Digital Foundry Article Technical Discussion [2023]

Status
Not open for further replies.
been playing this version for a while, the first couple of levels, but after playing it my goal is to complete the Quake 2 RTX version. Running a Path Traced game at a rocky, stable 60 fps and a power consumption of 81W, the GPU doesn't emit a single noise, it's priceless. My eyes almost got teary from the emotion. Please someone hugs me.

NrjJet6.jpg


pycWokt.jpg


so yeah, I'm continuing to play the RTX version til completing it.

Settings to save power consumption:

- windowed mode. (this doesn't matter but still)
- temporal upscaling: yes
- dynamic resolution scaling: yes
- target framerate: 60fps
- GI: High (the max setting, the other RT options set to default)
- minimum scale: 25%
- maximum scale: 50% (hence the 81W power consumption, we are talking about a 32" 1080p TV at 1.5 meters, but the game looks good as you can see in the pictures)

Maybe we are going to see Path Traced games in mobile devices soon, which is Intel's goal, and a new Steam Deck and the likes of it are going to be a dream come true. I cried today. 😭
 
Last edited:
the only thing missing in Quake 2 (2023) superb update.


If they added something like that N64, why not full RTX compatibility? I like some features that the RTX version don't have, like the way they show you the path when you get confused in a level.
 
the only thing missing in Quake 2 (2023) superb update.


If they added something like that N64, why not full RTX compatibility? I like some features that the RTX version don't have, like the way they show you the path when you get confused in a level.
The RTX version misses shadows and AO in a lot of scenes especially outdoors. Why?


edit: AO is certainly not there. Shadows are there but strangely either too subtle or not casted if there is no strong lightsource somewhere.
 
AO is a hack when ambient lighting cannot be contained, not real light transport. There should be no AO!
AO encompasses non-screenspace methods too including RT based techniques as well! The quality of all lighting or shadowing effects are only as good as their data structure's scene representation ...
 

Some interesting discussion here! Really enjoyed the discussion about upscaling and such. Looking forward to the referenced Immortals of Aveum interview.

Minor correction @Dictator to 1:14:30ish... Unreal does by default bias both the texture and Nanite detail levels to the upscaled resolution when using TSR or similar. We do *not* currently bias the VSM resolution up because that indeed would be far too expensive (although VSM resolution on Epic shadows is already quite high, targeting more than 2x the density of screen rays).

This is all configurable of course and so no guarantees about specific games (or platforms), but that is the default setup. See these cvars if you are interested in fiddling with it:

r.Nanite.ViewMeshLODBias.Enable
"Whether LOD offset to apply for rasterized Nanite meshes for the main viewport should be based off TSR's ScreenPercentage (Enabled by default)." (note - applies to other upscaling methods as well)

r.Nanite.ViewMeshLODBias.Offset
"LOD offset to apply for rasterized Nanite meshes for the main viewport when using TSR (Default = 0)." (note - this is an offset on top of the raw upscaling ratio)

r.Nanite.ViewMeshLODBias.Min
"Minimum LOD offset for rasterizing Nanite meshes for the main viewport (Default = -2)."
 
Alex's talk starting at 1:12:39 should be disseminated throughout the gaming twittersphere. It is a message that really needs to be drilled into the gamer mindset. How to successfully redirect attention from numerical specs such as native resolution and to a smaller extent framerate that no longer carry as much significance as they have in the past? Maybe when these situations arrive for a game where on paper resolution/framerate may appear unimpressive at first glance, DF really hammers the unique or cutting edge rendering tech(s) that are being implemented into said game. Try to do a better job of explaining why the tech or what the game is doing is special.
 
@Dictator Any plans to cover Immortals of Aveum?

This game is heavy but scales poorly for reasons that are as of yet unknown.

A few hours with the RPS test PC suggests that these requirements are fairly accurate. Though the closest I could get to the minimum spec was a Core i5-11600K and an RTX 2080 Ti, some 1080p benchmark runs ended with all-Low settings producing 67fps and all-Ultra settings averaging 55fps. Weirdness of seeing a former flagship brought so low besides, these do point to the RTX 2080 Super getting around 60fps on lower settings (or Ultra with upscaling).


Hopefully there’s a DLSS 3 fix in the works as well, but Immortals of Aveum has by far the worst implementation of its frame generation feature (currently exclusive to RTX 40-series cards) that I’ve ever seen. Usually, there are some imperfections in the AI-created frames that DLSS 3 slides between the ordinary rendered ones, but in ever other game, these are so small and are on screen for such little time that they’re essentially invisible. Here, on the other hand, the UI is plagued by flashing afterimages of incorrectly generated imagery.

It’s not just ugly but unplayable, as it makes the aiming reticule bounce from side to side whenever you move or look around. And, even if frame generation was visibly fine, Immortals of Aveum doesn’t let you enable it without enabling DLSS upscaling at the same time. This robs DLSS 3 of its flexibility, as 1080p would be ideal for enjoying the frames-per-second benefit of frame generation without a noticeable drop in sharpness that upscaling would impose. Other games recognise this and let you toggle the two component features separately, but Immortals of Aveum locks the two together.
 
