• Ever wanted an RSS feed of all your favorite gaming news sites? Go check out our new Gaming Headlines feed! Read more about it here.
  • We have made minor adjustments to how the search bar works on ResetEra. You can read about the changes here.

dgrdsv

Member
Oct 25, 2017
11,885
I especailly don't like extreme black crush
This "extreme black crush" of yours is you seeing how light should be propagating, creating proper occlusion/contact shadowing where necessary.
But since you're used to seeing this lighting being totally and utterly fake in pretty much all titles of the last 15 or so years, you think that it's something incorrect when in fact it's the opposite.

The thing is - in this case, I don't think the RTX cards are a good value right now. If I WERE reviewing things and talking about value for your money I'd come to the same conclusion as many other reviewers. That's not what I'm about, though. I also don't think we need THAT much more rasterized performance RIGHT NOW. I could already play everything at native 4K at 60fps with a 1080ti class card - no problem. Sometimes it required some tweaks but I feel like performance was great. By the time we NEED more performance, new cards will arrive. I just don't see the burning need for dramatically more classic performance...and if you don't, then don't buy it!
One thing worth remembering is that most people who are angry with 20 series cards price/performance had never even considered buying anything like 1080Ti you've mentioned and are using something like 1050Ti/1060. For them 20 series is useless since there is no - and there won't be - 20x0 card sold on such pricing. They are using cards with performance which isn't really enough to handle 4K workloads - or even 1440p ones, unless you're willing to go with some really low quality settings. And they can't upgrade to 20 series because even the slowest 2060 costs some 50-75% more than 1060 (3/6 GBs). This is where the most rage and disappointment in Turing is coming from I feel. The rest is just the usual rationalization.

Now we have these seemingly arbitrary restrictions on DLSS being tied to different resolutions on different cards, and no explanation as to why.
DLSS 1x will only improve performance if a game is shading a frame in DLSS resolution slower than tensor cores are rebuilding the resolution with DLSS. After some point of engine's performance, DLSS will become more of a limiting factor than shading which may lead to performance _loss_ with DLSS enabled compared to just rendering in native res without DLSS. I'm not saying that this is the reason for a strange limitation on DLSS we've got in BFV but it's a possibility.
 

pulsemyne

Member
Oct 30, 2017
2,641
Can someone else try something? I just played BFV with DLSS on but using high textures instead of ultra. Now I'm not sure if it was some placebo effect or other but it looked better. Seemed to be less blurry. It might have been the maps I played but it did seem better. MAybe the DLSS implementation is odd with ultra textures.
 

Serious Sam

Banned
Oct 27, 2017
4,354
This "extreme black crush" of yours is you seeing how light should be propagating, creating proper occlusion/contact shadowing where necessary.
But since you're used to seeing this lighting being totally and utterly fake in pretty much all titles of the last 15 or so years, you think that it's something incorrect when in fact it's the opposite.
Not true. In real life you almost never see pitch black shadows if there is natural day light coming through the windows because real life "rays" are bouncing infinite times and create some sort of illumination in shadows, even if it's minimal. And even if it's night time with dim lighting inside the room you still don't have pitch black areas, there is always at least SOME light even in the darkest shadows. RTX GI in Metro has some cool effects in some places, but it isn't real-life accurate, what I mean by that is that it is trying to be accurate, but technology of today isn't sufficient to do that in real time.

I5TjLxA.jpg
 
Last edited:
Nov 2, 2017
2,275
Not true. In real life you almost never see pitch black shadows if there is natural day light coming through the windows because real life "rays" are bouncing infinite times and create some sort of illumination in shadows, even it's it's minimal. And even if it's night time with dim lighting inside the room you still don't have pitch black areas, there is always at least SOME light even in the darkest shadows. RTX GI in Metro has some cool effects in some places, but it isn't real-life accurate, what I mean by that is that it is trying to be accurate, but technology of today isn't sufficient to do that in real time.
True for daylight but not true for night time unless I have really bad eyesight at night or you have cat eyes. All I need to do is look around and I see plenty of pitch black areas in my room and there's plenty of light coming from my monitor.

