• Hey Guest. Check out your NeoGAF Wrapped 2025 results here!

Digital Foundry: Death Stranding 2 PC Review + Optimised Settings - Sony's Last Big PC Single-Player Game?

This pretty much confirms what we thought.

Last time AF was expensive was on early 200x GPUs.

Matt Damon Grandpa GIF
 
Check out that comparison above that uses APU with shared RAM pool.



Yeah, it's obvious they don't care, they don't notice far more obvious flaws in some games or lifted HDR black levels.

I remember that PS4 AF discussion very well.
Yep. This reminds me Dead Space remake with its terrible VRS artefacts notably on PS5. The image was completely destroyed, pixelated. People complained so they removed it with zero performance penalty.

Many developers just don't care at all about IQ. But I also remember some experts (including NXGamer) explaining patiently to us that was not a big problem (easy to say when you can play if fully maxed on PC without VRS) and they were pretty sure the game needed that to run at a stable framerate. And this technology was very good actually, like when we hear DLSS 5 slop is the inevitable future of gaming. Bullshit.
 
Last edited:
I was disappointed the ghosting (no pun intended) still happens in certain scenes with DLSS, unless it's a setting I missed.
 
A lot of people take not being able to set the AF how they want as a reason to purchase one machine over another.
 
Last edited:
The low AF thing with consoles has always been a puzzle to me as on PC turning to 16x loses you like 1 frame, literally.

I have just been doing it at the driver level for like 15 years or so. And its a settings that can make a huge difference in graphics qiuality, especially if its a game with a lot of flat surfaces, walls, etc.
 
I don't buy these excuses either. There's a thousand other things that are far more intensive than AF yet bring much less visual improvement.

Nobody likes chromatic aberration and similar junk. But those are apparently fine to use while AF is not.
 
Could it be that Pragmata's PC port takes advantage of the CPU and GPU being in separate pools, which enables Deck to not run into the same contention issues PS5/Xbox Series/Switch 2 do? Or maybe the 200% resolution scale is introducing a new bottleneck that hides the cost of aniso?

Bojji Bojji , try to run at 720p or lower res to hit 60fps and report on the cost of anisotropic filtering then.
 
Last edited:
And this even applies to PC ports. Many games don't have AF settings, or they are baked in "texture quality" combined with other settings.

Some games don't have any AF at all... You won't have it until you force it in the driver.
I'm going to try this, Death Stranding 2 just looks like mud to me!
 
And this even applies to PC ports. Many games don't have AF settings, or they are baked in "texture quality" combined with other settings.

Some games don't have any AF at all... You won't have it until you force it in the driver.

Call of Juarez Gunslinger, one of my favourite arcady shooters of the 360 era, is such a game.

no AF options in game. textures look like a blurry mess. forcing it through Nvidia control center massively improved the visuals
 
Last edited:
Could it be that Pragmata's PC port takes advantage of the CPU and GPU being in separate pools, which enables Deck to not run into the same contention issues PS5/Xbox Series/Switch 2 do? Or maybe the 200% resolution scale is introducing a new bottleneck that hides the cost of aniso?

Bojji Bojji , try to run at 720p or lower res to hit 60fps and report on the cost of anisotropic filtering then.

Resolution scaling is one of the reasons why I think they don't care about AF. "We upscale the game from 720p to 4k, they won't notice any AF change anyway".

I'm going to try this, Death Stranding 2 just looks like mud to me!

I think the biggest DS2 problem (and Decima in general) is how shit it handles mid distance LOD. SlimySnake SlimySnake talked about this a lot. So AF change might not help with this.

Call of Juarez Gunslinger, one of my favourite arcady shooters of the 360 era, is such a game.

no AF options in game. textures look like a blurry mess. forcing it through Nvidia control center massively improved the visuals

Yep, like I said - with forced AFx16 I don't even know which games are doing it lol. And devs are using partial AF, on some surfaces only, "optimization" (that saves them what, 0.12% of performance?) they say. Global AF works on all surfaces.

Also, let's remind ourselves how it looks like

crop-of-the-comparison-between-16x-anisotropic-filtering-v0-4mj68y555zgd1.png


Developers want you to play version on the right, they think you are stupid and won't notice.
 
Last edited:
Bandwidth limitation maybe
There are no "bandwidth limitations" associated with AF, whoever dropped that shit into public discourse was an idiot.
All AF fetches are cached on chip since about GeForce3 or so, which is why it's so cheap on PC GPUs, and it's the exact same situation on console GPUs as well, there is no difference whatsoever.

The reason why AF gets butchered on consoles is due to broken optimization priorities on the developers side.
It's easy to win some +5% in literally every scene by dropping AF down to 2X from 8X, and when you have a task to win anything at all costs you just go and change that variable - done, you're a "genius".
The reason why its "broken" is the resulting IQ hit from such an optimization which is quite often well above the performance win you get from doing this.
So an "easy win" at the cost of high impact to IQ vs more work to claw back the same 5% through doing a code optimization pass w/o IQ impact - money talks and thus the cheaper easy option wins most of the time.
Then someone will tell you that it's "because of bandwidth limitations", yep.
 
Bojji Bojji , try to run at 720p or lower res to hit 60fps and report on the cost of anisotropic filtering then.

I think you tagged the wrong guy, which is why I didn't see your request lol.

but I did redo the test at native 720p (1280x800 output, the game has black bars so it's rendering 1280x720, 100% scale, no reconstruction)

here's the result:

Trilinear
VTPSVzPIDsAQGgTt.jpg


16x AF
N3XPuTzuMycm1456.jpg


0.3ms difference. and that could be down to random variation just as I took the picture, as the frametime isn't 100% stable, it's varying between like 0.3~0.5ms just standing still
 
Last edited:
I think you tagged the wrong guy, which is why I didn't see your request lol.

but I did redo the test at native 720p (1280x800 output, the game has black bars so it's rendering 1280x720, 100% scale, no reconstruction)

here's the result:

Trilinear
VTPSVzPIDsAQGgTt.jpg


16x AF
N3XPuTzuMycm1456.jpg


0.3ms difference. and that could be down to random variation just as I took the picture, as the frametime isn't 100% stable, it's varying between like 0.3~0.5ms just standing still

And steam deck has like 88GB/s bandwidth? Half of PS4.

Let's not make excuses for devs...
 
And steam deck has like 88GB/s bandwidth? Half of PS4.

Let's not make excuses for devs...

yup, that's why I ran it at 1440p at first, really hammering that memory bandwidth.

but I have an OLED model, so technically it's 102.4GB/s.
still significantly lower bandwidth than the PS4's 176GB/s tho.
 
Top Bottom