Ray tracing the new "physX"?

[Spjut]

I don't see DXR taking off unless the next-gen consoles support it with good performance. And I'm doubtful if they'll be able to do both the expected minimum 1080p/1440p checkerboard rendering, High/Ultra graphics and then raytracing on top of that.

It's worse than PhysX. You could run PhysX on most Nvidia cards that were available at the time, and if you had an old card, you could use it as a PhysX processor. This ray tracing stuff seems to have a huge performance hit and Nvidia wants about 5 people on the planet to be able to afford it. It may be the future, but I don't think it will be Nvidia proprietary or RTX. I give it 10 years before it becomes any kind of actual thing. Seriously.

I still feel bitter when thinking of how Nvidia disabled using a PhysX card in combination with AMD

 

[psolord]

Would it be possible for a standalone Ray tracing card to exist? Like that old Ageia card that brought physx? Something that a user could add to his system and collaborate with any gpu? Wouldn't that bring die costs down? Maybe at the expense of some latency?

 

[sxr7171]

It's quite early days. The BFV presentation made it most clear showing RTX on/off in many shots and explaining the limitations of old techniques and how RT "just works".
https://www.youtube.com/watch?v=WoQr0k2IA9A

At some point in the next couple of years I expect RT reflections and shadows will commonly be part of "Ultra" settings, and if you don't have a RT capable card, you don't get "Ultra", and I also expect that developers will start doing less work on faking high quality shadows/reflections at lower quality tiers, because the old way is more work for less quality.

Developers are all going to want RT to succeed, because it is higher quality for less Dev work.

RT is inevitable.

Totally agree. It’s too much work to do this prebaked lighting and these game budgets are insane. It will happen but this is early and you will pay a premium. Honestly I can’t resist technology and will probably be one of those early adopters. However I won’t cry next year when it gets done right on proper 7nm.

 

[dacostafilipe]

The idea of RT is great, but do we really need that extra hardware for those "side-effects"? I mean, the RTX demos looked great, but it's not like the games looked bad with those effect off, specially the Tomb Raider one.

Is it really impossible to run RTX-like effects on todays hardware, even with lower effect quality (ex: half resolution real time reflections)?

For now, I'm not impressed. Waiting for more ...

 

[Jaskalas]

Developers might enjoy RT if it means they don't have to spend all that time developing specialized shadows. Unfortunately, they won't get that break until all hardware can run RT. Assuming RT becomes "affordable" in 2 years, "mainstream" in 4... it'll be 10 years before they can drop standard shadow technology.

 

[LTC8K6]

You'd probably have to have two 2080ti cards using NVlink with their combined 22gb of VRAM to run RT at any decent resolution.

 

[PrincessFrosty]

The idea of RT is great, but do we really need that extra hardware for those "side-effects"? I mean, the RTX demos looked great, but it's not like the games looked bad with those effect off, specially the Tomb Raider one.

Is it really impossible to run RTX-like effects on todays hardware, even with lower effect quality (ex: half resolution real time reflections)?

For now, I'm not impressed. Waiting for more ...

A better question would be, what else can we spend the hardware on? We've basically solved geometric complexity problem with high resolution models and tessellation allowing the scene to scale up dynamically with polygons, we're reaching diminishing returns throwing hardware at more polys. We've basically also solved the texture problem, games like GTAV are a testament to that, you can throw 80Gb of textures at a huge open world and we can stream those high res textures into the game without filling something like an 8Gb vRAM on a modern video card. We seem to have near infinite freedom with shaders and dynamic effects these days. The last real area is in accurate shadow and lighting and for that if you want better you really need to move to RT, we've faked things just about as far as practically we can with current technology, we've bumped into the limits of current techniques and the only real way forward is ray tracing.

Others here are right it'll probably be a order of a decade before RT is mainstream and every developer is using it, but we have to start somewhere, we always have to start somewhere, in that decade we get to learn all the lessons, do the optimization. It's like this with every new step forward in tech with graphics, I've seen this every since the first GPUs it's just part of the maturation process. If they didn't have something new to spend all that power on, most gamers would never buy the next GPUs unless essentially you're running 4k and you need that last drop of power which like 95% of gamers are not. If you're in 1080p or 1440p then you'd have no reason to get the next GPU, and then how does the industry progress?