The longer I look towards the darkness the less pitch black spots remain but it takes a while for your eyes to properly adjust.
 

dgrdsv

Member
Oct 25, 2017
11,885
Not true. In real life you almost never see pitch black shadows if there is natural day light coming through the windows because real life "rays" are bouncing infinite times and create some sort of illumination in shadows, even if it's minimal.
You're looking at game graphics through a monitor, not at a real life. This is a limitation of a display device more than anything else at this point. Also of note: there are no light bounces in Exodus in scenes where there is no sun light as only sun light generate DXR rays but there is DXR occlusion shadowing in such scenes. There's also just one light bounce from the sun as well - for obvious performance reasons.
 

signal

Member
Oct 28, 2017
40,199
I watched this entire deep dive into RTX 28 minute video. They've found and showed plenty disappointing and undercooked RTX implementations. Have you watched it? I especailly don't like extreme black crush that is so abundant with RTX on, as if RTX implementation was rushed and this wasn't taken into account in development.


I liked the discussion in this video about sometimes preferring RTX off because it looks more like a game / "what we're used to games looking like" even if it's less realistic or whatever. Interesting to wonder then how much of the comparisons are "better vs. worse" rather than "different than my what I'm used to / my expectations and thus worse vs. better"
 

Vash63

Member
Oct 28, 2017
1,681
Static time of day probably not a good use case for single bounce realtime GI

This. To elaborate, DICE can already bake fake GI at a very high level of quality in BFV because the lighting doesn't change very much. The only thing to account for is updates when walls are destroyed and such and those can just be a single time pass during the destruction, not constant updates like a moving sun or flashlights.

The difference for real GI in BFV would be less dramatic.
 

rashbeep

Member
Oct 27, 2017
9,464
This. To elaborate, DICE can already bake fake GI at a very high level of quality in BFV because the lighting doesn't change very much. The only thing to account for is updates when walls are destroyed and such and those can just be a single time pass during the destruction, not constant updates like a moving sun or flashlights.

The difference for real GI in BFV would be less dramatic.

AO would be nice to see. can't stand the way it looks in any FB game
 

Tokyo_Funk

Banned
Dec 10, 2018
10,053
So the latest patch for BFV and NVIDIA drivers has given me a boost on my 2080Ti using DLSS. 1440p with RTX and Ray tracing has increased from 63-68fps to 80-85fps average. DLSS seems to leave a softer image though overall, with finer details in the distance seeming blurred a bit. Nice improvement overall despite the blur.
 

plagiarize

It's not a loop. It's a spiral.
Moderator
Oct 25, 2017
27,559
Cape Cod, MA
Not true. In real life you almost never see pitch black shadows if there is natural day light coming through the windows because real life "rays" are bouncing infinite times and create some sort of illumination in shadows, even if it's minimal. And even if it's night time with dim lighting inside the room you still don't have pitch black areas, there is always at least SOME light even in the darkest shadows. RTX GI in Metro has some cool effects in some places, but it isn't real-life accurate, what I mean by that is that it is trying to be accurate, but technology of today isn't sufficient to do that in real time.

I5TjLxA.jpg
We only get there via single bounce implementations first.

I'd still argue it's more accurate. If the above image was Metro with it's other GI implementation, the wall the window is in, would be too light, the stuff on the shelves would be lit all wrong. The reflected light on the table would be the wrong colour. Etc Etc.

Again, if the devs say the ray traced GI is closer to what they want the game to look like, knowing that it's also closer to how light really works, that's more than good enough for me.

Of course, if we just call it an artistic choice, people are welcome to disagree with it...

But it's certainly more 'correct' for how 4A Games want Metro to look.
 

Jijidasu

Member
Oct 25, 2017
75
Japan
What's the word on chip reliability right now? I'm going to be building a new PC within the next couple of months and curious as to whether the faulty/at-risk chipsets have been ridden out of circulation. I'm in Japan so I'm not sure if things are slower or not here, but I'll be buying through Amazon (so I'll have some warranty for returns in any case, but wish to avoid any complications).
 

icecold1983

Banned
Nov 3, 2017
4,243
dynamic objects would have it though and contribute to it - it could make for interesting destruciton looks or when levels are filled with moving objects.

