Support NeoGAF

[Killer8]

The Pro version appears to have a weaker depth of field effect. You can see it most obviously in this shot:

Pro:

PS5:

Probably a consequence of PSSR scaling as resolution differences are known to change the intensity of effects like DOF and bloom.

 

[Zathalus]

When DF reviewer messed with not normally exposed game settings and essentially faked the FSR 1 review, botching the image, did they apologize or double down?



Who cares that reviewer shows clear bias up to quite noticeable criminal energy applied to shit on competitors? Hm.

As for "FSR 1 is shit" down to unusable - this demonstrates actual awareness and typical IQ level in this thread. Perhaps you should pull your heads out of DF's biased arse more often, pardon my French.

Dude you’re just salty AMD has been utterly dominated by upscaling technologies and is becoming increasingly irrelevant in the PC space, with DLSS being one of the major reasons for that. Imagine trying to stan for FSR 1 of all things. I read your thread about the reviews of FSR 1, it was remarkable in how you tried to make it more than it was and dunked on DLSS every chance you got. How is FSR 1 doing these days, does anyone ever give a shit about a spatial upscaler? DF were absolutely right on the money to be completely dismissive of it.

The other reason I was dismissive of your comment was that your other “contributions” to this thread were complaints about DF and Nvidia from years ago, and had sweet fuck all to do with this topic. Like your post about 8k gaming, DF actually did a video about 8k gaming on the 3090 and came to the conclusion that while possible with DLSS Ultra Performance, it is mostly pointless.

Call me a DF shill or greenboi (can’t believe you actually use that), but I’ll be the first in line to buy an AMD card again once they decide to actually compete at the high end.

 

[Braag]

Sony about to be new the new AI gods. If you're smart, you invest all your money into Sony stocks right now.

/s
for anyone dumb enough

 

[PaintTinJr]

Well here for example there is no extra hair I can see:

https://assetsio.gnwcdn.com/6__Pro.png?width=3840&height=3840&fit=bounds&quality=100&format=jpg&auto=webp

(PS5 Pro image from DF article)

So I presume it "pops" up only in certain key moments of the game? Like a special model which gives Abby a mane of hair only around the side of part of her cheek. Unless she shaves normally and just had a bad shaving day in that particular cutscene.

What do you mean?

Zoom in on that image and you can see hair faintly in that image at similar positioning, and also see a skin colour lobe, which may or may not be related to you thinking it was a skin colour - if the lobe is contributing tiny amounts of colour through the faint hair.

 

[Puscifer]

I'd like to better understand where/how the additional 45% GPU power is being applied because many of the games implementing PSSR have simply adopted the performance mode of base PS5 without improving graphics. In the case of TLOU, what once took them 16.67ms/frame should now take them 11.5ms/frame; taking 2ms PSSR budget into account, that would equate to a theoretical render budget of 13.5ms with 3ms+ to spare, which is a LOT for 60fps budget. Now someone may counter by saying "up to 45% GPU lift", but my response would be that both TLOU Remastered and TLOU Pt. 1 both run ~75fps range most of the time so I am actually being conservative with my numbers (By extension, we should be seeing TLOU games easily running 100fps+ on average in the modes DF has been testing).

For games like Spider-Man and Ratchet, it's even more mind boggling because they already had render budget reserved for ITGI, so net PSSR cost should be more favorable. So again I ask, where is the additional power going??

I wonder if there's some internal statistics they you can enable akin to afterburner but for PlayStation to see CPU and GPU usage.

 

[Gaiff]

I wonder if there's some internal statistics they you can enable akin to afterburner but for PlayStation to see CPU and GPU usage.

They definitely have internal profiling too for devkits.

 

[Lumpyoatmeal]

I'd like to better understand where/how the additional 45% GPU power is being applied because many of the games implementing PSSR have simply adopted the performance mode of base PS5 without improving graphics. In the case of TLOU, what once took them 16.67ms/frame should now take them 11.5ms/frame; taking 2ms PSSR budget into account, that would equate to a theoretical render budget of 13.5ms with 3ms+ to spare, which is a LOT for 60fps budget. Now someone may counter by saying "up to 45% GPU lift", but my response would be that both TLOU Remastered and TLOU Pt. 1 both run ~75fps range most of the time so I am actually being conservative with my numbers (By extension, we should be seeing TLOU games easily running 100fps+ on average in the modes DF has been testing).

For games like Spider-Man and Ratchet, it's even more mind boggling because they already had render budget reserved for ITGI, so net PSSR cost should be more favorable. So again I ask, where is the additional power going??

 

[FireFly]

What do you mean?

Zoom in on that image and you can see hair faintly in that image at similar positioning, and also see a skin colour lobe, which may or may not be related to you thinking it was a skin colour - if the lobe is contributing tiny amounts of colour through the faint hair.

I'm talking about the edge of her cheek, which in the video has an outline of what would be beard length hair that cuts off at her jaw (where she forgot to shave?).

 

[hemo memo]

Ftfy

 

[panda-zebra]

Near the ear just above a sideburn height is a natural height for sweeping tied back non-male hair, and the reason this would be visible on the Pro would be due to an increased FOV, which because AA/AAA game engines render a parallax FOV - like human vision looks in the brain - by default so the render doesn't look flat to our parallax vision looking at a 2D flat display, and instead allows us to see both sides of things slightly like our vision would allow us looking at a diorama of the game world scene.