 

[Midwayman]

I feel like ray tracing is one of those inevitable technologies. I feel like we've been talking about it for decades and it just hasn't been possible. Someone just had to make a first serious attempt at a consumer product. Its success hinges on a couple things though. Mostly how easy it is to implement. If its literally just a render setting they can turn on it becomes something that just hides in ultra on future titles. It'll probably become something that's just available in most middleware packages. For it really to become popular in hardware I think it'll take them figuring out how to do multi gpu on interposer. That will allow them to avoid massive die sizes and make sure lower end cards still have reasonable RT performance. For a long time they will have to support both raster and RT paths and that would allow them to easily mix and match performance between the paths.

 

[lifeblood]

Unlike PhysX, Ray tracing and VR are the future, the only question is whether the future is now or later. Real-time Ray tracing is necessary for photorealism in games. We’ve been talking about it for a decade. Intel’s Larrabee from 2010 was supposed to be good at ray tracing. Given ray tracings inclusion in DirectX I2 I think the industry wants it and will eventually make it happen. The problem has been and still is the amount of hardware required to execute it in real-time. Die space is expensive and until the die space required for meaningful use is reasonable it won’t get included. Note here, If that die space can be used for other things as well (AI, whatever) than it becomes more acceptable for including it.

Nvidia owns the high end and will for the near future. Vega and Navi are both meant as console GPU’s, and that means midrange. AMD tried pumping Vega up to a high end GPU and it didn’t work so well. Maybe Navi will scale better but I’m not holding my breath. So really Nvidia is going to dictate the rules (and prices) at the high end until at least 2020. Will that be enough to force games to include real ray tracing effects? I don’t think so.

AMD needs to support ray tracing. They currently own console GPU’s and that sets the standard for PC games. If the console makers want ray tracing, AMD will implement it. If not, then games won’t include it in any meaningful manner. I would argue that including ray tracing is a no-brainer, unless it takes too many dedicated hardware resources than it probably won’t show up in consoles any time soon. As I said earlier, die space is expensive.

I don’t know if Navi includes ray tracing. If not then maybe the new uArch AMD is planning post Navi will include it.

The disrupter in all this is Intel. If they come out with ray tracing in their new GPU, and the GPU is actually competitive with Nvidia and AMD’s, then I think that could do the trick. And with a competitive GPU Intel could also take the console market away from AMD and that could spell doom for AMD.

 

[JeffMD]

That's fair.

Nevertheless, in a fluid, fast paced game environment, I'm not sure I'd really notice.

I play many games that arn't always fast paced. RPGs and adventure. Heck cyberpunk is right around the corner.

Also I belive we do not see much active gameplay use of reflections because of the limitations of its accuracy, imagine though game worlds finally being designed around the intention of exploiting it. The flame thrower demonstration was a fairly simple street corridor layout with strewn vehicles as cover. Before you couldn't see what was happening while behind cover unless you popped your head up. On the other hand boom, introduction of shiny windows and you can now determine where (gun)fire is coming from.

Is it worth buying $500 right now for new cards for it? maybe not.. but I think its a good future investment that we can expect to all have on our next video card.

 

[dacostafilipe]

Great video that explain RT on the BFV demo : https://www.youtube.com/watch?v=8kQ3l6wN6ns

 

[Rifter]

I think until Nvidia and AMD price RT cards capable of RT 1080P 60FPS in the sub $200-$300 price range that we are getting way way ahead of ourselves.

Nothing requiring $2k+ in video cards is ever going to take off, ever.

Maybe in 10-15 years but certainly not going to become mainstream any time soon.

 

[LTC8K6]

I think what I have seen of NV's RT does indeed look very good and it does look promising for helping you play games better.

I also think that it does not look good enough to justify what it seems like you'd have to give up to use it.

