Super FX Chip

The Super FX runs while the SNES is running but the SNES must halt the Super FX to read the results of the Super FX's calculations. For this reason it is not considered a parallel processor. This is more than just a pedantic observation since such definitions effect import and export between countries. As for effects of system performance, the chip gets halted every 1 to 4 cycles of the SNES' 65C816, meaning that, at best, the whole operation can only deliver slightly less than 20 frames per second. The chip can be overclocked for improved frame rate but this is of course an unofficial hack that could damage the chip, game, and SNES (no commercial games run the chip above 21.477 MHz). Another limitation is 192 horizontal lines of resolution. The SNES displays 256x224 creating an 8 pixel wide frame above and below that the chip cannot access. The SNES still can so, some games fill this area with regular 2D graphics. The SNES also has 512x448 mode which would require very creative solutions to include Super FX effects.

Towards the end of the 80s, Nintendo was in a tough spot with the NES. Better hardware including 16-bit systems were winning market share. But more than just customers, publishers (licensing fees) and developers (quality games) were shifting away as well. In 1990, customers had become accustomed to long lived consoles (the Atari 2600 was 14 years old and still manufactured) OR backwards compatibility (Atari 5200/7800 could play 2600, MegaDrive could play Master System). Nintendo could have offered add-on upgrades for the NES and Famicom systems or through the cartridge interface as CPU chips in the cart or in a pass-through cart (like the later 32X would do for the MegaDrive). Or they could have made a backwards compatible 16-bit system (note, the 65C816 CPU can run 6502 CPU code, Apple made the Apple IIgs, with a 65C816 CPU that could run Apple IIe/c/p software intended for a 6502 CPU). Instead of those options, they decided to do what would become fairly standard in the future; make an entirely new system without considering backwards compatibility. They knew consumers were not ready for that. So, they launched a propaganda campaign that shockingly treated their target audience (kids) as if they were intelligent. They didn't use buzzwords, ("Blast Processing"), make appeals to be cool, or throw confrontational catch phrases at their target audience. Instead they talked about real technology; scaling, rotation, colors, Mode 7. Terms that were essentially real instead of vague language with pliable definitions that could be altered to fit whatever tactic the marketing department decided to try next ("Blast Processing"). Sadly, Nintendo would later decide their customers were stupid and resort to buzzwords ("Morphmation") and catch phrases. Their intelligent approach worked, in the sense that kids correctly understood that the SNES could do things that the NES could not, and it would be better. Better games that could not exist on the NES were implied (and this was a reasonable conclusion). But it failed, because parents were the ones with the money. Many parents didn't see the information provided by Nintendo or didn't care to understand it or simply disbelieved it. A common attitude was that the SNES was only a way for Nintendo to get more money from parents. Perhaps Nintendo should have been even more revealing and direct such info to parents. Launching a new console with no backwards compatibility was a huge risk. They wouldn't be getting money from parents, they'd be losing money for every SNES unit sold. Only after this loss could they hope to make some money back from selling their own games and collecting licensing fees from 3rd party publishers. The profit the NES was providing could be expected to drop off quickly as NES game sales, developers, and publishers shifted to SNES. Only after all that, would they start making a profit. The risky complex plan was not about getting _more_ money from parents. It was about getting all money, hopefully, that parents were already spending instead of a large portion of it going to competitors. They'd happily take more if course, but such wishful thinking was not reasonable marketing. Well, they had a backup plan. Street Fighter II was an SNES exclusive. 6.3 million sales later, it is safe to conclude their backup plan succeeded beautifully.

But, beyond all that, Nintendo had taken steps to delay the next transition; when the SNES became outdated (less competitive). The SNES was designed from the beginning for expendability. The cartridge port had 16 extra pins and there was a general purpose expansion port on the bottom of the unit. It is obvious to think the NES had those those too, they were never used, the SNES expansion options were just as useless. Not, exactly. The NES had 8 extra pins, but they were not connected to anything except the expansion port. Any memory and/or computing hardware in the expansion port may as well reside on the cartridge. Sure, there is a cost advantage to not duplicating hardware for each cartridge for each game that used it. But the Famicom didn't have the extra pins and would have to put it on each cartridge anyhow (games were almost always made for the Famicom then ported). The expansion port, which more or less existed on both NES and Famicom systems, was severely limited. Mainly, it was for controller type accessories and very low bandwidth communication. It only had a single serial bus address pin. It was suitable to load data from a floppy disk. The expansion options on the SNES were planned much better. It had 8 parallel bus address pins and 8 parallel data pins. All of them could write as well as read. The extra pins of the cartridge port have 8 multipurpose pins for parallel data or bus address read and write. Additionally, the extra cartridge pins and expansion port can communicate directly. A CD-ROM expansion was always part of the plan. Not because JVC was creating one for the MegaDrive. No, Nintendo was worried about the PC Engine CD-ROM² (TurboGrafx-CD). The Super FX chip, known as the MARIO chip at that time, was not shown to Nintendo until 1990. It was intended as an NES expansion. However, this was almost exactly the type of expansion Nintendo had designed the SNES to utilize. Nintendo wanted to include the MARIO chip as a co-processor in their CD-ROM add-on. In this configuration, it could have been used as a co-processor for CD games as well as for cartridge games (as the cartridge could directly access hardware attached to the EXP port). If this had happened, the chip may have been forever stuck at 10.74 with an 8Mbit limit. Then Nintendo went with a 32-bit design. The MARIO chip didn't really add much to the new 32-bit design so it was dropped. But, not forgotten. Nintendo would play the long game with the MARIO chip and release the first Super FX game, Starfox, in 1993, late in the 32-bit era. The Super FX chips along with other enhancement chips and some notable quality development by 3rd party companies helped to extend the life of the SNES so that new games were still being published in 1999 (a few years after the N64 was released)

Multiply two positive 8-bit numbers on a 65C816 (assembly code).
LDA NUM
STA RESULT
LDA NUM+1
STA RESULT+1
ASL RESULT
ROL RESULT+1
ASL RESULT
ROL RESULT+1
CLC
LDA NUM
ADC RESULT
STA RESULT
LDA NUM+1
ADC RESULT+1
STA RESULT+1
ASL RESULT
ROL RESULT+1

Multiply two positive 8-bit numbers on a GSU-1 (assembly code).
FROM R5
UMULT R1

Not just fewer instructions, way fewer clock cycles. And, the GSU ran 3 to 6 cycles while the 65C816 was running 1.

Multiply two positive or negative 16-bit numbers on a GSU-1 (assembly code)?
FROM R5
LMULT R1

Multiply two positive or negative 16-bit numbers on a 65C816 (assembly code)?
LDA #$05
LDX #$03
STA Result
STX Temp
LDA $C000
ASL A
STA $C000
LDA $C001
ASL A
STA $C001
LDA $C000
CLC
ADC $C001
STA $C002

(24 cycles)