Its the same the other way around for those who are opposed to hw-RT.
Its the same the other way around for those who are opposed to hw-RT.
Whom is opposed to HW-RT in this discussion? I missed that bit.
Again this not against HW-RT @JoeJ who is a dev like @Andrew Lauritzen complain about the limitation of DXR. Most developer pushing the envelope don't like DXR and some think we are back to DX7 with current raytracing.
Basically game developers complain since 2018 Brian Karis and Sebastian Aaltonen alias @sebbbi talked about LOD in 2018 just after the release of DXR specification.
Random Thoughts on Raytracing · Aras' website
aras-p.info
Andrew Lauritzens posts have confirmed my former speculations and critique, besides Karis comments on twitter.
It's true: RT APIs can't do LOD.
But if your agenda dictates, prefer to classify truth as speculation, just because you don't like it to be true.
Their users decide which path to use.
Neither of us blames them to give both options.
But both of us should blame the failure on API design, so we can get that RT future you desire so badly, and i need as well.
I did not mean mobiles. I mean PC platform. Last time i checked RT support on Steam HW survey it was less than 20%. Idk what's the percentage now, but i would support old non RT HW during the entire console generation, at least.
Currently RT is the option, not the standard. Due to bad economy, Moores Law stagnation, and success of low power portable gaming, i'm no longer sure we'll ever have 'powerful enough' HW RT an all platforms.
I believed in this when RTX was announced much more than i do now.
Regardless of my opinion, SW alternatives to lighting, if they are faster and better, do not hold back progress of RT, nor do they turn RT redundant.
RT is not the best tool for every problem, but it's nice to have it. There is not really a disagreement about that. But still we manage to divide us into two camps, forcing both sides into narrow and extreme corners, turning arguments weak and attackable, making both of us look silly.
Yes, but it's all about games. World design, artist workflow, compression requirements are more important than photorealism.
And it's not a HW issue - it is a API issue, so actually SW. That's my whole point of why i think we should not argue even if details about our visions differ.
It can be fixed, so we should focus on agreement on the fix, instead focusing on disagreement on which path will bring us to the common goal.
Why am i almost the only one here proposing such fix? Why do so many defend a broken status quo?
Yes, but my point is: A software implementation of RT could support Nanite. Because there would be no API blackboxes preventing a 'lodable' BVH.
HW can support Nanite. It's just triangles in a tree, so yes. There is no need to change the HW.
All we need is to open up BVH. So we can modify the data structures as needed to implement LOD changes.
That's the only way forwards.
The alternative is custom formats to implement something like Nanite in HW directly, using custom, blackboxed, probably even proprietary data structures.
But that's the bad way. Forwards for some time at limited focus, but again backwards on the long run. Because:
Nanite and LOD is just an example to illustrate balckboxed BVH is not acceptable. We need control over dynamic geometry in general, and BVH is too useful to restrict it to RT only.
Now in a time where HW progress is stagnating, we need flexibility more than ever. Because felxibility enables SW progress, and SW progress becomes now more important and faster than HW progress.
Notice this is not my guessing or predictions - it is the only conclusion besides having no more progress at all, which i'm not willing to accept.
You say that even more fixed function is the proper reaction to the stagnation, but time already proved you wrong. Because of that raster -> T&L -> VS+PS -> general compute progression, which became the winner. So the trend is away from fixed function, not towards it.
The real and only problem here is that RT, as an entirely new and complex feature, does not evolve fast enough towards more flexibility.
And the irony is that it's not related to fixed function HW, but only comes from API design failure, politics and bureaucracy issues.
So again, our differing opinions do not really matter and are pointless to discuss.
However, personally i can not get out of my 'contra RT corner' as long as you guys do not accept the fact that there is a problem.
There is a reason why every professionell software company switched over to hardware accelerated raytracing. It is >3x faster than a pure compute solution.Again this not against HW-RT @JoeJ who is a dev like @Andrew Lauritzen complain about the limitation of DXR. Most developer pushing the envelope don't like DXR and some think we are back to DX7 with current raytracing.
Basically game developers complain since 2018 Brian Karis and Sebastian Aaltonen alias @sebbbi talked about LOD in 2018 just after the release of DXR specification.
Random Thoughts on Raytracing · Aras' website
And all of them use DXR and all of them are quite happy with it? Could you ask them for me because I personally don't know them all.
Then maybe this belongs more in an nVidia RT thread? Unless you're just trying to fanboi nVidia up in AMD's RT thread...
It's possible and plenty of demos with billions of triangles have been shown. Take a look at NVIDIA research for example, and hopefully you will find these multi billion polygon path traced scenes traced in real time on 3090.
There are many ways how you can get away without LODs - with instancing, with micro meshes, etc.