I think Rifter nailed it with the idea that a 1080P 60fps RT card should cost ~$250.

For a $600 and up card, you ought to be able to run wild, especially with all the RT talk and presentations, along with the CPUs we have now.

Would you buy a $250 card that could do great looking RT graphics at 1080P and 60fps minimums?

But, the market is what it is, and if NV continues to be successful, then we'll have to live with what is offered to us until some competition comes along.

 

[PeterScott]

If all the negative attitudes here, were reflected in the dawn of 3D cards, we might still be playing 2D games.

Think how few games supported 3D initially. Think how long it took before there was a card that was good at both 2D and 3D.

I was a PC gamer in the era of the 2d-3d transiton. My first PC gaming cards were fast frame buffer cards with chips from Tseng. I didn't go 3D until Voodoo 3.

Ray Tracing is not another PhysX, it's the biggest upgrade since 3D itself, and the issues are trivial.

Just being slightly higher cost, and you can always wait that out, since no one is saying everyone need be early adopters, but acting like there is some great penalty for RT HW on this is release is nonsense compared to what you had sacrifice, or the hoops you had to jump through in the first generation or two of 3D gaming HW.

 

[Rifter]

If all the negative attitudes here, were reflected in the dawn of 3D cards, we might still be playing 2D games.

Think how few games supported 3D initially. Think how long it took before there was a card that was good at both 2D and 3D.

I was a PC gamer in the era of the 2d-3d transiton. My first PC gaming cards were fast frame buffer cards with chips from Tseng. I didn't go 3D until Voodoo 3.

Ray Tracing is not another PhysX, it's the biggest upgrade since 3D itself, and the issues are trivial.

Just being slightly higher cost, and you can always wait that out, since no one is saying everyone need be early adopters, but acting like there is some great penalty for RT HW on this is release is nonsense compared to what you had sacrifice, or the hoops you had to jump through in the first generation or two of 3D gaming HW.

Id hardly call 90%+ of the market being priced out a trivial issue.

Bottom line is the vast majority of PC gamers are running sub $300 cards. Unless they get RT working well on sub $300 cards its never going to take off, because no game designers will waste the money coding on something no one has hardware to play.

So sure, it may very well be amazing, but until prices drop its not going to gain much traction. Maybe in 10 years it will happen, but no time soon.

 

[PeterScott]

Id hardly call 90%+ of the market being priced out a trivial issue.

Bottom line is the vast majority of PC gamers are running sub $300 cards. Unless they get RT working well on sub $300 cards its never going to take off, because no game designers will waste the money coding on something no one has hardware to play.

So sure, it may very well be amazing, but until prices drop its not going to gain much traction. Maybe in 10 years it will happen, but no time soon.

You mean much like $300 3DFX Voodoo 1, priced out 90%+ of the market.

Did it take 10 years for 3D to gain traction?

 

[Rifter]

You mean much like $300 3DFX Voodoo 1, priced out 90%+ of the market.

Did it take 10 years for 3D to gain traction?

Are you seriously trying to compare going from 2d to 3d to going from 3d to rt?

 

[PeterScott]

Are you seriously trying to compare going from 2d to 3d to going from 3d to rt?

In a thread that compares PhysX to RT sure.

It isn't as big as 3d, but it's the biggest thing since 3d.

 

[Midwayman]

In a thread that compares PhysX to RT sure.

It isn't as big as 3d, but it's the biggest thing since 3d.

No, thats VR. RT is nice but ultimately its a fidelity upgrade. VR totally alters the game similar to the 2d to 3d jump.

 

[PeterScott]

No, thats VR. RT is nice but ultimately its a fidelity upgrade. VR totally alters the game similar to the 2d to 3d jump.

VR hasn't changed Video cards.

But Raytracing will quickly be a must have for VR. Accurate lighting/shadow/reflections might not be a huge deal on a 24" monitor but when you are actually in the VR world, where you move your head changes the effect of lighting/reflections/shadows, that extra accuracy/realism will be a huge deal.