I just question whether the performance hit would be worth it. It would also have to be in addition to the system they are using now as otherwise too many areas would likely see a reduction in quality. Would probably be quite a headache for Dice. AO however would be good as frostbite AO has been weak for so long now. Im surprised they never just used nvidias HBAO+. Im sure theres a good reason however given dice proficiency with rendering
 

Xx 720

Member
Nov 3, 2017
3,920
Something I've noticed with dlss in battlefield - if I play on my monitor the loss of detail is noticeable but on my tv, it's way way less easy to tell a difference. If a game has good ray tracing in it like Metro I might play on the tv @ 4k
 

icecold1983

Banned
Nov 3, 2017
4,243
Something I've noticed with dlss in battlefield - if I play on my monitor the loss of detail is noticeable but on my tv, it's way way less easy to tell a difference. If a game has good ray tracing in it like Metro I might play on the tv @ 4k

Are you sitting at the same distance? Whats the resolution/screen size of each
 

Xx 720

Member
Nov 3, 2017
3,920
No, and that is probably why my tv is oled 4k I'm sitting about 9 feet away on the monitor I'm right next to it so easier to see. On the tv I dropped bfv to 1440p and then 4k dlss I liked better with the 4k
 

snail_maze

Member
Oct 27, 2017
974
I liked the discussion in this video about sometimes preferring RTX off because it looks more like a game / "what we're used to games looking like" even if it's less realistic or whatever. Interesting to wonder then how much of the comparisons are "better vs. worse" rather than "different than my what I'm used to / my expectations and thus worse vs. better"
Thats sort of where I am. I don't have an RTX card so I may be somehow biased to justifying the status quo, but from what I've seen of raytracing I don't really feel like it's always better for a game. I don't at all deny that it adds realism but I sort of prefer it when shadows aren't as dark or lighting differences as pronounced. Many of the "RTX on/off" comparisons are lost on me. Sometimes I prefer one other times the other and when I only look at one of the screens then I quickly forget what the other looked like.
 

low-G

Member
Oct 25, 2017
8,144
Very impressed with the performance in Metro Exodus on my 2080. All that preview build stuff was running at 30fps-ish. My GPU utilization is hitting highs of 65% on ultra settings, hairworks, physx, and RTX on. I could either try for extreme settings & RTX Ultra, or crank it up to 1440p seemingly.

And here I was fretting about not having done the step-up to a 2080Ti.


I must really give kudos to Nvidia, Dice, 4A, and anyone else that helped because they've brought the early preview stuff an extremely long way.
 

Xx 720

Member
Nov 3, 2017
3,920
Very impressed with the performance in Metro Exodus on my 2080. All that preview build stuff was running at 30fps-ish. My GPU utilization is hitting highs of 65% on ultra settings, hairworks, physx, and RTX on. I could either try for extreme settings & RTX Ultra, or crank it up to 1440p seemingly.

And here I was fretting about not having done the step-up to a 2080Ti.


I must really give kudos to Nvidia, Dice, 4A, and anyone else that helped because they've brought the early preview stuff an extremely long way.
So, so glad to hear this. I have a 2080, definetly want to play with rtx on, was worried after battlefield that it would be too taxing.
 

dgrdsv

Member
Oct 25, 2017
11,885
GI is a much better use of RT than reflections in BFV
I'd argue that it's not actually. Reflections are a lot more visible while GI is both difficult to implement in addition to a regular fake solution (can't avoid this as its needed for consoles and non-RTX PC cards) and provide tangible benefits only with lots of dynamic lighting which isn't as widespread as one might think, especially since it's hard to fake properly without RT.

do you guys suspect the 3080ti will get announced this fall?
No. Don't expect 30 series this year, it'll likely come out next spring. Also don't expect 3080Ti launching straight away either, it'll likely be more like with Pascal, with a new Titan launch first and a top Ti card coming a year later or so.

