Support NeoGAF

[dgrdsv]

I don't understand.

If some frames are missing pixels, doesn't that mean the frame isn't 4K?

Missing pixels (2x2 quads really) are still there even if they are just black or transparent. The image dimensions are still 3840x2160, it just have about half of its pixels filled with solid color (or zeroes which is the same really). The next frame rendered is what's filling the missing parts of the first image or, more accurately, it's the first image which is filling the missing parts of the next one. So it's temporal reconstruction, sure, but it's not really upscaling.

 

[Nostremitus]

I don't understand.

If some frames are missing pixels, doesn't that mean the frame isn't 4K?

They are arguing a technicality. It is a method of rendering a lower res image and converting it to a higher resolution image.

They are saying that because the PS4 Pro increases the resolution by spacing each pixel out and inserting blank pixels between them that are filled in with approximate gradients based on the actual rendered pixels of another smaller frame I the reconstruction process it is somehow native 4k instead of a lower res image converted to 4k...

The PS4 pro is still only actually rendering a smaller image. It's just stretching it up to 4k and then filling in the empty pixel created by doing so.


Like I originally said, the PS4 Pro uses a different method to perform the same task that QB's "upscaling" option does.

 

[dogen]

inserting blank pixels between them that are filled in with approximate gradients based on the actual rendered pixels

No, they use the fully rendered pixels from the last frame and reproject(using motion and depth information) to fill in these gaps. The idea is to avoid interpolating as much as possible.

In most cases the end result will be much better than if it worked as you described.

 

[Nostremitus]

No, they use the fully rendered pixels from the last frame and reproject(using motion and depth information) to fill in these gaps. The idea is to avoid interpolating as much as possible.

I'd expect the end result in most cases to be far better than if it worked as you described.

It's still converting a lower resolution image to a higher resolution image. It's still not native 4k. And it's still a process that serves the same purpose that QB's upscaling does.

 

[dogen]

It's still converting a lower resolution image to a higher resolution image. It's still not native 4k. And it's still a process that serves the same purpose that QB's upscaling does.

I'm not arguing with that, just explaining some of the details.

 

[Nostremitus]

I'm not arguing with that, just explaining some of the details.

Gotcha thanks, this whole conversation is because I said those things and apparently it bothered some people.

Thanks for the clarification though. An article I read following the reveal mentioned gradients. I can't find it now, though. It must have been redacted...

 

[low-G]

How's this run on a 1070? I'm about to get mine in a few days.

My other specs are:

i7 6700hq
32gb RAM

I'm playing the crappy DX12 version and I get 60fps near max details except upscaling, so in DX11 it'll be even better.

Forget about playing in 1080p though unless you want 30fps.

 

[Comrade Crunch]

I'm playing the crappy DX12 version and I get 60fps near max details except upscaling, so in DX11 it'll be even better.

Forget about playing in 1080p though unless you want 30fps.

Alright, thanks man.

 

[SimplexPL]

If you disable upscaling and set Xbox One level of graphics (all on medium except textures), you may be able to get 1080p60 with some occasional dips below 60.

 

[Lonely1]

With a 1070 at XBOne settings? You are lowballing the performance hard on my experience.

I finished it on a 980ti with mostly maxed settings @1440p with upscalling on, which means it was rendering very close to native 1080p, and the game mostly stayed at well above 60 fps, I would say that around 70fps was the average.

A 980ti is slightly slower than a 1070, so you can expect similar performance even though my 980ti was OC to ~1.4Ghz. My other specs are: 64GB of RAM and a 5930k @4.5Ghz, Windows running from a PCiE SSD and the game from a SATAIII SSD.

 

[Beer Monkey]

I'm getting 60 at 1080p with some drops to 59 and 58. Upscaled, medium everything but ultra textures and geometry. 3770k plus GTX 1060.

May mess with tweaking further but it plays and looks nice to me. Not getting any major hitches at transitions.

 

[SimplexPL]

With a 1070 at XBOne settings? You are lowballing the performance hard on my experience.

Xbone graphical settings, but at twice the framerate and over twice the pixel count (so over 4 times more of pixels per second). I don't think that's lowballing it too much. The game is badly optimized after all.

 

[Tomodachi]

Does this happen all the time or is it a bug in some section?

All the time but only in menus and when visualizing documents and such. It's annoying but not a game breaker (the pointer disappears in actual gameplay and every cutscenes).