 

[ozzy702]

VR hasn't changed Video cards.

But Raytracing will quickly be a must have for VR. Accurate lighting/shadow/reflections might not be a huge deal on a 24" monitor but when you are actually in the VR world, where you move your head changes the effect of lighting/reflections/shadows, that extra accuracy/realism will be a huge deal.

I agree, and we're 5+ years out from RT seeing mass adoption. This is a good step in the right direction but it's certainly not one that will make me jump from a 1080 to a 2080. This generation is simply not attractive to anyone concerned with value who has at least a 1070. 7nm or bust.

 

[ZeroRift]

RT is a significant and (IMO) inevitable paradigm shift in real-time rendering.

RT has always been the gold standard for image quality in offline rendering, and the only reason it wasn't commonly used at the very beginning of real-time 3D was its massive compute costs. I still remember playing with mirrored surfaces in U:ED back on my Voodoo 4, and being amazed at how cool RT was and also how it would drag my system to a crawl.

RT obviates the need for multiple rasterization techniques, such as AO, which are becoming quite computationally expensive in their own right. So it's only natural to move to the simpler coding solution as soon as the compute resources become available.

It's my belief that NVIDIA chose this moment to release RT as a strategic decision to move the industry forward knowing that they could still beat AMD in conventional rendering even with the "wasted" die area for games that don't benefit from hardware accelerated RT.

However, I also don't predict any sort of abandonment of "traditional" rendering pathways happening until every sub ~$300 GPU has reasonable RT performance.

In the short term, I expect RT to have the same kind of industry effect as tessellation. NVIDIA will be a lot better at it than AMD for several generations, and will leverage their advantage by "optimizing" ray counts in green titles just like they did with tessellation factors.

 

[Thala]

I agree, and we're 5+ years out from RT seeing mass adoption. This is a good step in the right direction but it's certainly not one that will make me jump from a 1080 to a 2080. This generation is simply not attractive to anyone concerned with value who has at least a 1070. 7nm or bust.

Why do you care for mass adoption? Using RT is the like new Ultra-settings in games which support it - without RT you will not get the best possible quality and still have to live with fake reflections, shadows and global illumination. So it is attractive for anyone, who where shooting for Ultra-Settings, this includes those with 1070, 1080 and even with 1080Ti.

 

[PeterScott]

It's my belief that NVIDIA chose this moment to release RT as a strategic decision to move the industry forward knowing that they could still beat AMD in conventional rendering even with the "wasted" die area for games that don't benefit from hardware accelerated RT.

However, I also don't predict any sort of abandonment of "traditional" rendering pathways happening until every sub ~$300 GPU has reasonable RT performance.

In the short term, I expect RT to have the same kind of industry effect as tessellation. NVIDIA will be a lot better at it than AMD for several generations, and will leverage their advantage by "optimizing" ray counts in green titles just like they did with tessellation factors.

I think you nailed on the timing of NVidia introducing RT now. If NVida and AMD were neck and neck, I suspect we wouldn't haven't Ray Tracing yet, as they couldn't afford the die area losses with a competitor breathing down their neck.

Though you won't see an abandonment of "traditional" rendering pathways soon, I suspect we will quickly see them given less polish, in favor of using RT pathways to showcase a games best visual presentation.

 

[ozzy702]

Why do you care for mass adoption? Using RT is the like new Ultra-settings in games which support it - without RT you will not get the best possible quality and still have to live with fake reflections, shadows and global illumination. So it is attractive for anyone, who where shooting for Ultra-Settings, this includes those with 1070, 1080 and even with 1080Ti.

RT incurs way too much of a performance hit even on the $1,200 2080TI. When the top tier GPU can push 100fps+ @ 1440p or better with RT it will be a viable technology and one that would get me to pay $$$ for a new cutting edge GPU. As is, even the 2080TI can't effectively implement RT so it's a non-factor in my purchasing decisions. 1st gen 7nm may make RT something that matters or it may take 2nd gen 7nm GPUs to get there, only time will tell.