Post "Console RT" debate, what can we learn from it?

I thought hw RT acceleration was not gonna make it for this gen of consoles, because I though that would be a bet too high for yet unproven tech. I was clearly wrong.

So what?

 

After a quick read of thread, I'm going to throw in this nugget.
Yes we know it's hardware accelerated, how it's implemented, performance, compromises if any, are things we don't know.
Possibly amd gen 1 RT, first gen of things aren't always amazing.

Hence why things like discussing possible bandwidth requirements and how that could impact it is more progressive than everyone thinking they was shouting into the wind or being misrepresented. Lot of effort being used in this thread.

 

I don't think anyone here failed to notice the paradigm shift - the debate was about its actual implementation in the gaming consoles.
My doubts were specifically based on 1) DXR performance issues with GeForce RTX cards, with little room for software optimization, and 2) silicon costs for a mid-range APU part, which are typically used in gaming consoles (with first-gen Navi unlikely to include any form of RT).

I don't recall anyone dismissing the raytracing technology outright. Yes, there were doubts, primarily about cost and performance - and there are still doubts, as neither AMD nor Microsoft or Sony disclosed any benchmarks or other significant implementation details. We've seen it many times when a certain feature is advertised like it's a second coming, and then the actual hardware implementation leaves much to be desired.

I can live with that. The list of people/corporations/governments that failed to predict something significant would be dramatically incomplete without me and the Beyound3D Forum!

 

The difference is what you interpret as strong signs, others saw as mild pointers. There were only really two signs AFAICS - MS releasing DXR and nVidia releasing RTX. These are also released quite close to new hardware, so it's not like the consoles, already deep into development, could easily change plan to incorporate RT if they weren't already planning it. Most importantly, AMD didn't have a hardware RT solution and we knew AMD were providing the hardware for the next consoles. If AMD had RT hardware, yeah, the evidence would be quite solid. But without that, with AMD using compute for their RT implementation of DXR, it's not at all obvious that the consoles would be adding RT hardware to AMD GPUs ahead of AMD's own PC parts. Meanwhile on the software side, we had loads of RT-alternative techs coming and going. Things like VXGI producing great results without needing specialist hardware.

It wasn't at all one-sided as you suggest. There were plenty of alternative possibilities with high probabilities that the writing wasn't on the wall and it wasn't possible to determine which way the consoles would go. That you believed it all pointed to RT hardware, and were proven right, doesn't actually change that interpretation of the data we had. Data == inconclusive. Outcome == unknown.

It'll be more revealing when we learn the details of the hardware, as to whether it's anything we expected or something out of the blue.

 

I didn't know it was my duty to track everything I said in the past and follow up on every bit of news that might alter my previuos opinions. I participate in discussions when I have time to contribute some different point of view. New raytracing pipeline and implications for GPU design and game programming / content creation - that's interesting; simple facts like which part gets it at what time, not much.

You mean, stock market analysts?

 

My first thought was that if you're going to do RT, you need to do it properly, it can't be half-arsed, otherwise you're just wasting die-space on something that's not going to be of much use.

At which point I think, can a console implement a meaningful/effective amount of RT hardware in 2019/2020; I was doubtful.

However, what I found to be more doubtful was a console platform existing in say, ~2025 without RT and the industry still waiting for a new generation so that devs can shift over to RT-only for certain effects in a given piece of multiplatform software.

All the main platforms (particularly consoles) need to switch over to spark a meaningful shift. Even if it's only for select effects and not to the highest quality, once a dev can say we're using RT-only in this specific area, across PC, PS5 & Scarlett, then things can begin to snowball.

I thought it's kind of unreasonable to expect it in 2019/2020, but it's even more unreasonable to expect a platform not to have it ~5 years down the line. As such, I found my expectations weighted towards it being included anyway. There is of course the argument for pro/x iterations but the fundamentals of the game are built around the base platforms and you'd still have to have both non-RT and RT approaches to a given effect, negating one of the big advantages of going RT-only; development efficiency..

You can apply the same line of thinking to the shift to SSDs (which will have vast implications not just to loading but gameplay potential, world building, efficiency and development), though I'd argue that relatively speaking, it's a much more reasonable proposition from the get go.

I think they're both things that needed to be forced up front to get the ball rolling even if the initial investment is big.

On balance RT just had to happen or else the industry would have been waiting another 6-7yrs to really get going with it. It still would have been happening on PC but not to the extent with it being available across the board. You could argue the technology would be "more ready" for prime time, but I'd argue the years of experience to be gained doing select effects to a lower quality and having to do it very efficiently will directly translate to further generations; and be a bigger net gain.

 

If companies based decisions on furthering humanity, yes. However, choices are made on furthering profits.

Hypothetical case, Console A includes RT, Console B just has loads of compute. Console B eclipses console A in launch titles using conventional methods of rendering. Gains critical mass, goes on to become the dominant console, software focuses on it with RT being used as an 'also-ran' feature on console A so it never really hits its stride.

Time and again the inferior tech became the more popular one with consumers and set the standard. That's a very real possibility that these companies face, so they have to balance risk and the cautious choice is a perfectly fair one from a business POV. Nintendo has rarely advanced the technological arts but they've been very successful as a business as a result.

 

@DavidGraham what can we learn from the debate? That you're a bad winner

I seem to recall it was a consensus that RTRT hardware would definitely be in our hands by the time of either PS5Pro or PS6. It was the timing of AMD going from nothing at all, to affordable/mid-range, within 2 years that left some skeptical.

The PS3 released with separate vertex and pixel shaders in spite of the trend towards unified shaders. NVidia then released a series of cards with unified shaders about 6 months after the PS3 launched. So it's not as though there's no precedent for this sort of thing.