#467 – Tim Sweeney: Fortnite, Unreal Engine, and the Future of Gaming

Tim Sweeney: If you learn any engineering discipline, you learn massive amounts of math and you learn the rigor of problem solving, not just what you find from the Wikipedia article, but going through all of the exercises of solving complex problems and building up series of solutions to derive an answer, it's valuable and it embodies the knowledge that you need as a programmer. And people often go to the university and think, okay, my goal here is to get good grades so I get a diploma and I prove to an employer that I'm valuable. No, that's just kind of the superficial bookkeeping of the university. The real purpose of all of this is to learn. And whether you learn formally or you learn on your own, it's the learnings that are really valuable in a career.

Tim Sweeney: I think this is something that's kind of changing in America. So much focus on grades and homework and structure around kids' lives. When I was growing up, my mom would feed me and my neighbors, my neighbors and moms would feed them breakfast and they'd be like, 'Well, be back by dark.' And we'd go out and we'd play, and we'd do all sorts of things. We'd explore the woods, we'd build go-karts, we'd salvage old pieces of electronics and build what we thought were our spacecraft control panels for the fake spaceships we were building as play. And we'd have an enormous amount of freedom. The freedom there and the organic learning that occurred there, I think it's something that is really critical to the American experience. I worry is increasingly going away as society is ever more protective and sheltering and makes it harder to get these experiences.

Tim Sweeney: At first I was really depressed. I think because the wizardry of Doom especially was so incredible that I gave up on programming for like 6 months. I was like, I don't ever be able to compete with this. I have no idea what we're going to do. We just keep making 2D games and hope that the business goes on. But that was the nature of Carmack's wizardry. He had done things that were like not just one innovation leap ahead, but like a dozen simultaneously interplaying in a way that you couldn't pick them apart into their component pieces. But a funny thing happened. Michael Abrash, long-timer in computer graphics, wrote a book on the techniques for 3D graphics and texture mapping, and he wrote some articles in one of the programming magazines of the day and explained it and showed assembly code to do texture mapping, you know, drawing these 3D graphics on the screen. And it was actually really simple stuff. I was like, oh, I can do that. And so a bunch of us at Epic independently went off and started writing our own 3D graphics to figure it out.

Tim Sweeney: Abrash's original articles took like 12 CPU cycles to render a pixel. And everybody else had figured out how to get it down to six or sometimes even four cycles. And that involved lots of different trade-offs of caching and memory hierarchy and so on. It was just like a magical time where a human could actually understand exactly what the CPU was doing under the hood and could write code that exactly targeted that. And that's largely lost now. When we talk about optimization software now, it's largely about heuristics. And statistically, this memory access is likely to hit the cache. And this algorithm is faster than that algorithm. Because CPUs now have such advanced out-of-order execution that you really can't micromanage what's happening on an instruction-by-instruction basis. You can only manage the aggregate performance of code. And so there's kind of this lost art. Some people miss it, some people don't, in which the programmer had absolute control over the machine and could work miracles in special cases if you tried. The optimization problems have just moved around. And a system like Nanite, the virtualized micropolygon geometry system that Brian Karras, a brilliant engineer with Epic, built, was just one of those multi-year optimization efforts that required him understanding everything from the highest levels to the lowest levels of the hardware to figure out how to make this breakthrough technique work in a way that was actually maximally performant on GPUs.