Battlefield V with DXR on tested from Techspot(Hardware Unboxed)

[kiska3]

Source:

https://www.techspot.com/review/1749-battlefield-ray-tracing-benchmarks/

Quote

So how does DXR actually look in Battlefield V? Above is a direct comparison of the four presets, and the first thing you should notice is there’s a large visual difference between Ultra and Low, but there’s not as much of a difference between Ultra and Medium. In fact, there is basically no difference between the Ultra, High and Medium settings for DXR reflections, and it’s not just in this scene. We tested several other sections of the game and couldn’t spot any difference between the three higher settings.

This leads me to conclude that there are only two DXR modes: Ultra, which applies the full complement of ray tracing reflections in the game, and a cut down Low mode for better performance. Both the low and ultra modes apply reflections to surfaces like water, puddles and shiny objects, but it’s only the ultra mode that also applies these reflections to more matte surfaces like mud and your gun.

You’ll spot in this scene for example that the ultra mode reflects the fiery vehicle wreck off the muddy surface in the front, as well as on the barrel of your weapon at some distance away. Switch down to low, and the reflections disappear on your gun and from the mud, leaving just reflections from the water’s surface.

From what we can tell, this is the main difference between the two modes. The quality of reflections in terms of resolution, accuracy and draw distance is unaffected, so you’ll get the same sort of experience looking at those shiny surfaces whether you’re playing on Ultra or Low. Switching it down a notch only affects the amount of materials and surfaces that RTX reflections apply to. Ultra, in my opinion, gives a far better and more realistic presentation and really shows off the quality of ray tracing.

Spoiler

So cut perf about 60% or more to get more realistic reflections... I think not, at least for a fps. And this certainly doesn't reflect the price.

[suicidalfranco]

But what's the noise level like?

[voiha]

I guess people will skip the DXR option in-game totally, maybe if it had a Story mode that you can enjoy with the DXR ON, maybe yes, in fast paced multiplayer and 60% fps cut...I doubt it. Battlefield yet again looks good even on lower settings so it's a no brainer for me - only if they could present the game properly without the stupid stuff..

[GoldenLag]

oh, ye. below 50 fps is starting become kinda "meh" in terms of framerate. 

 

i mean, hopefully optimizations help in the future, but i doubt we will ever see more than 70 fps on the 2080ti. and at that point id call it pretty good. 

 

atm its good, but not quite there yet.

[asus killer]

like i said on another topic i think Nvidia's plan for this not to happen is to implement a good SLI with DXR. And that's why SLI is now a thing again.

[Nexxus]

Spoiler alert, Raytracing is really really really REALLY fucking hard to render in realtime.

The fact its doable at all is amazing.

[westport17]

Just hope the next 7nm nVidia GPUs can handle it well enough..

[Nexxus]

2 minutes ago, VegetableStu said:

people don't get this though ._. they see 60 fps evaporating and people go OMG RAYTRACING SUX

Frankly, I was expecting 30ish at 1920x1080, the fact its getting about 20 over that is blowing my fucking mind lol

[mr moose]

5 minutes ago, westport17 said:

Just hope the next 7nm nVidia GPUs can handle it well enough..

It would be nice with a new architecture and node shrink to hope for something kick arse and compelling.   It's starting to get rather droll with Nvidia did this and nvidia did that.  Having something to compare and talk about rather than simply addressing pointless hate would make for a nice change on these forums. 

[Nexxus]

2 minutes ago, huilun02 said:

But it only works with some reflections and even then not done properly

If the scene has no reflections, it is identical to having raytracing disabled

The hit to performance is unjustifiable for what little it gives

 

It is really fucking annoying during livestreams of the game and everyone constantly asking if RTX is turned on

Because if it does so little that it isn't visually identifiable and you have to ask if its on, then its pointless to care about the feature at all

You dont know how shitty it feels when you cut your framerate to a third and still nobody can tell if its enabled

Being its physics-based it is not unreasonable to assume that its still working behind the scene even if nothing is utilizing it at that moment. Since you just don't know when its gonna need to kick in for example in MP.

