Is 60fps the new thing for next-gen?

The problem is if you don't introduce latency you have to project into the future, that means you will always be wrong in the case of any discontinuity of motion.
Which will introduce artifacts.
As to using it for reflections, if it takes 2ms, how does that compare to actually rendering them?
2ms is a long time when you only have 16 for the entire frame.

It's an interesting idea, the value would be very much dependent on the artifacts, they don't have to be very severe before it's a net loss.
 
Could interpolated-framerate be used partially, like for obviously predictable movements (scenery), and so free up some time every other frame for other computations?
 
60fps is nice, but 30fps is more than acceptable for most games, especially if it's a smooth 30fps without dips. I'd be more interested in making sure they all run at 1080p.
 
The problem is if you don't introduce latency you have to project into the future, that means you will always be wrong in the case of any discontinuity of motion.
Which will introduce artifacts.
As to using it for reflections, if it takes 2ms, how does that compare to actually rendering them?
2ms is a long time when you only have 16 for the entire frame.

It's an interesting idea, the value would be very much dependent on the artifacts, they don't have to be very severe before it's a net loss.

Since most motion will be natural, it should work most of the time. However their results seem fine, things do not seem to overshoot. There's also a 2 frame version with which adds a frame of latency, without this problem, which could be used for cutscenes for example.

For the reflections, it could be used for cases where your game runs at 60fps, but you don't have that much budget for reflections and it would normally run at 30fps (like the forza example), anything you can render at 30ms can be made to look like running 60fps, so you have about 30ms and not 2ms. But you do have to have velocity buffers. Also, it's quicker for reduced resolutions, which would be the case for reflections. 720p takes 1,5 ms on that PS3 actually. Also, for reflections you may not need as complicated solution, you might not need per-object velocity data, just information on camera velocity could do the job.


Could interpolated-framerate be used partially, like for obviously predictable movements (scenery), and so free up some time every other frame for other computations?

This is exactly mentioned in the future work section! Separate scenery from objects, and render the scenery at 20 or even 15fps. It surely could add a lot of image quality, with minimal artefacts, you obviously need to interpolate shadows and reflections separately, if I understood the method correctly. You'll see camera pans work perfectly in the 15fps -> 60fps part of the video, while there's visible artefacts on the objects (which is not noticable at 30fps -> 60fps).

60fps is nice, but 30fps is more than acceptable for most games, especially if it's a smooth 30fps without dips. I'd be more interested in making sure they all run at 1080p.

This technique could allow you to play those steady 30fps games at 60fps, which is quite a nice bonus, IMHO.. It could be optional, tho.
 
60fps is nice, but 30fps is more than acceptable for most games, especially if it's a smooth 30fps without dips. I'd be more interested in making sure they all run at 1080p.

Visually sometimes I agree, but there's controller to screen reaction latency to worry about as well, which may be the more important reason for the success of CoD's multiplayer.
 
Yeah, it depends on the game. The faster the response times required, the more you're going to want the higher framerate. CoD is probably the best example on console because of the 60fps thing, but that's been the standard for PC shooters for years.. I'm not sure I could even play a shooter at 30fps any more.
 
Visually sometimes I agree, but there's controller to screen reaction latency to worry about as well, which may be the more important reason for the success of CoD's multiplayer.

I do not think the controller latency matters that much, a bigger reason is that you actually see motions clearly. And I am not sure if this technique will help or not.
 
I do not think the controller latency matters that much, a bigger reason is that you actually see motions clearly. And I am not sure if this technique will help or not.
What "technique"? We're talking about native rendered framerates now, I think we've stopped talking about frame interpolation. If you really want to talk about it, it wouldn't have any effect on response times or reaction.. the player would still effectively be playing the game at 30fps.

Personally, I think frame interpolation should be avoided at all costs.. way too many visible artifacts, no matter which method you use. And for anyone that knows how to spot artifacts like that, it would drive us absolutely batty seeing it all the time.
 