I'd more expect a pricedrop of the 20xx series when Navi hits.
Wouldn't count on this either as it looks like the only 20 series card in danger of directly competing with Navi is 2060. Maybe NV will cut the prices on high end Turings in 2H19 just to increase sales though.
 

laxu

Member
Nov 26, 2017
2,782
I liked the discussion in this video about sometimes preferring RTX off because it looks more like a game / "what we're used to games looking like" even if it's less realistic or whatever. Interesting to wonder then how much of the comparisons are "better vs. worse" rather than "different than my what I'm used to / my expectations and thus worse vs. better"

I wonder if they have used the RTX on setting to fine tune the lighting in RTX off. It would make sense as raytracing would give you more accurate lighting so adjusting the faked lighting it to make it look better, while reducing the difference between the two, would make the game look better for those without RTX hardware or for those on GPUs too slow for it.
 

Galava

▲ Legend ▲
Member
Oct 27, 2017
5,080
RTX can also be used for sound (afaik). Making objects cast sound they make as rays that bounce off the level "automatically", thus achieving sound that "evolves" as it travels through the level, bouncing and going through other geometry. Don't know how expensive performance-wise that would be though.
 

laxu

Member
Nov 26, 2017
2,782
I'd argue that it's not actually. Reflections are a lot more visible while GI is both difficult to implement in addition to a regular fake solution (can't avoid this as its needed for consoles and non-RTX PC cards) and provide tangible benefits only with lots of dynamic lighting which isn't as widespread as one might think, especially since it's hard to fake properly without RT.

On the other hand, the amount of significantly reflective surfaces depending on the game can be pretty low too. These two uses of RT are at the moment probably going to be less visible than Hairworks considering you are more likely to have a character with hair on screen at all times than mirrors, partially reflective surfaces or globally illuminated scenes.

It's kind of like you can't sell Boardwalk Empire as a CGI blockbuster because its CGI is things you won't even realize are not real like background compositions. People are hoping that raytracing makes this huge difference when we have more than a decade spent on faking realistic lighting in real-time graphics.
 

Smokey

Member
Oct 25, 2017
4,176
Very impressed with the performance in Metro Exodus on my 2080. All that preview build stuff was running at 30fps-ish. My GPU utilization is hitting highs of 65% on ultra settings, hairworks, physx, and RTX on. I could either try for extreme settings & RTX Ultra, or crank it up to 1440p seemingly.

And here I was fretting about not having done the step-up to a 2080Ti.


I must really give kudos to Nvidia, Dice, 4A, and anyone else that helped because they've brought the early preview stuff an extremely long way.

Hmm? Your GPU should be close to 100% usage...or did you can it at 30fps?
 

TheModestGun

Banned
Dec 5, 2017
3,781
s it was hard to imagine this being very good. Before that
Please help. Recently purchased a 4K tv. Have a pc that was considered pretty good a few years ago?at 1080p (gtx 980, i7 4970k, 16 gigs). Latest games just don't do even 1440p at anything above 30 fps. Playing at 60 fps it's very hard to compromise better graphics for lower frame rate, but I want the frame rate AND the resolution at least at 1440p/60 without compromising much.
So, I'm going to upgrade very soon. Looking at a new gpu, either ASUS rtx 2070 or 2080. 16 gigs of ddr4 ram and an i7 9700k. Would that be a suitable upgrade? Also the motherboard ASUS rog strix z390-f
How will that do? I'm more concerned about what happens with next gen consoles and not jumping the gun too early on any parts. Please help as I'm no expert in this area.
it might be worth waiting on the GPU end, hard to say. But a 2080 will definitely hit 60 FPS at 4K in many games with very little compromise and is 100+ at 1440p. 2070 I personally feel is probably going to feel outdated the fastest.
 

dgrdsv

Member
Oct 25, 2017
11,885
On the other hand, the amount of significantly reflective surfaces depending on the game can be pretty low too.
Most games have SSRs these days and RT reflections basically are always better than SSRs. Although with low amount of reflective surfaces RT performance hit can be hard to justify, that's true.
 