By increasing the FOV on a Pro version the amount of both sides of an object the camera will see will be increased slightly too - as it appears in that comparison.

An interesting theory, but occam's razor. Looking back at the clip in slow motion a few times it seems more likely the haloing effect is due to simply bringing over and reinforcing temporal data from when the character moves to the viewer's right past brighter background images. There's just enough rotation on the character's head position during this time that I think it would reduce and increase this effect if it was simply out-of-view hair causing it. Once that rightwards motion ceases and moves back to the left, the halo effect instantly vanishes. That kinda confirms it for me -it's not out-of-shot hair, it's a compounded error (bug). not something that won't be ironed out as the algos/learning improve.

 

[Gaiff]

Dude you’re just salty AMD has been utterly dominated by upscaling technologies and is becoming increasingly irrelevant in the PC space, with DLSS being one of the major reasons for that. Imagine trying to stan for FSR 1 of all things. I read your thread about the reviews of FSR 1, it was remarkable in how you tried to make it more than it was and dunked on DLSS every chance you got. How is FSR 1 doing these days, does anyone ever give a shit about a spatial upscaler? DF were absolutely right on the money to be completely dismissive of it.

The other reason I was dismissive of your comment was that your other “contributions” to this thread were complaints about DF and Nvidia from years ago, and had sweet fuck all to do with this topic. Like your post about 8k gaming, DF actually did a video about 8k gaming on the 3090 and came to the conclusion that while possible with DLSS Ultra Performance, it is mostly pointless.

Call me a DF shill or greenboi (can’t believe you actually use that), but I’ll be the first in line to buy an AMD card again once they decide to actually compete at the high end.

Explains everything. I was starting to think this was a bot lol.

 

[SpokkX]

When DF reviewer messed with not normally exposed game settings and essentially faked the FSR 1 review, botching the image, did they apologize or double down?



Who cares that reviewer shows clear bias up to quite noticeable criminal energy applied to shit on competitors? Hm.

As for "FSR 1 is shit" down to unusable - this demonstrates actual awareness and typical IQ level in this thread. Perhaps you should pull your heads out of DF's biased arse more often, pardon my French.

FSR … is shit

1 was worse that whatever AA already out there and even the latest 3.1 looks absolutely terrible, especially in motion

DLSS is so far ahead it is not even funny.

PSSR seems to be better than FSR (not hard) and actally somewhat close to earlier versions of DLSS in some cases.

 

[PandaOk]

FSR … is shit
1 was worse that whatever AA already out there and even the latest 3.1 looks absolutely terrible, especially in motion

this screed is hilariously disconnected but there’s no better proof you don’t know what you’re talking about than claiming “1 was worse that whatever AA already out there”.

FSR1 didn’t have an AA pass, it relied on the in game AA. When the game had good AA, FSR1 was good. Heck FSR1 was good at *lower scaling ratios with higher base resolution* it’s far from useless so far as a basic upscaler goes.

Generally speaking FSR2 requires more work to implement than DLSS and devs just don’t bother! This is exactly why FSR2 mods can often look better than official implementations, because devs don’t bother using FSR2 exclusive features like reactivity masks.

Also on FSR1 for a moment, the performance impact is less than FSR2, so if you’re sticking with the highest quality preset on FSR1 performance gains are typically equivalent to FSR2 Balance. If you’re gaming at higher base resolutions even the first FSR can have its uses.

 

[Senua]

Wow I haven't seen defence for FSR1 in a long time

 

[Shaz12567]

FSR … is shit

1 was worse that whatever AA already out there and even the latest 3.1 looks absolutely terrible, especially in motion

DLSS is so far ahead it is not even funny.

PSSR seems to be better than FSR (not hard) and actally somewhat close to earlier versions of DLSS in some cases.

Just my 2 cents. I use a 4090 and a 7900 XTX for gaming (I am using a Neo G9 57 which requires DP 2.1 and only AMD has that to access 240hz while Nvidia is stuck at 120hz). At higher resolutions such as 4k and above, FSR 3.1 holds up pretty well unless you are pixel peeping.

 

[PaintTinJr]

An interesting theory, but occam's razor. Looking back at the clip in slow motion a few times it seems more likely the haloing effect is due to simply bringing over and reinforcing temporal data from when the character moves to the viewer's right past brighter background images. There's just enough rotation on the character's head position during this time that I think it would reduce and increase this effect if it was simply out-of-view hair causing it. Once that rightwards motion ceases and moves back to the left, the halo effect instantly vanishes. That kinda confirms it for me -it's not out-of-shot hair, it's a compounded error (bug). not something that won't be ironed out as the algos/learning improve.

It isn't a bug as it would have discontinuities, as all graphical bugs invariably do when zoomed in.

So it is quite the opposite with Occam's razor, the most likely thing when the rest of the image is flawless is that the observers - like DF - are looking for issue where there aren't any because they don't understand what is being renderer, how it is composed in parallax and how things creeping into shot but are not fully in shot will give rise to dangling side hair in that position when it is tied back and the hair moves lower and higher between relaxed and taut bun positioning - as is the supreme quality of animation in Naught Dog games since UC1.