Digital Foundry Article Technical Discussion [2023]

[pjbliverpool]

This is sufficiently bad that when I started the video my first thought was “wow they didn’t even bother to add TAA.” There are a myriad of games on the prior gen consoles with far superior TAA. Doom games, UE4 games, Naughty Dog games, the AC trilogy, Insomniac games, God of War, The Order etc.

But it's not a TAA solution. It's an upscaling solution that uses temporal information as one of it's inputs. It's not just antialiasing the image, it's upscaling it from a lower base resolution so it of course should be expected to look worse than games running at a higher resolution + TAA.

In terms of upscaling solutions FSR2 obviously can't compete with the ML solutions but it seems to be reasonably competent vs other solutions.

 

[techuse]

But it's not a TAA solution. It's an upscaling solution that uses temporal information as one of it's inputs. It's not just antialiasing the image, it's upscaling it from a lower base resolution so it of course should be expected to look worse than games running at a higher resolution + TAA.

In terms of upscaling solutions FSR2 obviously can't compete with the ML solutions but it seems to be reasonably competent vs other solutions. In Spiderman for example several outlets including DF ranked it higher than IGTI.

In the case of RDR, DF says FSR is being used as a TAA solution as the resolutions are native. I disagree that FSR 2 looks superior to IGTI, which despite apparently being worse on PC, still looks better. FSR 2 is sharper but in every other metric it looks quite bad. It's a shimmery, artifact ridden mess. The ghosting, disocclusion and smearing issues are so excessive.

 

[pjbliverpool]

In the case of RDR, DF says FSR is being used as a TAA solution as the resolutions are native.

Ah apologies, I've not watched the video so wasn't aware of that.

I disagree that FSR 2 looks superior to IGTI, which despite apparently being worse on PC,

I've heard this said a few times but I'm yet to see any direct evidence of it. Not saying it's not true but it feels like it's just been picked up as a narrative now without any substance to drive it. The only 'source' I've seen of this was from NXG's Spiderman PS5 vs PC video in which he shows clear differences in image quality without accounting for what was likely a lower underlying internal resolution on the PC.

still looks better. FSR 2 is sharper but in every other metric it looks quite bad. It's a shimmery, artifact ridden mess. The ghosting, disocclusion and smearing issues are so excessive.

I guess it's a personal preference but Alex does cover these elements in his comparison video's but still concluded FSR2 was better*

*I think - this is from memory so I stand to be corrected if wrong. EDIT: I may have been thinking of a different game / video as I can't find that comparison now.

 

[DavidGraham]

The analysis of CPU performance of Baldur's Gate 3 from DF.

Act 3 from the game is severely CPU limited due to the increased number of NPCs and AI on screen, here Vulkan performs consistently worse than DX11 by 10% or more on both AMD and NVIDIA GPUs.

 

[mpg1]

I mean I sort of get why they did this with RDR. BC of this game on Xbox sort of limits total potential sales. Switch is not overly powerful enough handle a remaster. And by just targeting the PS4 it then hits all PS consoles.

 

[Shifty Geezer]

I guess it's a personal preference but Alex does cover these elements in his comparison video's but still concluded FSR2 was better*

Better is qualitative comparison, so better in what way? If in a way one doesn't prefer, such as shimmer, then it's not 'better' in terms of 'preferred'.

Basically, the term 'better' shouldn't ever be used unless it's clearly better in pretty much every way.

 

[Silent_Buddha]

Better is qualitative comparison, so better in what way? If in a way one doesn't prefer, such as shimmer, then it's not 'better' in terms of 'preferred'.

Basically, the term 'better' shouldn't ever be used unless it's clearly better in pretty much every way.

Absolutely this. What is better to one person can be worse to another. IE - currently me and any form of temporal upscaling/reconstruction/frame injection where the added artifacts and image instability generally means that DLSS/FSR/XeSS is almost always worse when enabled than the default rendering in the game. The only real exceptions I've seen is when the game's own temporal reconstruction is even worse. But I totally understand that many people find value in those things because they don't notice those artifacts or instabilities when the game is in motion nearly to the degree that I apparently do.

Generally looks decent with no motion (although some games screw even that up with DLSS/FSR/XeSS), but then looks like ass (to me) when in motion. Which is amusing to me with something like DLSS frame generation as that's supposed to increased motion resolution but at the same time it destroys image quality (again to me) with all the newly introduced artifacts and image instability. Bleh.

Regards,
SB

 

[pjbliverpool]

Better is qualitative comparison, so better in what way? If in a way one doesn't prefer, such as shimmer, then it's not 'better' in terms of 'preferred'.

Basically, the term 'better' shouldn't ever be used unless it's clearly better in pretty much every way.

I may have been wrong about this being an Alex/Digital Foundry comparison in Spiderman. Just had a look at the video and I can't spot it in there so it may have been a different game.

In general though, when these comparisons are made they are split into various areas with each being compared separately on it's own merits. While I agree anything related to what looks better to people is always going to be personal preference, (hell some people might prefer aliasing over none!), I think if one solution demonstrates fewer visual anomalies in say 5 out of 6 tested areas, then unless that 1 area where it doesn't is obviously more impactful than the others, it wouldn't be unfair to judge one solution 'better' than the other, at least in that specific game.

FSR can sometimes beat DLSS in the odd situation but it's still fair to say DLSS gives the better quality overall.