And I haven't watched steams but the demos frankly blew my mind. I suspect a metric ton of these people saying is RTX on is people trying to be funny and reciting memes and or frankly dont know what they are looking for.

I wish I was getting BF5 to try RTX on my own but sadly Im not buying BF games anymore, I play them a ton day one and never touch em again, so im off that treadmill now.

[daimonie]

I'm just surprised that a shooter - were people care a lot more about FPS - is the first to implement it. I would've expected NVidia to throw money at e.g. Tomb Raider and have a cinematic game showing off the feature.  1080p Ultra @60fps is what I expected to be what Raytracing could do for us now, simply because it is the "sweet spot" for many gamers.  

[Questycasos]

For early, most likely somewhat rushed software (being that this is added in a post-launch patch rather than being integrated as part of the base game) running on first generation hardware, this is actually better than I would have expected. Sure, it's not practical right now. I doubt any implementation of RTX coming out in the near future is going to have an FPS hit that most would consider practical (outside of stuff like puzzle or horror games that don't need quick reactions). But as a proof of concept it shows that there are tangible benefits to the tech that are immediately noticeable for the performance cost (lest we forget the trials and tribulations of tessellation, where we paid a huge performance tax for bumpier rocks and the occasional underground ocean).

 

I'm frankly excited for what comes out of this in 2-3 GPU generations. Node shrinks (more than likely just 1 - 1 1/2 though) combined with refinements to the RT cores to boost efficiency and throughput and more optimized software could easily bring us to a point where this becomes a standard feature (assuming AMD comes up with something comparable that doesn't end up needing an entirely separate implementation. I'd expect them to have something working by at least Navi gen 2).

[westport17]

41 minutes ago, mr moose said:

It would be nice with a new architecture and node shrink to hope for something kick arse and compelling.   It's starting to get rather droll with Nvidia did this and nvidia did that.  Having something to compare and talk about rather than simply addressing pointless hate would make for a nice change on these forums. 

This is the way technology progressing. Inventions always come first and hardware improvements will follow.

Just be positive about what future will bring us.

[Taf the Ghost]

51 minutes ago, Nexxus said:

Frankly, I was expecting 30ish at 1920x1080, the fact its getting about 20 over that is blowing my fucking mind lol

Well, it's Hybrid Rendering and it's only handling some reflections. It's amazing they've gotten that far, but this is more "Can it run Crysis?" territory for at least 2 more generations of GPUs. It's going to either be the 5nm generation of GPUs or some sort of super Vector Math Unit that's built into the GPUs.

[Taf the Ghost]

To make the performance hit acceptable, for the limited implementation, it looks like at least 100 RT cores are going to be necessary. To make it really viable, it's probably closer to 200 RT Cores.

[Shally]

Not going to lie, those scenes with RTX on ULTRA looked pretty sick so in a few years time I'm really looking forward to playing ultra realistic AAA titles

[Ashley MLP Fangirl]

all those people who bought 2080Ti's expecting at least 60fps.... 

[kingmustard]

I am hoping that I can use a combination of technologies to arrive at 1440p/60 with my i7-8700K/RTX 2080:

 

1080p, ray tracing enabled in medium or high and DLSS to upsample the resolution back to 1440p.

 

That'll do me for a first-gen. product.

[SupremeGOAT]

oh man this is gonna look so sick on my PC Classic! Everything shining and glistening... RTX rubbing baby oil on everything...you aint gaming unless you shoot a man with a glistening baby oiled gun fool...

 

EDIT: I meant shoot a person... I did not mean it to come out that way, I believe in equality. Please let me keep my job...

[mynameisjuan]

8 hours ago, Nexxus said:

Frankly, I was expecting 30ish at 1920x1080, the fact its getting about 20 over that is blowing my fucking mind lol

I remember trying raytracing in blender back in 2012ish? It boosted a render from 2-3 mins to 20 secs....20 secs for 1 frame.

