For all the coding to the metal discussion. I am developer but not a game dev.(Though I have wrote some raytracing and wrote my own 3d engine 19 years ago that ran quake 3 levels). However one of the most innovative, smartest, a true game dev expert amongst the top .1 percent and gifted 3d engine dev ever, John Carmack said this about consoles vs PC's :
What Carmack is saying is fundamentally true, but I don't think people understand the context in terms of modern-day consoles and consumer electronics. It's not like in the past where the architectures themselves had intrinsic differences in how they did various functions and calculations, algorithms etc. where coding to the metal was necessary to get the most out of the systems. And consoles in the older days, having MUCH simpler OS (or in some cases none at all) meant there was no OS overhead or OS acting as an abstraction layer for the games to run on, which freed up all system resources for the game itself during runtime.
Nowadays, virtually all consumer products are running on either x86 (or x86-64) or ARM processors, They all use either DDR-based or GDDR-based RAM. They all either use that ram in a unified or split memory pool. They all feature support for almost all the same connectivity standards such as USB, PCIe, etc. They all feature support for almost all the same industry-wide interfaces such as SATA, UHS, NVMe, etc.
Featuring such similar architectures and feature support, modern electronics these days generally see their performance differences based on inclusion of things such as amount of RAM, memory bus bandwidth sizes, generation of processors used, what GPU they use, amount of storage, specific connectivity standards they support, features of the OS they run etc. But none of that involves 'coding to the metal'. Even in regards to PS4 and XBO, developers cannot code to the metal on those systems because if they happen to exploit very unusual/niche architecture features that successor consoles don't have via hardware, then that game will not be backwards-compatible unless significant re-coding is done in the game. Depending on the severity the game used those architectural features as exploits originally, the entire game may have to be re-coded from ground zero to run on that newer system.
Sony and MS want to ensure backward-compatibility as standard, so while they may have APIs that cut down on the abstraction levels some, they are not letting any developer (even their in-house ones) code "to the metal" the way we saw devs doing with even PS3, let alone systems like PS2, N64, Saturn, SNES etc. Those days are over for mass-market consoles; they've BEEN over for several years, in fact.
We can tell that, as well, simply looking at the relative visual gains in, say, PS4 exclusives going from the start of the gen to where we are now. Yes, there have been gradual improvements on-average, and most PS4 exclusives are visually arresting like few others, but...the gap between, say, Killzone:ShadowFall and Death Stranding is noticeably smaller than Uncharter to Last of Us was for the PS3. And it has to do with more than just diminishing returns: coding "to the metal" for AAA games of these size is just unrealistic in this day and age.
You may get very specific or simpler functions and features coded at a lower level that's as analogous to pure assembly coding as possible, but the majority of the game is not going to be coded that way. It's simply infeasible from a time and budget perspective. The visual gains we've seen this gen have mainly been thanks to re-use of textures and assets created at the start of the gen, freeing up dev time from needing to generate completely new ones from scratch (not saying there are NO new assets being created completely from scratch, but it's much less than it was at the start of the generation, hence why cross-gen was being pushed so hard early on). The visual increases are also coming mainly thanks to increases in budget; that means more artists, more animators, more programmers etc able to either create more assets and/or polish created assets even more within a development time frame in-line with what's already there.
I feel Carmack 100% understands this context, but it's not something you can literally put in a tweet while keeping it simple. Unfortunately I feel a lot of others are still under the impression that differences in big console AAA games and games on other platforms is down to "secret sauce" and "coding to the metal" when those days ended in 2013. You want to know
why there are few PC games with visuals comparable to God of War IV, GT Sport etc.? Because there isn't a large enough install base of high-performance GPU cards on PCs to justify the budgets and manpower necessary to create those level and quantity of assets. That's the only major difference between console and PC gaming at the AAA level these days, besides perhaps consoles having unified memory (keep in mind there are still some advantages to DDR/GDDR split pools too, and when memory amount is high for both it mostly cuts off the unified memory advantage consoles have).