It's actually you who think that UE5 solution of LODs is only practical solution (though it's not even fully featured and does work only for non deformable geometry)
It depends on what kind of LOD it's. UE5's LODs are practically voodoo for anyone but Epic at the moment. Classical discrete LODs are industry standard and I don't see why other major game engines would want to go UE5 way when a simple SW rasterizer or even RT with classic LODs would suffice for micro triangles.
The same way it's done now with discrete LODs, told it to you before, it's also how it's done in UE5 for distant objects BTW.
Sorry, but you sound like a broken record and argue with phantoms in your head. I didn't say that LODs are useless anywhere to begin with. Complex LODs are obviously not your standard discrete LODs that were implemented into every major engine decades ago. Also, virtualized geometry is the same stuff as virtualized textures vs textures, you don't need UE5's LOD system to make games with micro triangles in the same way you don't need megatexture to make high density and variable textures.
Some great innovation comes now form compute shader Nanite, strand hair software rasterizer analytic AA of Frosbite hair. I remember the reaction on twitter of the technical director of Sony Santa Monica after the first demo of Nanite without technical detail, good luck to do a full game with so many details.
still not seeing how this is a failure point. To me, and this is no disrespect to any developer, things have to mature to the point where you are allowed to get full control over the pipeline.
It took forever to get to compute and we're still just on the edge of it, but how many years was this really in development for. RT really just started. I think people are asking for too much too soon.
I think we need to ask the obvious question of whether or not, a programmable RT pipeline would have had performance, because that's like saying all of our initial GPUs should have gone with compute from the get go and there should not have been fixed function hardware, and I don't think that's true.
It's an evolution that requires innovations on both the software and hardware side of things.
The difference compute was not existing when first GPU release now it exists and we just have probably the biggest compute innovation with Nanite and not enough flexibility is a problem but at least they solve the problem with software lumen.
Needs of professional software company or virtual production aren't the same than videogame developer.
I think the RT/ML fanboys are worse, as they ignore the valid arguments speaking against the hype.
Contrary, there are no conservatives or AMD fanboys who say those things are totally useless, or AMDs RT is equally fast or better, or such things.
It feels like doubtless convictions and predictions about a hyped potential future and presenting this as already state of the art, vs. rational concerns about cost / benefit ratio and real economical limits.
However, being one of the latter camp, surely i can't judge. I know this. But as said - because the discussion never arrives at consens, you are sadly right. No matter who's more wrong, the outcome is annoying, heated, and pointless.
Because there is no consens, probably both sides are wrong in their polarizing attitude.
Basically the disagreement, including industry veterans, confirms how fucked we are, and how uncertain actually everybody is.
Idk, but maybe we need to let loose from the idea of a unified gaming culture and economy. The thing became so large, it no longer fits under a single hat.
There is no more common vision of how gfx should look, what's the proper cost to achieve improvements, what's the expected HW power needed to play modern games.
For example, while we sit here in our enthusiast echo chamber, shitty Switch outsells next gen consoles and 4080 does not sell at all. So to me, the enthusiast corner feels much more old school and conservative than anything else in gaming.
But maybe that's the wrong view to look at it. Something becoming more and more niche does not mean it's wrong or bad business.
Well, only future will tell.
I have a 6950 XT in the house, btw. Not mine, but my son ordered one. So i'll soon have a look on glorious RT and high refresh display myself for the first time. : )
Still need to ask him why he preferred AMD this time. He likes NVs 'progress', and Elons as well, so i truly wonder about that...
You used Optix if you want to research improvements on importance sampling, denoising, etc. Classical RT is perfectly fine for that, as it is for offline movies.
But you would not have used it if you wanted to research raytracing dynamic LOD geometry, for example.
The problem is that Optix API became DXR, without spending a thought on level of detail being still an open problem. They closed the door to solve this.
The question is why. I assume they thought games do well without addressing LOD so it's not yet important. Til this changes, classical RT should give devs enough options to achieve progress for years.
If so, they did just underestimate game devs and their role in achieving progress. It also shows they did not talk enough to them.
That's no argument at all. Nanite relies on instancing to achieve compression, so it builds on to of that.
But instancing did not and won't give us practical high detail.
I won't search for it. Their demos are impressive, but it's just that. Demos running on unaffordable high end, spending all resources on restricted facades scenes. No game, self running, etc.
I mean, i know realtime PT is possible. But that does not mean it is the best or even a practical option for games.
If i would think this i would not still work on a different LOD solution.
But i agree with all their conclusions, though. Read Karis Path to Nanite slides - he's right with pretty much everything. Alternatives are even more specific and restricted, true continuous detail with texture mapping is impossible, and clusters of discrete LOD triangles are more compatible with HW RT than voxels, points, or morphing triangles. Deformable geometry is possible, but not important.
