Digital Foundry Article Technical Discussion Archive [2016 - 2017]

Status
Not open for further replies.
Depends how they sell it when they respond I guess. If they admit its 720p but goes higher when stationary I don't think anyone will have a problem.

Try and make out it's 900p when clearly it isn't then sleazy as a sleazy thing on sleazy day.

(Assuming df are correct of course)
 
You guys need to put the pitch forks away. There's nothing sleazy happening here, they are pushing the system as much as they can, where they can and that takes more effort than just setting the resolution to 720p and calling it a day.
 
Last edited:
I demand to know what the resolution is, and how the AA solution works! Even though I can look at screens and watch plenty of video that's available, and decide from that whether I like the way the game looks, and whether I'd want to play it, I'm entitled to answers ... The devs must answer me, even on weekends!
 
4c581me2fnp5j68qwx7f.jpg
 
Their AO solution is really nice, adds lot of depth to the scene... also, good looking wood material, a shame they couldn't afford a better texture filtering solution.

Can't wait to see what the PC version looks like at 4k.
 
Remedy's official response:
Quantum Break’s 1080p output is a temporal reconstruction from four previous 720p 4xMSAA frames. This approach gets us high pixel quality in combination with complex shading and effects, allowing us to achieve a cinematic look. However, varying sample counts between passes and temporal upscaling makes talking about resolution, as it is traditionally understood, complicated in the case of Quantum Break. Since the start of “Quantum Break’s” development, the most important thing for Remedy and Microsoft has been delivering a compelling gaming experience with superior artistic quality. This is what Remedy is renowned for. We’re confident that we have achieved this, and can’t wait to hear what fans think on April 5 when they play the game.
http://www.inquisitr.com/2911665/qu...p-in-its-opening-moments/#EfaIdYuU2VHDkzcd.99

Edit: this is compared to Killzone MP which uses the 3 previous frames for the horizontal axis/pixels.
 
Last edited:
temporal reconstruction from 4 previous frames? I'd be worried about input lag more than I would about aliasing. Should be interesting to hear what they're doing.
 
temporal reconstruction from 4 previous frames? I'd be worried about input lag more than I would about aliasing. Should be interesting to hear what they're doing.
It may not work that way. When you draw say trails on missiles you keep the last X vectors of the missile and draw the trail behind a missile (connecting the previous vectors) that missile isn't behind on update code.

It's possible they are just writing the 720p buffer back to some place each frame after it displays, and recalls the last 4 in the queue to form the image.
 
Last edited:
Not surprising that it falls to pieces when you start moving then. It looks good when you stay still for a while but I can't really see the point if it breaks down to a significant degree in motion.
 
Not surprising that it falls to pieces when you start moving then. It looks good when you stay still for a while but I can't really see the point if it breaks down to a significant degree in motion.
just occurred to me, that what if the reason they went this route wasn't to save performance. Like there's so much time warping happening all over the place, if you're doing all this weird reverb/time splits and shatters, it would be to your advantage to have a frame buffer to instantly rewind to.
 
Not surprising that it falls to pieces when you start moving then. It looks good when you stay still for a while but I can't really see the point if it breaks down to a significant degree in motion.

Yes, it looks especially bad when it quick cuts to the kill animations. Jaggies everywhere all of a sudden.
 
temporal reconstruction from 4 previous frames? I'd be worried about input lag more than I would about aliasing. Should be interesting to hear what they're doing.
Why does this have to introduce input lag? It really doesn't have to, you can accumulate as you go along. Haven't watched videos extensively, but you can even hide artefacts during scene changes by employing a 4 frame fade in / out.
 
Is the reason we see so much more temporal AA/reconstruction techniques applied recently the relative high amount of memory the new consoles offer? If I understand it correctly, typically all consoles up to now were quite memory starved. Now, the consoles are quite low on CPU but have lots of memory...is this a new trend?
 
Is the reason we see so much more temporal AA/reconstruction techniques applied recently the relative high amount of memory the new consoles offer? If I understand it correctly, typically all consoles up to now were quite memory starved. Now, the consoles are quite low on CPU but have lots of memory...is this a new trend?
A more generic answer: if I understand correctly after the design for the game is fleshed out the engine people do the math to ensure the game runs according to design.

Some games like open world games have much different requirements from linear games. So if your goal is 1080p @60fps and your player can only move in specific arcs and see specific arcs you can determine from the hardware the maximum number of triangles and or other you can stream in at a given moment. IE: you can move at 1m/s forward and see and arc of 90 the
budget on triangles for a single scene is X, but you can only stream Y triangles maximum per second while moving, what is your viewing distance how many objects can we have on screen at one time etc.

The engine guys begin doing a lot of budgeting math, on available memory bandwidth as both instructions and data needs to be pulled from RAM, not to mention audio etc.

So once all that budgeting is done they look at what's left for them to reach their targets and we see things like temporal reprojection start to crop up because it fit the bill for the remaining resources. i mean we still need to take account for AF and AA etc. theres so much to account for, its never as simple or straight forward. Without knowing the design choices of the game its difficult to pinpoint why they make specific graphic choices. i feel that rewind had an important aspect in QB and since they had 4 buffers in rewind might as well take advantage of it and use it for temporal reproject.
 
Correct me if I'm wrong but regardless of the inevitable comparisons, this isn't very similar to Killzone Shadowfall right?.

Both use temporal reprojection but in a different manner from what I can observe, with QB you've got a full 720P image that is then later reconstructed using 4 frames (4x 720P is 1080P so makes sense), but is not able to maintain this in motion but only in stills.

Killzone on the other hand displays a full 1080P image all the time that is constructed using an interlaced image half of which is new data and half of which is reprojected data from the previous frame. The advantage of this would be that the game always displays a 1080P image even in motion (albiet with minor artifacting).


Both are ofcourse very different from MSAA reconstruction seen in Rainbow Six Siege/PS3 Ratchet & Clank games. So I think it'd be a bit incorrect to bunch all these techniques into one type.
 
Is the reason we see so much more temporal AA/reconstruction techniques applied recently the relative high amount of memory the new consoles offer? If I understand it correctly, typically all consoles up to now were quite memory starved. Now, the consoles are quite low on CPU but have lots of memory...is this a new trend?
I'm really digging this new trend like I mentioned elsewhere. All those previous frames contain too much valuable info to discard. Surely, the memory amount should be helping.
 
Status
Not open for further replies.
Back
Top