I'm looking for information whether there has ever been an official statement by Commodore back in the days on why the Original Chipset was limited to doing 6 bitplanes only when running in low horizontal resolution.

The Amiga DMA Allocation Channel diagram from the Hardware Reference Manual, 3rd edition: Amiga DMA channel allocation.png clearly shows that 2 more slots are free in low resolution.

Enabling 8 bitplanes in low resolution would have saturated the chip ram bandwidth, much like 4 bitplanes in high resolution, therefore making the machine sluggish as 4 bitplanes high res makes it. But then, why not limit high res to 3 bitplanes?

Adding 2 more bitplane pointers, 2 more shift registers to Denise, and more CLUT registers does not seem to me like a huge hurdle, even in 1984/85.

What other reason could it be?

  • 4
    I think the size of the CLUT was the limiting factor. Increasing it from 32 to 256 entries would have been significant. It's the equivalent of 168 more 16-bit registers (since each CLUT entry was 12 bits). – Brian H May 27 '17 at 19:07
up vote 7 down vote accepted

IIRC, the primary reason for the decision on 6 bit planes was die area. Registers (control and shift, but especially if the LUT was expanded) take up measurable die area. The original bottom end price target was for an under $400 Hi Toro game machine. This allowed only 32 kB of DRAM, and 8 bit plane layers was far more than fit in memory. So why increase costs of the higher volume target?

The higher price target home computer decision came after the chips were floor planned... too late to change things and floorplan larger ASICs, given the extremely limiting R&D budget and aggressive schedule.

Also, hold-and-modify was an option for game designers who wanted to display more colors.

There is no official statement by Commodore because they had no clue how Hi-Toro and Amiga, Inc. designed the system and OCS over a year before their purchase of the technology.

  • Guess that I must accept this as a the correct answer, coming from one of the original Hi-Toro designers :) – user180940 Jun 18 '17 at 21:18
  • @user180940: well, it’s not contradicting. The original 32kB, however, is an interesting additional fact. I always wondered about the original price target, as the first Amiga’s release price was so far away from a home computer… – Holger Jun 19 '17 at 7:41
  • Though now I realize that 32KB is not enough to hold a full 6 bitplane low res screen. 320 x 200 x 6 / 8 requires 48KB. So maybe the 32KB later became 64KB, or maybe there originally was a super-low-res mode with 280ns pixels for showing HAM pics in 32KB ? Or was the part of the cartdrige ROM address map meant to be addressable by Agnus in the original games machine design? – user180940 Jun 20 '17 at 22:07

You are underestimating the on-chip costs of additional colors at that time. Supporting six bitplanes was already maxing out the possibilities, which is best illustrated by the fact that there were not enough CLUT registers even for six bitplanes, but only for five (32 colors), so six bitplanes could only be used via EHB, HAM or Dual Playfield mode.

If adding another 32 CLUT registers was that easy, there was no need for the EHB compromise. Not to speak of adding another 224 color registers…

Besides that, reserving all cycles to the display DMA implies that neither Copper, Blitter nor CPU (in the absence of real FastRAM) could do anything while the display data is transferred, i.e. could only work during the blanking phase and that with even more graphic data to process. Such a display mode is not very useful for real applications (you may try a 256 color SuperHires mode on an AGA machine without FastRAM to get the idea)… So the benefit would not have justified the higher chip costs.

  • Expanding the CLUT would have been expensive, but adding more overlay modes might have been helpful. For example, adding an to overlay a single 640-pixel bitplane onto what would otherwise be a 320-wide mode could have been helpful for a wide range of purposes. – supercat Jun 13 '17 at 22:27
  • @supercat: that special overlay would require chip space too and the fact that it would steal cycles from blitter and cpu still doesn’t change. But of course, we could discuss a lot of modes that could have been possible with the available technology, not knowing which were considered and rejected and which were never considered. – Holger Jun 14 '17 at 6:23
  • The chip space required would have been far less than for expanding the CLUT. It might have been necessary to do something like have the 640-mode layer display alternating pixels from two 320-mode bitplanes, but sufficient bandwidth existed to overlay a single 640-mode bitmap on something like a HAM screen, and drawing text would have been much more efficient than doing all the gymnastics necessary to write text to a HAM screen. – supercat Jun 14 '17 at 14:15
  • 1
    @supercat indeed, in the end Commodore with the Hombre chipset (and probably the AAA before that) tried to implement the very same concept: the chipset had 4 8-bit "byteplanes" that could be combined in many ways, to either give 24-bit chunky (3 byteplanes), 16-bit chunky (2 byteplanes needed) or 8-bit CLUT indexed (1 byteplane). So you could have had a 24 bit playfield with a 8 bit playfield overlayed, or two 16-bit playfields. – user180940 Jun 18 '17 at 21:27

Your Answer

 

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Not the answer you're looking for? Browse other questions tagged or ask your own question.