 

[Daingurse]

All the time but only in menus and when visualizing documents and such. It's annoying but not a game breaker (the pointer disappears in actual gameplay and every cutscenes).

I don't seem to encounter this on the UWP version either. Maybe cause it's always running in a Borderless Window? It's pretty annoying on the Steam version when using Exclusive Full Screen.

 

[ussjtrunks]

I don't seem to encounter this on the UWP version either. Maybe cause it's always running in a Borderless Window? It's pretty annoying on the Steam version when using Exclusive Full Screen.

Does running the steam version in Borderless Window fix the bug?

 

[Daingurse]

Does running the steam version in Borderless Window fix the bug?

I think it did when I was testing it, but I had much worse performance than I did in Exclusive Full Screen mode. There is something weird about the v-sync on that option, it does not behave correctly. Should be Triple Buffered using borderless windowed mode, but I swear it was acting like Double Buffered V-Sync. . .

 

[Filben]

How's this run on a 1070?

Pretty good,1080p (without upscaling) and almost constant 60fps if you keep volumetric lights on high instead of very high and the screenspace reflection on medium (you won't notice much difference anyway as long as you keep it on). You may have to turn the shadow resolution from ultra to high as well.

If you aim for 30fps you can keep shadows on highest level and increase the resolution to 1440p or even a bit higher when you reduce some settings.

 

[sfried]

Does anybody have more AMD figures? I'm having second thoughts now wheather I should get the Steam version or the Windows Store version due to DX12 performance benefits on AMD cards. That, and I don't know if I'm willing to pay to upgrade my card to the GTX 1070 OC ITX... (as I probably don't even need it atm)

 

[dgrdsv]

Does anybody have more AMD figures? I'm having second thoughts now wheather I should get the Steam version or the Windows Store version due to DX12 performance benefits on AMD cards. That, and I don't know if I'm willing to pay to upgrade my card to the GTX 1070 OC ITX... (as I probably don't even need it atm)

https://www.computerbase.de/2016-09/quantum-break-steam-benchmark/
http://www.pcgameshardware.de/Quantum-Break-Spiel-15745/Specials/Technik-Test-Steam-1209192/
http://gamegpu.com/action-/-fps-/-tps/quantum-break-test-gpu-steam-versiya

Generally, the only reason to go with UWP/DX12 version is when you have a slow CPU (basically an AMD one) and an AMD GPU, otherwise Steam/DX11 would be a better choice.

 

[SimplexPL]

That EG blurb is wrong, 2x2 into 4x4 is not checkerboarding and it doesn't make any sense pixel numbers wise considering specs.

So this quote is also wrong?
"The principle behind the 'checkerboard' scaling is pretty straightforward - a 2x2 pixel structure is extrapolated up to 4x4, apparently using new hardware built into PS4 Pro's GPU, so there is no cost to developers"
Source: http://www.eurogamer.net/articles/digitalfoundry-2016-4k-gaming-on-ps4-pro-tech-analysis

 

[dgrdsv]

So this quote is also wrong?
"The principle behind the 'checkerboard' scaling is pretty straightforward - a 2x2 pixel structure is extrapolated up to 4x4, apparently using new hardware built into PS4 Pro's GPU, so there is no cost to developers"
Source: http://www.eurogamer.net/articles/digitalfoundry-2016-4k-gaming-on-ps4-pro-tech-analysis

Well, I don't think that we have any solid info on how PS4Pro really work but 2x2 into 4x4 is not "checkerboarding" and it makes little sense to upscale 1080p into 4K on a machine which is twice faster than the original PS4. With PS4Pro being twice in flops and checkerboarding meaning a checker board rendering pattern (which gives exactly twice more pixels in a 4K buffer compared to 1080p) it makes more sense to actually render twice more pixels than in 1080p and reconstruct into 4K from there. But this can't be handled in the way this quote is describing the tech since 4x4 is 4 times more pixels than 2x2.

 

[sfried]

https://www.computerbase.de/2016-09/quantum-break-steam-benchmark/
http://www.pcgameshardware.de/Quantum-Break-Spiel-15745/Specials/Technik-Test-Steam-1209192/
http://gamegpu.com/action-/-fps-/-tps/quantum-break-test-gpu-steam-versiya

Generally, the only reason to go with UWP/DX12 version is when you have a slow CPU (basically an AMD one) and an AMD GPU, otherwise Steam/DX11 would be a better choice.

My German is quite limited, but I can't make out the comparisons between AMD GPUs in DX11 and DX12 on those sites...

