Impact of nVidia Turing RayTracing enhanced GPUs on next-gen consoles *spawn

[OCASM]

BF:V RTX deep dive. It seems not every reflective surface is reflected or correctly reflected. Lower res reflection in the final game than the demo, very noisy water reflection, RTX off in final game looks better than the demo with RTX off.

Their conclusion is the visual gain with RTX is not worth the performance, resolution and AA sacrifice.

Some people are satisfied with console-level graphics, that's to be expected.

 

[troyan]

Their conclusion is the visual gain with RTX is not worth the performance, resolution and AA sacrifice.

And what credential has this youtube channel?I mean claiming that SSR looks better than DXR isnt something somebody would have said if he had any clue...

/edit: Ah, read it wrong. BTW i dont believe that SSR was updated from the Gamescom demo more like they dismissed the cubemaps for the train...

 

[London Geezer]

Some people are satisfied with console-level graphics, that's to be expected.

Looking at BFV, you can’t exactly praise its graphics. Some (lots of) parts are so ugly, even on a 2080ti with RTX on!

Also, when the agreement across the board is that it does look slightly better (of course), but the performance hit is just not worth it, then you really can't keep arguing about it...

 

[Shifty Geezer]

Pc gamers tend to be very much on the fps thingy, and for mp i can understand.

That's an oft repeated piece of general ignorance, but my poll here suggests very much otherwise. Unless someone has really good data on PC gamer preferences, it'd be best for us all to stop repeating this as if fact. For myself, I'm going to stop believing that until presented with some good evidence.

 

[Shifty Geezer]

... then you really can't keep arguing about it...

He can, and he will.

 

[Scott_Arm]

That's an oft repeated piece of general ignorance, but my poll here suggests very much otherwise. Unless someone has really good data on PC gamer preferences, it'd be best for us all to stop repeating this as if fact. For myself, I'm going to stop believing that until presented with some good evidence.

A poll on this forum is only representative of this forum. That said, I have no idea which way the general PC gaming population swings. My guess would be the fps crowd pretty much pushes consistent 60Hz, but I have no idea for the overall PC population which covers a lot of genres and hardware.

 

[Scott_Arm]

https://www.eurogamer.net/articles/digitalfoundry-2018-battlefield-5-rtx-ray-tracing-analysis

There are plenty of Battlefield 5 DXR performance benchmarks out there right now, and some of the numbers look low - but revised code is forthcoming that addresses a number of issues that should address the most egregious frame-rate drops. For example, all levels right now are affected by a bounding box bug making ray tracing more expensive than it should be due to the existence of destructible terrain. Certain 'fake' god ray effects or a certain type of foliage can also impact performance negatively, sending out far more rays than they should. It's difficult to get a lock on how much performance is hit by using DXR, as the computational load changes according to content - there is no flat cost here.

It's almost as if grand conclusions shouldn't be made from the first attempt with a new API, on a new game, with brand new hardware (edit: and a delayed OS patch to enable DXR in the first place).

I say give it a full two years before devs have a real grasp on RTX/DXR.

As things stand right now, the DICE developers responsible for the DXR implementation see it as a work-in-progress. Further optimisations are due, both in an imminent patch and also down the road as the title receives further support in the coming months. Even Nvidia driver updates are expected to deliver further boosts to frame-rates, such as the ability to run ray tracing compute shaders in parallel.

Drivers can't dispatch compute and RT jobs at the same time yet ... utilization anyone?

 

[Shifty Geezer]

A poll on this forum is only representative of this forum.

I know. however, it's one data-point against the accepted norm. Until someone presents at least one data-point proving Pc gamers are more interested in framerate than graphics, there's more evidence against that suggestion than for it, and it's illogical to think otherwise.

"We have very little data to go on. The little data we have suggests PC gamers are not interested in framerate. Ergo, I conclude PC gamers are very much in favour of framerate."

Does not compute.

 

[Shifty Geezer]

I say give it a full two years before devs have a real grasp on RTX/DXR.

A sensible option not available to engineers designing consoles coming out within that full two years.

 

[iroboto]

A sensible option not available to engineers designing consoles coming out within that full two years.

Neither was 4K. But the hardware engineers were able to profile and determine how to free up bottlenecks using production game code pushed to 4K.
We saw this happen with X1X, and that console produces 4K very well.
Nothing stopping MS from profiling DXR code from BFV, Metro, Shadow of Tomb Raider and other non released titles like Halo Infinite to see how it would perform.