Deleted member 12317

Account closed at user request
Banned
Oct 27, 2017
2,134
RTX can also be used for sound (afaik). Making objects cast sound they make as rays that bounce off the level "automatically", thus achieving sound that "evolves" as it travels through the level, bouncing and going through other geometry. Don't know how expensive performance-wise that would be though.
Some games already use a similar tech like Rainbow Six Siege and several racing games, sounds bounce on walls, floors and ceilings.
 

dgrdsv

Member
Oct 25, 2017
11,885
TU116-vs-TU106-GPU-1000x500.jpg


TU116 is roughly the same size as GP104 which might mean that tensor cores are still there.
 

Flappy Pannus

Member
Feb 14, 2019
2,342
One thing I'd like to see when sites like Digital Foundry review DLSS implementations in games is to take the approach Hardware Unboxed did in its DLSS investigation, which is not simply compare image quality/performance to either the 'native' DLSS rendering res (such as 1440p in DLSS4K) or the final framebuffer res, but choose an intermediate non-DLSS resolution that more closely mirrors the performance profile you get with DLSS 4K.

DLSS incurs a performance hit above whatever res it's starting from, meaning that DLSS 4K which natively renders at 1440p doesn't give you 1440p performance - you get something in between that and 4K. So it's certainly not 'free', that's not exactly surprising. However, many reviews still seem to make the mistake of giving it too much credit by comparing it to native 1440p res with TAA in terms of image quality (I mean it should be much better but apparently with Metro Exodus it may even be blurrier than that currently!), or conversely being somewhat unfair by comparing it with native 4K image quality - albeit I say 'somewhat' as per Nvidia's marketing we should expect something very close to that.

Ideally though, set either a custom resolution or use the resolution slider in certain games until you're within the performance profile of the DLSS mode used, and then compare image quality and performance. That's what HB unboxed did by setting Final Fantasy and the Nvidia demos to 1800p, and why they came away far less impressed than others. When setting a custom res is so trivial to do, I don't really care than DLSS 4K looks better than 1440p, or that it performs better than native 4k - I want to see how it looks/performs compared to the native resolution which can equal its performance.

If it can't offer a noticeable upgrade over 1800p or some other arbitrary intermediate resolution using just regular upscaling to 4K, especially considering how long patches are taking to come out and other artifacts it can bring, it deserves some harsh criticism IMO.
 

Anarkin

Member
Oct 27, 2017
121
I hope someone here can help me out, since I can't create my own thread and don't know where else I can ask.

I bought a RTX 2080 card and was doing some benchmarks of course and everything looked good and on par with benchmarks online, except for Doom.

Game runs only between 60-100 fps not matter what. The strange thing is, that no matter the settings, it will run the same (low vs. ultra). I also tried both vulkan and opengl, no change.

My system:
Intel i7 5820 @4,5Ghz
Asus RTX 2080
16GB ram
Windows 10
G-Sync 144 Hz monitor

I'd like to play the game again but want this issue resolved first.

Here is a screenshot with detailed ingame info:

 

Skeleton

Banned
Oct 26, 2017
1,240
if anyone is getting into anthem tonight on 2080 please let me know average frames etc.

Interested to see what you get and also full machine specs would be super helpful!
 

low-G

Member
Oct 25, 2017
8,144
So, so glad to hear this. I have a 2080, definetly want to play with rtx on, was worried after battlefield that it would be too taxing.

What problems did you have with BFV? I can also run that >1080p 60fps absolutely maxed...

Hmm? Your GPU should be close to 100% usage...or did you can it at 30fps?

Hmm? Why is that? No I'm at 60. Matches what GamersNexus shows...


Even a 2060 can do 1080p 60 on ultra + RTX high...
 
Last edited:

Lump

One Winged Slayer
Member
Oct 25, 2017
16,034
if anyone is getting into anthem tonight on 2080 please let me know average frames etc.

Interested to see what you get and also full machine specs would be super helpful!

70fps average at all ultra settings at 3440x1440 on a 2080ti (the ultrawide FoV really does drop the framerate)
 

Bjones

Member
Oct 30, 2017
5,622
So weird going from a Vega 64 to a 2080.

Not just performance but there is a lot difference in the way the image is displayed and the way things are rendered.

The ocean effects in AC oddessy is a good example of this. The breaking waves on the shore are much more fluid / natural looking on the 2080 compared to more rigid synthetic look on the Vega 64.
Like there's a randomization in the shader that just works better on the Nvidia hardware.

