• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

The thread which sometimes talks about RDNA2

Status
Not open for further replies.
ljt said:
The amount of games is irrelevant, there are only going to be more and more as time goes on.

Pretty much the majority of the games with RT also have DLSS, because without it they would be unplayable.

The point I'm making is without some form of DLSS type of technology on either vendor, ray tracing becomes practically useless, especially at higher resolutions.

If I were buying now and wanted to catch up with some games like control or metro for example, and wanted to try ray tracing then it makes it difficult to choose a 6800xt because frame rates at 4k would make it unplayable, because there is no ai upscaling available to help with the heavy lifting ray tracing requires.

Then we have cyberpunk releasing soon, which again will probably mean without DLSS it would be unplayable with RT enabled, meaning it will only probably be playable with RT on nvidia until AMD sort their 'super resolution' out
Click to expand...
If DLSS is not supported in more than few games, then the majority will just disable RT if the performance hit is too big.
 
LtMatt said:
If DLSS is not supported in more than few games, then the majority will just disable RT if the performance hit is too big.
Click to expand...

But DLSS is pretty much in every RT title and it will be that way going forward too.
 
EastCoastHandle said:

Performance per watt 6800 wins







In Valhalla, a newer title, SAM shows massive performance uplift.
Click to expand...

For that benchmark yes. For doom eternal @ 4k sam on 172fps and sam off 187fps. Metro Exodus sam on or off @ 4k 60 fps @ 1440p both 94fp. Assassin's Creed Unity the same no change in performance. Borderlands 3 66fps both on and off @ 4k. Same with Control and Control RTX.


source https://www.eurogamer.net/articles/digitalfoundry-2020-amd-radeon-rx-6800-and-6800-xt-review?page=6

There are more SAM benchmarks here https://www.techpowerup.com/review/amd-radeon-sam-smart-access-memory-performance/2.html

For my resolution 4k most games dont see the upto 11% AMD stated. Most get a few 1-2fps at best. Sometimes SAM hurts performance.
 
When Nvidia brought out DLSS 1.0, it was quickly surpassed by AMD's RIS. Nvidia had to improve with DLSS 2.0. I have no doubt that AMD will respond sooner rather than later with their own technology that will be as good or better.
 
Anonagrog said:
Isn't that just sophistry? Whilst I agree on financials, that has nothing to do with the differences between a "tacked on" versus a "next gen/proper" implementation.
.
Click to expand...
You mentioned financials. All those questions are financial question which drive the amount of money that a company is willing to spend on the implementation/development. A company isn't going to spend more than it thinks it will get in return hence you will get tacked on implementations. And for quite a while we will continue to get tacked on implementations till the mass market (under £300) has the power necessary to run the "next gen/proper" implementations.


Anonagrog said:
Are our definitions the same here, I wonder? I'm looking at it from a technical/development perspective only. If what you consider as "next gen/proper" is an implementation that only requires h/w ray-tracing (ie. just as we all need gpus with gpgpu compute), then even that should only be an issue when that particular game doesn't have the usual fall-back rasterization/compute method for what their solving with ray-tracing (unlikely, save for an in-house engine maybe?), or simple toggle. Otherwise, if it's not technical but rather in terms of the narrative/gameplay (eg. story event relies on reflection of something beyond screen-space, dynamic diffuse indirect illumination controlling 'feel' of full environment, or maybe even non-graphics related use case) itself forcing ray-tracing as a baseline, then sure (I don't think any of those examples are possible considering perf of the new consoles, so it likely won't translate to pc - maybe I'm not creative enough in my thinking though).
Click to expand...

We are indeed on different pages. But in hindsight my bar is unfair, since one of the main advantages is that RT moves the time spent to achieve "realism" from the artist to the prcoessing unit.
I'm looking at it from the perspective of the full package for raytracing at a minimum lighting and shadows but in a way that provides a tangible improvement as an end user (There is the quality vs cost argument but that is for a different conversation). However with a well made game any differences will be small and will heavily affect the quality vs cost argument (against the RT implementation) hence i am being unfair. Regarding reflections, we simply don't have the horsepower to do reflections with roughness at the moment and I am not impressed by polished reflections on every surface. We may see devs just apply a blur filter on these in future.