 

[Scott_Arm]

A sensible option not available to engineers designing consoles coming out within that full two years.

How many console devs had a full grasp of async compute before PS4, X1 were released? Why do late-gen titles look much better than launch titles?

 

[Scott_Arm]

I know. however, it's one data-point against the accepted norm. Until someone presents at least one data-point proving Pc gamers are more interested in framerate than graphics, there's more evidence against that suggestion than for it, and it's illogical to think otherwise.

"We have very little data to go on. The little data we have suggests PC gamers are not interested in framerate. Ergo, I conclude PC gamers are very much in favour of framerate."

Does not compute.

You're weighing your "data point" against all of the other data points you dismiss, like forum posts on specific game forums, optimization guides for games, the vast array of pc benchmarks that are geared to high performance etc etc etc. No one knows the answer. I'm telling your what I'd expect to be true. I could be fully wrong. But I have zero confidence in your poll, and so should you.

 

[DavidGraham]

BF:V RTX deep dive. It seems not every reflective surface is reflected or correctly reflected. Lower res reflection in the final game than the demo, very noisy water reflection, RTX off in final game looks better than the demo with RTX off.

This is not a deep dive, he just collected all the bugs with RTX in BFV and posted them into a video, a lousy one at that as well.

 

[Scott_Arm]

This is not a deep dive, he just collected all the bugs with RTX in BFV and posted them into a video, a lousy one at that as well.

I found the vid interesting, because it showed all the places where their implementation fails. Always worth a look. Gamer's Nexus is probably the best PC hardware reviewer around.

Edit: Also, reading the DF interview illuminates some of the failure cases.

 

[DavidGraham]

I found the vid interesting, because it showed all the places where their implementation fails.

But it should be balanced with comparisons for RTX improvements, or at the very least be titled RTX limitations, instead of falsely calling it "deep dive".

 

[Scott_Arm]

The DF video (I'd highly recommend reading the article) shows a lot of really nice examples as to why even DXR Low is a significant improvement over DXR off. Interesting to see the subjective differences between DF's take and Gamer's Nexus's take. Ultimately what "looks better" is going to be an individual preference, but there's no doubt that DF highlights how DXR low is more accurate and adds a lot of detail that the screen-space reflections and cubemaps cannot handle. The bit at the end about the relative impact that this implementation has in BFV, compared to Quake and Crysis in the past is also pretty fair.

I'm not sure why people thought ray-tracing wouldn't come with performance penalties. It shouldn't have been shocking that there'd be compromises to resolution or frame rate if you understand what ray-tracing is. You're trading performance for quality per-pixel. Looking at this overview from DF, I'm going to be curious to see where DXR Low/Medium end up on the RTX2070 after bug fixes, planned optimizations and further optimization/tool improvements.

 

[DavidGraham]

Also, when the agreement across the board is that it does look slightly better (of course), but the performance hit is just not worth it, then you really can't keep arguing about it...

Agreement by who? half witted youtubers who have no clue and who only care about fps charts and causing drama for clicks?

Interesting to see the subjective differences between DF's take and Gamer's Nexus's take.

Gamers Nexus is not really a take, it's no more than a bug report.

I'm not sure why people thought ray-tracing wouldn't come with performance penalties.

Exactly. I mean if RTX were used in such a limited manner, people would have the right to complain. But it's use is so wide and far reaching, even characters eyes have dynamic reflections!

 

[Scott_Arm]

I think it'd be interesting if they'd let users adjust the roughness cut-off and the rt resolution independently.

 

[OCASM]

Looking at BFV, you can’t exactly praise its graphics. Some (lots of) parts are so ugly, even on a 2080ti with RTX on!

Also, when the agreement across the board is that it does look slightly better (of course), but the performance hit is just not worth it, then you really can't keep arguing about it...

"Agreement across the board of people who just want console graphics at higher framerates/resolution".

The complainers are always the loudest.

He can, and he will.

Oh you know me so well.

 

[function]

"Agreement across the board of people who just want console graphics at higher framerates/resolution".

Just what are "console graphics", given that PC and consoles share architectures, assets and shaders? And given that the 1X is faster than the top 4 GPUs in the steam survey, and faster than 15 of the top 20. Bleedin' "Haswell" graphics are at number 19!

Trying to separate graphics quality between "PC" and "console" is, especially when trying to use it as a weaponised, loaded term, pretty destructive.

And frankly, if you're running "console" graphics at higher frame rates and resolutions than any console can muster, then they're by definition not console graphics. Another reason not to get caught up in the semantics of platform wars.