 

[JudgmentJay]

I started playing this today. Is it just me or do the screen space reflections cause a lot of weird graphical anomalies? Like I can see the effect being cut off on the sides of the screen sometimes, and I can see splotches of light on surfaces that really shouldn't have any reflections. It looks weird and is pretty distracting, but when it's working it looks great. I'd hate to have to turn it off completely.

The AO causes some abnormal haloing around stuff too, but that's less noticeable.

 

[Daingurse]

I started playing this today. Is it just me or do the screen space reflections cause a lot of weird graphical anomalies? Like I can see the effect being cut off on the sides of the screen sometimes, and I can see splotches of light on surfaces that really shouldn't have any reflections. It looks weird and is pretty distracting, but when it's working it looks great. I'd have to have to turn it off completely.

The AO causes some abnormal haloing around stuff too, but that's less noticeable.

I don't know what exactly causes them, but yeah I see tons of graphical anomalies in this game, on both the UWP and Steam versions. Lots of weird flickering and light reflecting off surfaces oddly, like the shit you described. I just can't articulate it. Glad I'm not alone lol.

 

[JudgmentJay]

I don't know what exactly causes them, but yeah I see tons of graphical anomalies in this game, on both the UWP and Steam versions. Lots of weird flickering and light reflecting off surfaces oddly, like the shit you described. I just can't articulate it. Glad I'm not alone lol.

It's definitely the SSR because when disabled, the issues disappear.

 

[Tomodachi]

Does running the steam version in Borderless Window fix the bug?

I think it did when I was testing it, but I had much worse performance than I did in Exclusive Full Screen mode. There is something weird about the v-sync on that option, it does not behave correctly. Should be Triple Buffered using borderless windowed mode, but I swear it was acting like Double Buffered V-Sync. . .

I get the mouse pointer bug both in fullscreen and borderless window
I also get double buffering in fullscreen and triple buffering in borderless window, so I'm using the latter and having no issues on that front.

 

[elelunicy]

Does triple buffering work for anyone at all in exclusive fullscreen mode? I thought it worked fine for me just few days ago, and now it's always double buffered no matter what do

 

[dgrdsv]

My German is quite limited, but I can't make out the comparisons between AMD GPUs in DX11 and DX12 on those sites...

CB.de is the only site which made a direct comparison but you can look up old UWP benchmarks for other two although it won't be really correct. Anyway, as I've already said, UWP/DX12 version makes some sense only when you're using an AMD CPU+GPU system, otherwise Steam/DX11 is the better one.

 

[sfried]

CB.de is the only site which made a direct comparison but you can look up old UWP benchmarks for other two although it won't be really correct. Anyway, as I've already said, UWP/DX12 version makes some sense only when you're using an AMD CPU+GPU system, otherwise Steam/DX11 is the better one.

Isn't there still performance degradation using an AMD GPU with DX11?

 

[Arkanius]

CB.de is the only site which made a direct comparison but you can look up old UWP benchmarks for other two although it won't be really correct. Anyway, as I've already said, UWP/DX12 version makes some sense only when you're using an AMD CPU+GPU system, otherwise Steam/DX11 is the better one.

Gears of War 4 DX12 is solid for everyone. Both Nvidia and AMD. I guess they did the Render paths correctly.

 

[dgrdsv]

Isn't there still performance degradation using an AMD GPU with DX11?

Some, which can be fixed with drivers and even if they won't be - these couple fpses aren't worth it when they come with UWP and WinStore issues.

 

[Deleted member 325805]

Are the real life episodes streamed on the Steam version as well? If so are they smooth/good quality? I have a fibre connection so it shouldn't be a problem on my end.

 

[Bluth54]

Are the real life episodes streamed on the Steam version as well? If so are they smooth/good quality? I have a fibre connection so it shouldn't be a problem on my end.

Yes they are streamed on Steam and the one I watched looked fine (though I was struggling to get through the one I saw, it was pretty boring and I wanted to just keep playing the game instead of watching a 20 minute TV show).

 

[electroflame]

Are the real life episodes streamed on the Steam version as well? If so are they smooth/good quality? I have a fibre connection so it shouldn't be a problem on my end.

I've only seen the first one, but the picture quality of the show was fine, I thought. Overall quality was...lacking. Even as a huge Remedy fan, I was basically fast-forwarding through it to get the gist of the story elements.

 

[BBboy20]

Is the Steam version that much of a CPU hog? RAM is at 70 while my i5-2500k is dancing around 100 even at the lowest settings.

