You might be interested in this GDC talk about the PS4 Spiderman game, which dives into the performance optimizations they used, which are in a similar spirit (though at a higher level of abstraction): https://www.youtube.com/watch?v=KDhKyIZd3O8

Eh, kinda. There are still lots of performance tricks and clever hacks needed for games, just not on the storage capacity front. For example see this breakdown of the GTA5 rendering pipeline, it's still very complicated and employs lots of clever tricks.

https://www.adriancourreges.com/blog/2015/11/02/gta-v-graphi...

A lost art to be rediscovered. This type of insane hyper-optimization has a lot of applications today. Even in the exact domain of video codecs on modern systems.

Perhaps we cannot get much better perceived quality per bit, but I would strongly argue that we could improve other areas such as encoder/decoder latency.

I've got a little toy project where I am generating 1080p 24bpp frames in a C#/AspNetCore webservice and sending them to client browsers at 60fps over websocket. Nothing serious, I was just trying to test the limits of what I could do with current-year managed language and some crappy fun code. Think of it as a really bad attempt at Google Stadia. I currently see about 5 frames of latency with MPEG-1 using bidirectional pipes against FFMPEG. While this is an extremely encouraging first result, I feel there is some compromise I can make on quality+bandwidth in order to further reduce this frame latency (I.e. if I were to hand-roll some codec for this specific application).

If I could get the frame latency down to the ~30ms range on localhost, this opens the door really wide on some crazy project ideas I've had regarding novel real-time graphics approaches. I will definitely be printing this article off for deeper review.

Fitting things onto a 32MB cartridge isn’t a constraint anymore. Current consoles use 50GB discs and next-gen ones are going to use 100GB BD-XL discs. Makes it less necessary to do stuff like that.