HDR is another area in the 2080 I've had to dial back my picture settings because everything was getting blown out. It seems to have a higher contrast curve than my Vega 64 did.
 

plagiarize

It's not a loop. It's a spiral.
Moderator
Oct 25, 2017
27,559
Cape Cod, MA
One thing I'd like to see when sites like Digital Foundry review DLSS implementations in games is to take the approach Hardware Unboxed did in its DLSS investigation, which is not simply compare image quality/performance to either the 'native' DLSS rendering res (such as 1440p in DLSS4K) or the final framebuffer res, but choose an intermediate non-DLSS resolution that more closely mirrors the performance profile you get with DLSS 4K.

DLSS incurs a performance hit above whatever res it's starting from, meaning that DLSS 4K which natively renders at 1440p doesn't give you 1440p performance - you get something in between that and 4K. So it's certainly not 'free', that's not exactly surprising. However, many reviews still seem to make the mistake of giving it too much credit by comparing it to native 1440p res with TAA in terms of image quality (I mean it should be much better but apparently with Metro Exodus it may even be blurrier than that currently!), or conversely being somewhat unfair by comparing it with native 4K image quality - albeit I say 'somewhat' as per Nvidia's marketing we should expect something very close to that.

Ideally though, set either a custom resolution or use the resolution slider in certain games until you're within the performance profile of the DLSS mode used, and then compare image quality and performance. That's what HB unboxed did by setting Final Fantasy and the Nvidia demos to 1800p, and why they came away far less impressed than others. When setting a custom res is so trivial to do, I don't really care than DLSS 4K looks better than 1440p, or that it performs better than native 4k - I want to see how it looks/performs compared to the native resolution which can equal its performance.

If it can't offer a noticeable upgrade over 1800p or some other arbitrary intermediate resolution using just regular upscaling to 4K, especially considering how long patches are taking to come out and other artifacts it can bring, it deserves some harsh criticism IMO.

This is a great suggestion. Yeah, we should really be comparing like for like *performance*. Not necessarily exclusively, but since DLSS 4K runs at a lower frame rate than 1440p native, we should compare IQ between 1800p or whatever like you say.

One other thing I haven't seen yet, that we need to see however, is we should *also* be comparing IQ while the character is moving or at the very least while on screen elements are moving. Only showing TLAA examples on relatively static scenes really hides a lot of the issues TLAA has, and isn't that representative of in gameplay. I know that it's more difficult to get similar shots if stuffs moving, but it's arguably only fair to compare BOTH static and in motion things when comparing the IQ.

Like in FF15, with its poor TLAA implementation, in motion stuff looks terrible, and DLSS looks way better imo. I've seen precisely zero screenshots of what Metro Exodus's TLAA implementation looks like on stuff in motion. I can't imagine it's anything like as bad as FF15, but like I said, you can't adequately judge TLAA on a static scene, unless it's a game where you spend most of the time looking at static scenes.
 

DiceHands

Member
Oct 27, 2017
4,638
Buying it in-store with no online pre-pickup still gets you a game code?

I was gonna go pick one up in-store and price match Amazon for like $20 cheaper, but still wanted a free game.

I too am planning on upgrading from a 970. And yeah, it still performs fine, but I notice myself making more compromises on visual quality for performance, no to mention I've had the thing over 4 years now.

Yes, the game code came on the receipt.

The card they had available was the EVGA 2060. Do you see a price match somewhere? Id love to knock another 20 bucks off my credit card bill, lol.

I still havent been able to use it though because my PSU did not have the correct cables. I ordered a modular PSU from newegg and it should be here tomorrow, so Ill finally get to try this card that is just sitting here.

I, too, have been knocking down settings more than I would like with my 970. Apex legends really wasnt getting me the FPS I wanted and it was really the first time I noticed the age of my card. It didnt run bad, but it was not getting me 60 fps consistently in that game.

I probably could have splurged on a 2080, but it seems like the 2060 is plenty for 1080p 144hz. Hell, it is apparently even great at 1440p and can do some games in 4k.