Movies like The Matrix and video games like The Evil Within have popularized the idea of a simulated reality. I'm talking about a human connected to a computer and able to experience similar things like a person would be able to experience on Earth. The simulation doesn't necessarily have to simulate atoms and fundamental particles, rather like we see on video games structures of landscape and "reality" inside it (which of course would be limited to the size of computing power).

Also, I'm not talking about the entire Universe rather the interior of a city. What server-side physical infrastructure would be needed to achieve this?

  • $\begingroup$ Hi michael! I think we'll need a bit more information to help you answer this question, as right now it's very broad. Are you asking about the server-side physical infrastructure required to run such a simulation? The algorithms required to render the highest resolution a human eye can distinguish? Or the Kardashev-level society required to make such an undertaking feasible? I've voted to place your question on hold so you don't get swamped with unhelpful answers. You may also find the Sandbox useful! $\endgroup$ – Dubukay Sep 9 '19 at 5:12
  • $\begingroup$ Thanks for help! $\endgroup$ – michael smith Sep 9 '19 at 5:20
  • 1
    $\begingroup$ I made a few additional edits to help with grammar - feel free to modify anything more. Your edits have been super helpful figuring out what you're looking for, but you've introduced two separate questions here that would each be best as a stand-alone (physical infrastructure vs brain-body connection). I'd recommend leaving off the "breakthroughs in neuroscience" and stick to the physical infrastructure required to simulate a city $\endgroup$ – Dubukay Sep 9 '19 at 5:34
  • $\begingroup$ So you want me to delete part about neuroscience? $\endgroup$ – michael smith Sep 9 '19 at 5:50
  • 1
    $\begingroup$ The physical infrastructure size is a bit on the side of overly broad. The part about neuroscience definitely makes this question very broad. I would recommend limiting yourself to one question about one topic instead of trying to cover everything. Also what sort of city are we talking about? How many people live and move around? $\endgroup$ – Shadowzee Sep 9 '19 at 5:52

Simulating the physics is a tricky concept. If you were born in the matrix, how would you know any physics was "wrong?" Only those who escaped the matrix would know.

However, we can do some herustics on scene generation. Our movies are the closest to "photo realistic" we're going to get, so we can use that as a benchmark. Transformers: Dark of the Moon provided an interview in which they said the most complex scenes they rendered took 288 CPU hours per frame. Multiplying that by a crude 30 frames per second, and you need just shy of 900 CPUs to keep up. With, say, 10,000 residents, that's 9,000,000 CPUs, dedicated to nothing but rendering.

What does that look like? We don't make them that big. To give you a sense, here's Mira Mira

Mira is a Blue Gene/Q hosted at Argonne National Lab. She's got 65,000 processors.

So imagine Mira, cloned about 150 times, side by side. That's your scene generator.

Now, that being said, the problem is actually worse. Those speeds assume rendering is perfectly parallelizable. It isn't. So you'll run into issues there. You could also save some time by only rendering what's in the fovea of each person, but now you have to be able to respond to changes in the eye within the blanking period of a saccade. That's a hard real time challenge. But at least it gives you a sense of scale.

  • $\begingroup$ I'm confused by your maths here. If you need 288 CPU hours per frame, then you'll need 8,640 CPU hours per 1 second (at 30fps). But that's 8,640 CPUs running for an hour to generate 1 second of rendering. So this would mean you'd need over 31 million CPUs for real time rendering per person. $\endgroup$ – Kyyshak Sep 10 '19 at 15:22
  • $\begingroup$ @Kyyshak You're right. I dropped the hours-to-seconds conversion $\endgroup$ – Cort Ammon Sep 10 '19 at 18:07
  • $\begingroup$ @Chickensarenotcows I just wanted to provide a back-of-the-envelope lower bound estimate. Even just doing the oversimplifications of our input that I did, the result is still an astounding amount of infrastructure. $\endgroup$ – Cort Ammon Sep 10 '19 at 18:07

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.