 

[ussjtrunks]

Really enjoying the game but some odd technical issues drag it down

* Yeah why is the mouse cursor just in the middle of the screen when using a controller
* Cutscene transitions from 30 > 60 and from 60 > 30 are incredibly jarring, it takes far too long to make the transition and it looks awful
* I encountered an part in Act 3 so far where my autosaves were not working so I had to run back 3 times to complete the section
* Upscaling artifacts are noticeable on occasion especially when turning the camera

 

[Dmax3901]

Should I give up trying to get 60fps on my r9 290?

EDIT: what settings appear to be the most expensive?

 

[Vash63]

I finished the game already, but did anyone else notice the gamma gets messed up in Exclusive Fullscreen? I had to play it in Borderless Windowed to avoid really bad black crush.

 

[d00d3n]

Did they ever patch dx12 performance to steam dx11 levels, or is it still a ridiculous 30% performance hit?

 

[Locuza]

It's still the same and probably always will be.

 

[TSM]

Did they ever patch dx12 performance to steam dx11 levels, or is it still a ridiculous 30% performance hit?

Nvidia's drivers are responsible for that performance delta. Unless DX11 is coming to the windows store version, it is what it is.

 

[d00d3n]

It's still the same and probably always will be.

Nvidia's drivers are responsible for that performance delta. Unless DX11 is coming to the windows store version, it is what it is.

That is too bad ... It really takes the fun out of the experience to know that you are playing a gimped version of the game. I hope that I can resist rewarding these awful practices from Remedy/Microsoft when the Steam version goes on sale.

 

[Locuza]

It runs slightly better on AMD and far worse on Nvidia.
Having the different APIs DX11 and DX12 store exclusive is just awfully stupid.

 

[hendersonhank]

Pretty good,1080p (without upscaling) and almost constant 60fps if you keep volumetric lights on high instead of very high and the screenspace reflection on medium (you won't notice much difference anyway as long as you keep it on). You may have to turn the shadow resolution from ultra to high as well.
.

I'm getting very stuttery framerates with those settings on an MSI laptop, 6700HQ, 1070, 16gb of RAM, running off SSD. Right from the very start of the game.

Anything else I should check? Framerate limiter is off, I'm using the limiter in RTSS to cap it at 60, using Nvidia Inspector for vsync, exclusive fullscreen in QB options.

Even though I'm not getting 60fps, neither GPU nor CPU are hitting 100% and are usually in the 80s.

Steam version BTW.

 

[Filben]

Anything else I should check?

It seems like a big trade-off but try turning off v-sync, at least to just check the performance. V-sync seems bugged in this game because sometimes it works well and sometimes my fps fluctuate between 35 and 60fps by just panning the camera around the character. Without v-sync I have no problems at all.

 

[JaseC]

It seems like a big trade-off but try turning off v-sync, at least to just check the performance. V-sync seems bugged in this game because sometimes it works well and sometimes my fps fluctuate between 35 and 60fps by just panning the camera around the character. Without v-sync I have no problems at all.

That's because it's regular, double-buffer vsync. At 60Hz it caps your framerate to 60fps (60/1) and drops it to 30fps (60/2) if 60fps cannot be maintained, then to 20fps (60/3) if 30fps cannot be maintained, and so on.

 

[SRTtoZ]

Finally bought the game and I'm really impressed. I've been rocking a 2gb 680 and figured that would shit the bed with this game so when I finally got my new Evga 1070 SC I thought now was a good time to buy. I'm not the biggest fan of the upscale feature but it really doesn't bother me much, and everything else is cranked to Ultra settings, and I'm seeing a smooth 60fps at 1080p(upscaled obv). The game itself seems to be pretty fun, I really like the acting and the time manipulation story.

 

[hendersonhank]

I'm using triple buffered vsync through Nvidia Inspector. Without it I get tearing.

 

[dgrdsv]

I'm using triple buffered vsync through Nvidia Inspector. Without it I get tearing.

There is no way of forcing triple buffering in a DX game with Inspector.

 

[hendersonhank]

There is no way of forcing triple buffering in a DX game with Inspector.

Really? I thought that was entirely what it was for. Under what circumstances would it work, then?

At any rate, with vsync in-game off and vsync (w/triple buffering) in Inspector, I'm getting no tearing and framerates generally in the 50s with dips into 40s, just walking around the opening scene on campus, using the aforementioned graphics settings.

Vsync off is not an option as then I get tearing.