The PC Engine's CPU is probably its strongest aspect, Hudson did a great job here. Going with a really fast 8-bit CPU makes much more sense than Nintendo's choice of going with a slow 16-bit CPU, especially when the screen is 256 pixels wide so most of the calculations a game would be doing are only 8 bits anyway. Even when you have to do 16-bit calculations, the SNES CPU still has an 8-bit data bus so it doesn't have much of an advantage over the PC Engine (all 16-bit operations take extra cycles compared to their 8-bit counterparts). In practice, fast action games like shooters play a lot more smoothly on the PC Engine than on the SNES. This problem is exacerbated by the fact that the SNES CPU provides the option of running at ~3/4 speed if the publisher didn't want to pay for faster ROMs, so a lot of games have unnecessary slowdown because the publisher wanted to save 50 cents per cartridge.
Where the PC Engine ends up weaker is in the video processor. That generation of consoles was able to do more with the video hardware than previous generations (like the NES and Master System) due to faster RAM becoming available. The PC Engine used the additional bandwidth to make it possible for the CPU to access video memory while the screen is being drawn, while the SNES kept the restriction that you can only access video memory during vblank and instead used the additional bandwidth for more background layers. Being able to access video memory all the time is definitely a useful feature, but the result is that PC Engine games often look more flat than SNES games.
As someone who owns a SNES and a TG-16, I think I disagree. The TG-16 graphics really pop compared to the SNES.
The SNES can only render 256 simultaneous colours. The TG-16 could do twice that at 512. Its video processor was also full 16-bit.
I’m not sure where you’re getting that the video was the weakness of the TG-16. At the time, that was the Turbografx’s whole claim to fame in that it was superior graphically to the SNES.
Source: Old enough to remember the commercials and bought both consoles to compare like nerds did back then
The PC Engine could theoretically do more colours on screen (though things like colour math and mid-screen palette changes do complicate that a bit...), but only had 9-bit wide palette entries compared to the SFC's 15-bit colour depth, which allowed for a lot more subtlety and richness.
That's the biggest Achilles' heel of the Genesis/MD, too. The lack of color resolution. I helped make the palette more accurate for that MD port of Super Mario Bros, and I made two palettes - one to simulate the composite PPU palette, and one to simulate the RGB PPU palette. The composite palette was closer, and what Mairtrus ended up shipping with (the RGB palette, no matter which "side" of mid-values I chose, wasn't quite right due to the lack of color bits)
Its big weakness is that they didn't realize how much multilayered backgrounds made things pop.
The sprites look really good, but the single background layer makes things look flat compared to the SNES / Genesis. Devs started working around that by updating some of the background tiles on each frame or using some of the sprites.
The SNES also could do layer blending effects. So while all of the input data could only be 256 colours, the PPU was spitting out a high colour signal after the effects were applied.
The CPU was conceived by Hudson engineers who, despite their success in the NES business, felt they had hit the hardware's limits. Interestingly, they initially had no business plan. They were simply "chasing a dream" and supposedly had "no real intention of selling it," which led most semiconductor manufacturers to reject their pitch.
They managed to scrape together a few thousand samples, and by pure chance, NEC—a dominant PC manufacturer in Japan at the time—was looking to enter the console market. This lucky alignment of interests is what gave birth to the PC Engine.
While this story sounds reckless, it was only possible because Japan was in the midst of the real estate bubble, and capital was overflowing.
Once the bubble burst in the 90s, Japan entered a long recession. After a few major bank failures, Hudson’s funding dried up. They were eventually absorbed by Konami, and today, even the Hudson name is gone. It’s a real shame.
According to Wikipedia, Hudson originally approached Sharp before pitching the idea to NEC. While Sharp was enthusiastic and agreed that it had great commercial potential, the deal ultimately fell through.
The deal-breaker was Sharp’s deep relationship with Nintendo at the time. Apparently, developing a console with Hudson’s CPU was seen as something that would have jeopardized their partnership with Nintendo.
What Sharp was working on back then was the "C1 NES TV"—basically a NES-integrated television. You could think of it as the NES era's version of the iMac. It has a bit of a comical look that always makes me smile.
The Japanese version of that TV used an internal RGB connection with an RGB PPU, making it as clear and sharp as Nintendo's Famicom-based arcade hardware.
The US version of that TV used composite internally
Remember that the PC Engine came out in 1987, over three years before the SFC did. That makes Nintendo's CPU decisions even more questionable, but also explains the huge gap in graphics hardware - not just more background layers, but also effects like transparency, scaling and rotation, and richer colours.
Nintendo first demonstrated the Super Famicom hardware in November 1988: https://www.chrismcovell.com/secret/SFC_1988Q4.html . If you look at the hardware specifications listed, the only thing they changed between then and when it came out in 1990 is increasing the work RAM from 8 KB to 128 KB. I imagine they sat on the console for so long because the Famicom was still selling and they wanted to take the time to develop a good library of launch titles.
Where the PC Engine ends up weaker is in the video processor. That generation of consoles was able to do more with the video hardware than previous generations (like the NES and Master System) due to faster RAM becoming available. The PC Engine used the additional bandwidth to make it possible for the CPU to access video memory while the screen is being drawn, while the SNES kept the restriction that you can only access video memory during vblank and instead used the additional bandwidth for more background layers. Being able to access video memory all the time is definitely a useful feature, but the result is that PC Engine games often look more flat than SNES games.