But enlighten me: What is your proposal for a practical LOD solution, which also is (in theory) compatible with triangle raytracing?
Instancing is no LOD solution, so does not count. NVs DMM is no LOD either, it's a displacement amplification technique which can only scale up, not down.
So what? I'm all ears...
I could implement it in 6 months including the tools, i guess.
Not because i'm so smart, but because thy made everything public and gave multiple presentations. So i would not need to repeat their path of failures to get there (i was there already anyway).
Saw 2 guys on YT implementing their Nanite toy renderers already.
No voodoo at all, at least not to those with a wider horizon than NV research.
Oh, so your proposal is no progress at all, but instead sticking at discrete LODs? It somewhat works, but have fun with still competing Epic in 5 years.
Sadly this seems what the whole industry is doing. Disappointing if so.
I think you should try harder. You're reputated to deliver cutting edge (assuming i draw the proper conclusions from your nickname). Don't give those US techies the lead in absolutely everything without a fight >
Exactly. How nice it would be if we could start admitting some agreement.
However, it's entertainment business. At least that we get right and deliver \
/Silent_Buddha
Legend
Here's a summary of where many of us are at. And it isn't anti-RT.
Regards,
SB
- We all want RT. Full Stop.
- We don't think the hardware is quite there yet for completely convincing RT without significant compromises in other areas that some of us find worse than what current RT implementations bring.
- Which again, doesn't mean we are against RT.
- We want RT.
- Current hardware is a step towards that goal that is needed and companies bringing the hardware to us regardless of whether or not it's NV, Intel or AMD and regardless of how performant the hardware, it's still appreciated.
- Which again, doesn't mean all of us are happy with the compromises involved for current RT implementations.
- And again, doesn't mean we're against RT.
- We, well most of us probably, are happy that there are people that find current RT implementations worth the price of admission.
- That's good, because otherwise RT risks dying a short death.
- Thus having people that find current implementations well worth the compromises involved is needed for the market.
- What isn't needed is having some proselytize to the point where it almost feels like Jehovah's witness saying that you're damned if you don't agree with our collective and embrace your lord and savior RT into your heart.
- What isn't needed is having some proselytize to the point where it almost feels like Jehovah's witness saying that you're damned if you don't agree with our collective and embrace your lord and savior RT into your heart.
- Keep in mind, it isn't just "AMD fans" that aren't happy with current RT implementations, owners of NV cards as well are not always convinced and many run games with RT disabled even on their RTX 3xxx cards.
- Again, it must be stated that all of us (or at the least the vast majority) can't wait until RT gets to the point where it's actually worth it to have it on whenever the option for it is presented.
- What X person finds acceptable isn't necessarily what Y person finds acceptable.
- Hell, even developers that are fully on the RT train with attempting to implement RT into their games aren't entirely happy with the state of RT hardware and where it's at.
- That's why they question design choices and constantly talk about how the hardware could be better if they had more control over how aspects of RT is handled (the BVH tree, for example).
Regards,
SB
Bondrewd
Veteran
B i n g o.
Again, as far as your average 3060 buyer goes, RTRT is still a funny IQ gimmick feature he turns on once, sees it tank his FPS by quite a lot then turns it off to never toggle it on again.
That's the market perception and it's not gonna change in the near future.
Sure, but then, why would these people pay $999 for a GPU?
Bondrewd
Veteran
To drive their shiny new 4k@144 display, that's why.
I feel like a variation of these arguments always come up when something new and demanding comes up or even simple resolutions.
I would love to have ray tracing on all my new games. It looks good. But even my 3080 struggles to run the majority of them natively at resolutions I want to play at or heck even with reconstruction. Meanwhile there are still plenty of great looking games coming out in the future and games I am already playing that are traditionally rendered that i would like more performance out of
I would love to have ray tracing on all my new games. It looks good. But even my 3080 struggles to run the majority of them natively at resolutions I want to play at or heck even with reconstruction. Meanwhile there are still plenty of great looking games coming out in the future and games I am already playing that are traditionally rendered that i would like more performance out of
Why wouldn't they ?
Listen some people buy new $1000 gpu's every year or every other year. So someone can drop $1k on the 7800xtx and be happy that the current games they are playing are getting a huge boost in performance. Then they can go and sell the card in a year or two and buy a new one and get even more performance and maybe they will want RT performance this time and go with nvidia or maybe amd would have caught up. Who knows. But everyone has different needs and wants.
Look at it this way. The $900/$1000 radeon could trade blows in RT with the $900 geforce while offering better traditional rendering performance and what 8 to 12 gigs more memory ?
Similar threads