They will almost certainly cover it because it's a (still rare) UE5 game, as they said in the latest DF Weekly. Alex and (IIRC) Tom also interviewed some of the devs that they plan to offer as an article on Eurogamer (perhaps alongside a video?).
 
The RTX version misses shadows and AO in a lot of scenes especially outdoors. Why?


edit: AO is certainly not there. Shadows are there but strangely either too subtle or not casted if there is no strong lightsource
i don't think the PT version can "miss" the AOs cuz that just naturally comes after the technique (and obviously they have shadow rays, so not due to secondary light leakings).
My guess is that they might have some problem adapting a more physically based lighting system including nature skylighting. Somehow I remember them talking how difficult it is to balance between a natural sunlight intensity and the atmosphere of the original game.
So, here's my guess: the whatever auto exposure, tonemapping, and color grading they do reduce the visual perceiption of the brightness of the outdoor light source. It might be in fact very bright in the lighting calculation (meaning less noticeable corner AO effects), but compressed a lot in the post processing pipeline to reconstruct the original look & feel.
 
Still doesn't use ML and given how awful and riddled with artifacts FSR2 has been lately (especially on consoles) it just means we'll get 120 poor quality frames instead of 60.

Did you read the article?

So, how does FSR 3 look? At Gamescom, we had a demonstration of both titles running with the new technology active on a Radeon 7900 XTX running at 4K output. Both were running with v-sync on, which AMD recommends for frame-pacing purposes. In the very small Forspoken demo we saw, the game was running locked at 120 frames per second and looked just as a v-synced 120fps should look. The game was running in FSR 2 quality mode providing its own frame-rate boost, with frame-gen then taking you up to the limit. In terms of fluidity and clarity, FSR 3 looked a match for DLSS 3 - a view shared by Alex, Rich and John, who were all present to see the demos in person. A great start for FSR 3.

Immortals of Aveum was run in similar condititions: 4K output resolution in FSR 2 quality mode, with frame-gen then added to the equation to deliver a substantial boost to frame-rate. This time though, we didn't hit a locked 120fps, but the increase to fluidity was still profound and the quality of the generated frames again looked very good - comparable to DLSS 3.
 
Did you read the article?

Yes, I'm also well aware of how good FSR2 looked when that was first presented by AMD and we know how that turned out.

People were reporting how FSR2 looked very close to DLSS 2 then too.

These kind of demos are designed to showcase the tech at the absolute best case and very rarely represent what we get in actual games.

Nvidia are no better.
 
Last edited:
People were reporting how FSR2 looked very close to DLSS 2 them too.

Who/when? Certainly not DF.

These kind of demos are designed to showcase the tech at the absolute best case and very rarely represent what we get in actual games.

Granted marketing will always attempt to showcase products in most ideal scenarios, but Forspoken is an actual game. Immortals of Aveum is an actual game. The DF crew were able to see the demo running in person, not just relying on slide presentations. If DF and especially Alex is saying that AMD driver/software solution is comparable to HW ML based DLSS 3 visual quality, it is a very big deal even when accounting for unknown qualifiers, if any even exist in the first place.
 
Did they improve the quality of FSR2 in FSR3?

I'm doubting it's ability to remain as sharp and clear as DLSS while in motion.. which is where FSR2 really breaks down.. especially at lower internal resolutions. If they've managed to improve that, and the frame gen holds up, then massive kudos to them.
 
Granted marketing will always attempt to showcase products in most ideal scenarios, but Forspoken is an actual game. Immortals of Aveum is an actual game. The DF crew were able to see the demo running in person, not just relying on slide presentations. If DF and especially Alex is saying that AMD driver/software solution is comparable to HW ML based DLSS 3 visual quality, it is a very big deal even when accounting for unknown qualifiers, if any even exist in the first place.
There's a difference between them seeing it in a controlled setting, and them doing their own testing with their own tools.

FSR2 can look god awful and if they've not improved that base image with FSR3 (And without ML I don't see how they could) then you're still going to get shitty looking generated frames.

If they've somehow managed to improve the actual image then props to them, but Ill wait to see proper hands on testing by DF.
 
Last edited:
There's a difference between them seeing it in a controlled setting, and them doing their own testing with their own tools.

FSR2 can look god awful and if they've not improved that base image with FSR3 (And without ML I don't see how they could) then you're going to get shitty looking frames.

If they've somehow managed to improve the actual image then props to them but Ill wait to see proper hands on testing by DF.
I wonder if the ability to generate intermediate frames has given them the confidence to spend more GPU resources generating higher quality upscaled frames, allowing them to improve the quality further?

Of course they may have simply done both at the same time.. but again, I'm skeptical. Guess we'll find out soon enough. I'm particularly excited for what this tech could provide for handheld PCs like the Deck, Ally, and future iterations.
 
Status
Not open for further replies.
Back
Top