn't involve eating into your sprite palette area. And 6bpp eats up less vram than 8bpp. Let's see if he can figure it out.

 

 

 

That's enough evidence for me to conclude Nintendo messed up in calling direct color a 2048 color mode.  Thanks, @turboxray and @KulorXL.

 

When you have that much blank space on a page in your manual, there should really be some caveats listed, like "some of these are not real visible colors but fake, transparent colors"

Edited December 28, 2023 by jeffythedragonslayer

[KulorXL]

1 minute ago, jeffythedragonslayer said:

That's enough evidence for me to conclude Nintendo messed up in calling direct color a 2048 color mode.  Thanks, @turboxray and @KulorXL.

Yeah, I agree with the assessment that it's much more like a 255-color mode (with a fixed 255-color palette spanning the entire spectrum which you would probably not make full use of for real art assets). You *can* put 2041 colors on-screen with it, but not in much of a useful way, unless you really like displaying color wheels or something. If you're just trying to bang out unique on-screen colors for the numbers' sake, there's other ways to get many more simultaneous colors than that...

Totally open to novel uses of direct color though, so like I said, Kirk should draw up some artwork and show us what it's capable of.

[jeffythedragonslayer]

3 minutes ago, KulorXL said:

Yeah, I agree with the assessment that it's much more like a 255-color mode (with a fixed 255-color palette spanning the entire spectrum which you would probably not make full use of for real art assets). You *can* put 2041 colors on-screen with it, but not in much of a useful way, unless you really like displaying color wheels or something. If you're just trying to bang out unique on-screen colors for the numbers' sake, there's other ways to get many more simultaneous colors than that...

Totally open to novel uses of direct color though, so like I said, Kirk should draw up some artwork and show us what it's capable of.

Yeah, if there is better scene slang we can come up with that will reduce confusion I'm all for that instead of blindly submitting to whatever Nintendo called it.  And I agree on the other point to, I really don't like the idea of shutting down research into obscure features like Mode 6 just because they're unpopular.

 

Did you notice I edited my last post after you liked it?  I could've made it say anything...