Has any CS professor actually calculated how much processor and GPU would be needed for photorealism in games and how...

Has any CS professor actually calculated how much processor and GPU would be needed for photorealism in games and how many years are we left for that?

Attached: pepefroggie.jpg (569x454, 48K)

Other urls found in this thread:

youtube.com/watch?v=ag1-By-4qfw
youtube.com/watch?v=k7n5kRRHDpw&t=50s
twitter.com/SFWRedditImages

Why fuck would a CS professor do this. Is this a copy pasta Im not familiar with?
In my opinion its gonna take more time getting a display to look photo realistic

serious question, is a big topic for gaming companies to make the most realistic graphics.

Curious when we will get photoreal engines.

You could measure what it takes to render big budget movie CGI and then from that make an estimation what it would take to render it in real time

We cant even make full CG movies that look very good yet.
Or its too much work to model/texture

I don't remember who or where I heard it but I remember hearing current GPUs need ateast 60-70 TFLOPs for believable photorealism. It makes sense because if you've seen that Star Wars demo Nvidia showcased at GTC, it was run on quad-SLI Quadro GV100s which is 15 TFLOP each.

CS """professors""" are fucking hacks.
my recently tenured EE professor hinted at this.
CSlets were never and will never be as intelligent as EE.

Attached: 85262139daa4c2e46f1d2ae3b77e3169.jpg (256x256, 12K)

It's an art problem, not a power problem.
You see, hollywood still didn't managed to pull off photorealism with their insanely powerful computers, so there's no chance in hell realtime will be able to do it yet.
But if you go back to an older console and use new techniques on it (specially baking), you will be able to do much better graphical results than was available back then.
Check out this demo for example:
youtube.com/watch?v=ag1-By-4qfw

Wtf is this bull shit you faggot
Actually I don’t want to know

It is an impossible calculation to make and the answer is irrelevant.
What would you do if someone say it would take 50 years?
Make an example of when anyone would ever benefit from this knowledge.
Realism is hard to define.
The subject of what should be in "game" is hard to define.
As for "how many years left?" assumes technological advancements are linear.
It is not a resolution problem, old movies have a low resolution they look realistic.
The problem is artificially generating objects in a way that makes them look realistic.

Stupid frogposter.

Never gonna happen. You can get frames that look kind of photorealistic, but they are cherry-picked and there is no game that even looks CLOSE to photorealistic in an extended scene. Because it's fucking impossible.

>Or its too much work to model/texture
Thats why you make an estimation based on one realistic looking scene and then multiply it by how much you would need to do a full game in real time with that same quality or better.

>Because it's fucking impossible
hence the part of OPs post
>and how many years are we left for that?

reading comprehension breh

Do we even want photorealism? Esthetics are more important. What looks fun/nice is not the same as what looks real.

If we get the power for good 4k/vr at 100+ fps we are pretty much as far as we need. Let's hope 7nm CPUs and GPUs are up to the task.

We can do it if the person making it is skilled enough. I still can't tell which scene in Gladiator is supposed to be the one where they spent 6 million dollars on animating the slave master who's actor died during filming. On the other hand, Rouge One did a terrible job at remaking Leia. There is more than a decade between the movies.

Atleast a good 50 years left. And that's being optimistic. Even then, it'd be too pricey. I'd genuinely be shocked if vr ever takes off.
Just to put things in perspective, this video is from 2014 :
youtube.com/watch?v=k7n5kRRHDpw&t=50s

It's been 4 years and we haven't got much farther. And even the videos we see of vr suits and the like are unlikely to garner attention/funding/enthusiasm, so those bright ideas will die.

>medium that allows almost 100% freedom of expression
>choose to copy the physical world instead

>Do we even want photorealism? Esthetics are more important. What looks fun/nice is not the same as what looks real.
Yes

Attached: 1403837987365.png (400x389, 48K)

considering it takes like an hour for pathtracing to render a photo-realistic scene, i'd say we're a long way away

This.delete();
t. SE student

I wouldn't consider those renders to be photo-realistic, desu. Way better than what we're used to but definitely doesn't deserve to be called photo realistic.

That's too pessimistic. Remember, computer technology is continuing to evolve exponentially. I'll put it closer to 15.

never. CG render farms can do like 10 frames in an entire day (24hrs of continuous rendering).

Even if it gets developed, we won't know. Even if we know, we won't get our hands on it. Even if it's up for sale, it'd be too expensive.
Plus, progress isn't always uniform.