[CarlBar]

Question for anyone here who knows enough about programming with the new RTX in game engines. How hard is it to implement in time trms compared to traditional ways of doing the related parts. Honestly the way it sounds, it comes across as if it uses all of the existing art asset textures, geometry, and basic lighting sources to do it all automatically bypassing the need to do more than build an on switch into the engine, at which point NVIDIA@s hardware and drivers make over and build the raytrace, de-noise it, and integrate it. But i've no idea if thats acurratte.

 

Also for anyone familiar with any of the hardware computational costs, how acurratte do you think NVIDIA's 10 Giga Rays per second number on the 2080Ti actually is in pure "has the processing power" sense. Is it pure marketing of an utterly perfect scenario or is it just a modest  overestimation of the capabilities, (the way AMD's claim about EPYC 2's 4x Floating Point Performance is, acurratte and we'll see a real benefit close to that but not a number you'll hit 100% of the time). Because i've been doing some math and 1920x1080 @ 60FPS with 3 rays per pixel and just the reflection rays on bounce only comes to around 760 million rays per second. It's making me wonder if maybe something else isn't bottlenecking he Tensor cores real hard, (it's obviously not the Rasterization, but could be PCI-E, could be Memory, Could be Denoising, hard to tell).

[Tristerin]

20 minutes ago, CarlBar said:

Question for anyone here who knows enough about programming with the new RTX in game engines. How hard is it to implement in time trms compared to traditional ways of doing the related parts. Honestly the way it sounds, it comes across as if it uses all of the existing art asset textures, geometry, and basic lighting sources to do it all automatically bypassing the need to do more than build an on switch into the engine, at which point NVIDIA@s hardware and drivers make over and build the raytrace, de-noise it, and integrate it. But i've no idea if thats acurratte.

 

Also for anyone familiar with any of the hardware computational costs, how acurratte do you think NVIDIA's 10 Giga Rays per second number on the 2080Ti actually is in pure "has the processing power" sense. Is it pure marketing of an utterly perfect scenario or is it just a modest  overestimation of the capabilities, (the way AMD's claim about EPYC 2's 4x Floating Point Performance is, acurratte and we'll see a real benefit close to that but not a number you'll hit 100% of the time). Because i've been doing some math and 1920x1080 @ 60FPS with 3 rays per pixel and just the reflection rays on bounce only comes to around 760 million rays per second. It's making me wonder if maybe something else isn't bottlenecking he Tensor cores real hard, (it's obviously not the Rasterization, but could be PCI-E, could be Memory, Could be Denoising, hard to tell).

 

Im still waiting for someone to release some information on the power consumption when the RT and Tensor Cores are turned on, because the PCBs were already hitting max wattage OC'd without those cores turned on.  Maybe creating the bottleneck as its not performing on the Turing processor like it should.

[CarlBar]

6 minutes ago, Tristerin said:

 

Im still waiting for someone to release some information on the power consumption when the RT and Tensor Cores are turned on, because the PCBs were already hitting max wattage OC'd without those cores turned on.  Maybe creating the bottleneck as its not performing on the Turing processor like it should.

 

It's lower with them on, give me a bit to dig info up.

[Tristerin]

7 minutes ago, CarlBar said:

 

It's lower with them on, give me a bit to dig info up.

If that's the case than to me that sounds like intentional gimping of that GPU so that the next generation has an easy fix, more power OR smaller node (first easy, second in the works) or a combination of both.

[PhantomHawk11]

12 hours ago, VegetableStu said:

people don't get this though ._. they see 60 fps evaporating and people go OMG RAYTRACING SUX

It doesn't really matter if it's hard to render or not, when paying that much for a GPU people expect more than ~50 frames, and no amount of special shadows is going to change that. The fact is it's an exciting new development, but it hasn't been refined enough