 

[Shifty Geezer]

I think if one solution demonstrates fewer visual anomalies in say 5 out of 6 tested areas, then unless that 1 area where it doesn't is obviously more impactful than the others, it wouldn't be unfair to judge one solution 'better' than the other, at least in that specific game.

It can still only be 'better' when qualified. In this case, it'd be 'better' as an average, and the vernacular would expect it as much. But it's still unscientific.

Let's say ASS is measured as 5% better in 5/6 features over BSS, but BSS is measured as 50% better then ASS in the last feature, which is 'better'? One can pick any number of numeric analyses to pick a larger metric defining 'better'. It's still arbitrary and ultimately subjective.

I appreciate a DF conclusion is necessary for their audience, but we really shouldn't use the term 'better' without qualifying in what way(s). Except in those cases where something is better.

 

[Flappy Pannus]

fyi Oliver has had to flee from BC wildfires.

https://twitter.com/x/status/1692421369638854784

 

[Deleted member 11852]

Better is qualitative comparison, so better in what way? If in a way one doesn't prefer, such as shimmer, then it's not 'better' in terms of 'preferred'.

This is my peeve with Digital Foundry. Their preference is for visual fidelity (resolution over framerate) and, based on opinions expressed on Jedi Survivor and Starfield, they also prefer consistent a framerate. I do not, and I'm very tolerant of a variable frame rate because people have different cognitive tolerance for stimuli.

I also much prefer a balanced visual settings which I imagine publishers have data to support that being popular given many (perhaps most) console games include a fidelity, performance and balanced graphical mode. Publishers wouldn't be bothering with any of this if there was not data to support the time/effort required to include it.

Any group of individuals will thrive on diversity, people with different experience and preferences, but DF is a collection or four or five guys, who seemingly share the same preference.

Don't get me wrong, I really like their technical analysis but I find myself increasingly having to tune out their bias (preference).

 

[Cynicalking]

This is my peeve with Digital Foundry. Their preference is for visual fidelity (resolution over framerate) and, based on opinions expressed on Jedi Survivor and Starfield, they also prefer consistent a framerate. I do not, and I'm very tolerant of a variable frame rate because people have different cognitive tolerance for stimuli.

I also much prefer a balanced visual settings which I imagine publishers have data to support that being popular given many (perhaps most) console games include a fidelity, performance and balanced graphical mode. Publishers wouldn't be bothering with any of this if there was not data to support the time/effort required to include it.

Any group of individuals will thrive on diversity, people with different experience and preferences, but DF is a collection or four or five guys, who seemingly share the same preference.

Don't get me wrong, I really like their technical analysis but I find myself increasingly having to tune out their bias (preference).

Actually, they always recommend high framerate modes over high resolution ones. The only exception is when the performance modes are not consistent and fps vary a lot, in that you are correct, they go for resolution over inconsistent framerates.

But of course, one must take their conclusions as suggestions. I too, like you, prefer most times quality modes over performance. Their analysis let me know what to expect, but I still test every mode before choosing one.

 

[iroboto]

Actually, they always recommend high framerate modes over high resolution ones. The only exception is when the performance modes are not consistent and fps vary a lot, in that you are correct, they go for resolution over inconsistent framerates.

But of course, one must take their conclusions as suggestions. I too, like you, prefer most times quality modes over performance. Their analysis let me know what to expect, but I still test every mode before choosing one.

This may come down to motion resolution being a very large issue with digital monitors, such that so much detail is lost during motion that it's better to have higher framerates to produce better image quality; whereas I think with CRT monitors an overwhelming number of users may actual prefer quality modes, since motion resolution is not an issue, having higher resolution could be more beneficial.

 

[eastmen]

Actually, they always recommend high framerate modes over high resolution ones. The only exception is when the performance modes are not consistent and fps vary a lot, in that you are correct, they go for resolution over inconsistent framerates.

But of course, one must take their conclusions as suggestions. I too, like you, prefer most times quality modes over performance. Their analysis let me know what to expect, but I still test every mode before choosing one.

Yea I don't get their choices. Once I get to 60 fps I rather up resolution than try to tweak for better frame rates.

 

[SmooTh]

 

[Cyan]

what a freak is John he starts showing his 384 pages issue of a magazine. Haven't watched the video yet? I've played Quake 2 RTX an hour ago just before this video.

When looking for this game on Steam, adding the Remastered word during the search, Steam tells me that I have the game already. I also have Quake 2 on PC Gamepass 'cos I purchased it time ago.

If you aready have the game you get the remastered version for free or is it an entirely independent game?

 

[Flappy Pannus]

If you aready have the game you get the remastered version for free or is it an entirely independent game?

You get the remastered version for free. You can just select on startup from Steam if you want to launch the remastered or the Classic version.

 

[function]

Hate to be a party pooper, but I stopped playing the remaster because I didn't appreciate some of the gameplay changes.

The Berserker can now insta-jump huge distances in a way that doesn't fit in with the game, and then hit you even with level geometry in the way. It just felt cheap. The SMG also now doesn't recoil. Wat? Totally killed my retro thrills.

I appreciate this being a free upgrade and most of the changes are for the best, but I do think a toggle for the most character changing "updates" should have been included. If I could have the original Berserker and original SMG I'd still be playing.

 

[Cyan]

You get the remastered version for free. You can just select on startup from Steam if you want to launch the remastered or the Classic version.

thanks! I discovered that later in the video, I replied after starting to watch it. The only thing I miss is a RT option with XeSS support, but since I have the RTX version too -which is free as long as you have the original game- it's not the biggest deal.

 

[SmooTh]