What "technique"? We're talking about native rendered framerates now, I think we've stopped talking about frame interpolation. If you really want to talk about it, it wouldn't have any effect on response times or reaction.. the player would still effectively be playing the game at 30fps.

Did I miss a memo or something? :)
 
Could a focus on 60FPs be an attempt at keeping production costs under control?
 
Last edited by a moderator:
Many of us who are in our 30s can remember a time when the *vast* majority of high-end arcade games were a rock-solid, butter-smooth and consistant 60fps. Starting with Daytona and Ridge Racer.

I'm overall quite pleased with what I am seeing for this upcoming gen. A larger percentage of games are going for 60fps.
 
Many of us who are in our 30s can remember a time when the *vast* majority of high-end arcade games were a rock-solid, butter-smooth and consistant 60fps. Starting with Daytona and Ridge Racer.

I'm overall quite pleased with what I am seeing for this upcoming gen. A larger percentage of games are going for 60fps.

I don't expect that to last though, as soon as they'll need the power to make prettier pixels they'll ditch 60Hz.
It's easier to sell HQ images than gameplay, quite unfortunately... :(
 
I don't expect that to last though, as soon as they'll need the power to make prettier pixels they'll ditch 60Hz.
It's easier to sell HQ images than gameplay, quite unfortunately... :(

I;m not so sure, it may be more like how AA went this gen where the attempt was made initially but later replaced with more innovative ideas like mlaa. The same could happen with framerate this gen, be it with adaptive resolution or interpolated object framerate or something else.
 
Quite frankly, if a studio like Naughty Dog chooses to use the extra horse power to make prettier games instead of smoother ones, they certainly have my blessing.
Heck, save for traditional twitch-based arcade games, they all have.
 
There's nothing forcing developers to do 1080p60. In fact, I wouldn't be surprised if we eventually see 720p30. My sincere hope is that we don't dip below 720p though, I think that's one step too far.
 
Could a focus on 60FPs be an attempt at keeping production costs under control?

Unlikely. In fact, the amount of content expected by customers would be the same, but everything had to be more optimized to fit into smaller poly counts, smaller texture spaces, less draw calls, use simpler shaders and so on. That means extra artist time when crafting the ingame assets.
 
the simple truth is that photo realism is not needed to sell games. If a developer has a 60 frames per second game most likely they would not want to spend the time and resources to lower the game 2 30 frames per second and work to push photo realism. the fact is that people don't really care if games look realistic anymore. Developers know this. That is why at e3 only one to get demos actually approached near photo realism. Nothing else that was rendered in real time came close. This is proof that having 60 frames per second games even if they are not the best they could be at 30 frames per seconddo not need to be great visually to sell. People will buy less than the best.
 
ERP said:
As to using it for reflections, if it takes 2ms, how does that compare to actually rendering them?
2ms to get 2 16ms frames from 33ms, and in screenspace - so it scales with resolution of your render-targets. Cost footprint per-frame is comparable to screenspace motion-blur methods, so you can up-trade for "free".

Personally I think the real value of this isn't necessarily doubling the framerate as maintaining a constant framerate - as opposed to majority of late-generation games (TLoU is a pretty big offender there - to name the most recent). Velocity-based motion blur does nothing to mitigate the juddery impact of variable 20-30fps games, artifacts permitting this could be more pleasing to the eye as well (most shader-MBs are hardly artifact free to begin with).
Particularly could be valuable for stuff that doesn't target an actual hw-platform (eg. PC games).
 
Many of us who are in our 30s can remember a time when the *vast* majority of high-end arcade games were a rock-solid, butter-smooth and consistant 60fps. Starting with Daytona and Ridge Racer.

I'm overall quite pleased with what I am seeing for this upcoming gen. A larger percentage of games are going for 60fps.

Being somewhat younger I still remember that, it is quite a beauty for the eye such a smooth imagery, one of the reasons I still like Nintendo games visually is how well it runs, knowing that I will hardly ever notice a frame drop!

It maybe is just me but I think it even easy the need of some kind of fxs like motion blur and the like.
 
Last edited by a moderator:
Back
Top