Anonagrog said:
This makes no sense to me. Is there an example of a case where a hybrid-RT implementation would have been better off "turned off" in favour of the standard raster/compute approach? Mind you, I suppose "better" is pretty subjective here. I have a background in engine development, so I'm biased towards putting these seemingly small advances high up on a pedestal, where most consumers wouldn't... and that's understandable.
Click to expand...

Better is not the word that i should have used here. This is the quality vs cost arguement. Is the quality improvements vs the impact to framerate worth it? this is subjective as you've said.


Anonagrog said:
Sure, there will be differences between all engines. However, for the most part it's not too much of a stretch to assume most engines have rendering pipelines that lend themselves to dropping in ray-traced implementations without too much hassle (I define "hassle" here as a few programmers a few weeks to months, without having to change much existing code - depends on RT method. 'Justice' from 'Netease' is an example of this). Things like RT shadows, specular indirect, diffuse indirect, and ambient occlusion, are already handled in a pretty standardised way where inputs and outputs don't force swathing changes to be made. Over the years most game engine renderers follow similar design trajectories to each other due to knowledge sharing and openness in research. That helps in the sense that proposed hybrid-RT approaches have implementation requirements that end up somewhat satisfied already, making adoption easier.



That goes without saying. Justify your efforts using comparisons to the ground truth weighted against the performance cost and controllability. That's literally half of the job and implicit in everything a real-time graphics engineer does.
Click to expand...

My original statement and my line of thinking is mainly routed in the finances of it in relation to the payoff for both the developer and the end user. The questions that i am considering are as follows

You've given a broad range of time for how long you believe it would take to implement (Side note based off the time it took for Blender cycles to be developed i suspect that your timing may be generous but admittedly i am not a developer). How much is that going to cost? Will a developer recoup that when they implement it in older games? Looking at future games what is the size of the player base that can utilise this feature? how much time are they willing to invest to please a handful of PC players running graphics cards far north of £600?

I suspect that devs will put just enough time into it so that it just work on consoles but as we are already seeing they will probably come with the option to turn it off and they will want to maintain the quality of the game even with it off. Few devs will go the extra mile for the 10 PC players running high end graphics cards unless AMD or Nvidia greases their palms.

Which goes back to a previous post of mine
Chuk_Chuk said:
Yeah. In the mid cycle refresh (PS5 pro and Xbox something) or the PS6 and xbox Two X.

Either way few people will be rocking an Ampere or RDNA 2 graphics card by the time RT actually matters, and is not a tacked on feature that is sparingly used.
Click to expand...
 
LtMatt said:
How many titles?
Click to expand...

Why do you keep asking that?

It doesn't matter how many now, that number will only go one way, and that is up.

I dont care if there are currently only a few games, because there are games in those few that I wouldn't mind trying, so number of games are irrelevant.

I just find it odd to release an RT capable card but not the AI upscaling enabler to offset the performance penalty.
 
ljt said:
Why do you keep asking that?

It doesn't matter how many now, that number will only go one way, and that is up.

I dont care if there are currently only a few games, because there are games in those few that I wouldn't mind trying, so number of games are irrelevant.

I just find it odd to release an RT capable card but not the AI upscaling enabler to offset the performance penalty.
Click to expand...
Because it seems like a reasonable question to ask when considering the importance of a feature.
 
LtMatt said:
How many titles?
Click to expand...
To be fair and objective, the number of DLSS will increase a lot next year given the success and the universally good critical reception it is having, that is clear. People need to be thinking about how the situation will be in 6 months time and not just snapshotting how it is now.

AMD will get their own AMDLSS but it will likely take a while to catch up with Nvidia.
 
Richdog said:
To be fair and objective, the number of DLSS will increase a lot next year given the success and the universally good critical reception it is having, that is clear. People need to be thinking about how the situation will be in 6 months time and not just snapshotting how it is now.

AMD will get their own AMDLSS but it will likely take a while to catch up with Nvidia.
Click to expand...
TBH i think the reason we will see it more games is because AMD are so close to Nvidia. Had AMD dropped the ball and delivered a GPU that barely beat the 3070 (as some where hoping) Nvidia would have just continued to trickle in the games with DLSS. but if they had dropped the ball we wouldn't be having this conversation.
 
Status
Not open for further replies.
Back
Top Bottom