Why I'm Excited - An overview of HDRP and CBT Studies

[RayneCloud]

HDRP or the High Definition Render Pipeline..

What is it? 
HDRP is a hybrid rendering solution that supports a lot of modern rendering techniques, things such as rasterized ray tracing and path tracing. Among many other modern rendering features and solutions. The move to Unity HDRP has me (as a lay person, remember I'm just a QA lead that knows enough to understand the very basics here) very excited with the future goals of this project now. Let's see a few nifty bits about HDRP and what it can offer to any projects utilizing it shall we? Let's take a look at some lighting, atmospherics, and terrain videos.

Lighting

Quote

HDRP offers a range of techniques, from traditional rasterization (such as Reflection and planar probes), light baking, ray marching and ray tracing (SSAO, SSGI, SSR, Contact or ray traced shadows), full-frame path tracing, and multiple anti-aliasing options that can be easily toggled or mixed to scale performance and fidelity across platforms.

Clouds and Atmospherics

Enhanced Terrain Systems

Unity HDRP Workflow Demo and Terrain Demo by JaniYa

 

For CBT or "Concurrent Behavior Trees" I'd like to showcase something from SIGGRAPH 2021,

Remember, this is a year + old and has advanced much past this demo video.

 

Here's some nifty Photos with HDRP,

In short, the move to Unity HDRP and moving forward with a CBT based terrain generation system, that those of us in the industry have seen studies and early tech on, is one that has me extremely excited for the future. While I'm not even remotely capable of explaining this like Mortoc, would, or even truly capable of understanding all of it myself... I hope these videos and some of my own research can help you all get a little more excited about the future as well.  This is just me, as a note, exploring and researching and keeping track of tech in the industry and not me saying "This is coming for ksp!" I don't work for IG, I have no clue what they're doing anymore than you do outside current communications, my goal here is just to just educate and excite people and maybe inspire others to pursue a career in game development.

Please take this post at its core meaning, education. 

Edited March 12, 2023 by RayneCloud

[Astr0Guy5]

Oh yeah, same.  I wont pretend I know what these engines are or how they work, but I got very excited when I heard they were upgrading.  I was happy to hear that they were still using old and overloaded engines in KSP2 EA, because that means the framerate and quality will improve dramatically with the new tech.  The future looks very bright for KSP2!

[regex]

As if this game could get any more beautiful... We got a bright future ahead of us in KSP2.

[Rosten]

What's the target date for completing the switch to HDRP?  Also, how will CBT be extended to support planetary-scale terrain?

[RayneCloud]

1 hour ago, Rosten said:

What's the target date for completing the switch to HDRP?  Also, how will CBT be extended to support planetary-scale terrain?

I literally wrote in my post that I do not work on KSP 2. I do not work for IG. Did you read it?  

Edited March 12, 2023 by RayneCloud

[RocketRockington]

Reddit Thread between Mortoc and one of the originators of the CBT algorithm.

As with most things, it sounds more complicated than a silver bullet solution to KSP2s perf issues.  Will be interesting to see how/when any of this comes about.

[MarcAbaddon]

I think it's a good move, at the same time as I feel it's a bit late for such changes. Switching render pipelines usually takes a few months to get everything running again. So really hope the Take Two people are really patient. The game deserves it, but they have to wonder a bit about the many pivots going on.

Edited March 12, 2023 by MarcAbaddon

[The Aziz]

It's all cool and pretty but these are all demo scenes. The rendered environment it's all there is. Now how does that work when there's a full game around it? Are there games that use HDRP and fit somewhat inside KSP2s hardware requirements? Sure the devs said they're not set in stone but me, and many other I bet would hate to see a "hey guys, we're switching to a much more demanding rendering tech so the minimum requirements are going up, kthxbye"

Edited March 12, 2023 by The Aziz

[RocketRockington]

51 minutes ago, MarcAbaddon said:

I think it's a good move, at the same time as I feel it's a bit late for such changes. Switching render pipelines usually takes a few months to get everything running again. So really hope the Take Two people are really patient. The game deserves it, but they have to wonder a bit about the many pivots going on.

But in this case they kind of have to.  Those perf numbers Mortoc posted were abysmal and all related to one area of the code.  And he says it can't be optimized further.

It's a little shameful that they claim to have pushed PQS to it's limits...but sat around using it with those awful perf numbers Mortoc posted till this late in the day.  What else did they spend time on figuring out the rendering for, if not the planets?  There's very little other geometry KSP has to care about or optimize around.

But at least he's saying it's not that much code .. that's a good thing for the hopefuls - it means it's also not necessarily as big of a rewrite.  

 

24 minutes ago, The Aziz said:

It's all cool and pretty but these are all demo scenes. The rendered environment it's all there is. Now how does that work when there's a full game around it? Are there games that use HDRP and fit somewhat inside KSP2s hardware requirements?

Nebulous: Fleet Command and Hardspace: Shipbreaker both use it and are space games - but neither are expansive or graphics tour-de-force titles.  You can see a full list of steam games using it here: https://steamdb.info/tech/SDK/UnityHDRP/

Whats most concerning to me here is that it seems like a decision you would have wanted to make right at the start. Assets do not have compatible settings.  They'll have to go back and touch every asset they've built to look at it with new shaders. 

 https://docs.unity3d.com/Packages/[email protected]/manual/universalrp-asset.html#:~:text=You cannot%2C however%2C switch between,the render pipelines are incompatible.

Overall...it is a weird change to consider when you're having performance issues as it is. Especially when presumably things like atmospheric scattering and cloud tech was written without hdrp in mind and now you have a new lighting model to deal with.

Oh well.  I doubt we'll see any results from this for months and maybe they'll change their minds in the meanwhile.  At the moment though it's giving me star citizen object container streaming vibes though.

Edited March 12, 2023 by RocketRockington

[LameLefty]

2 hours ago, RocketRockington said:

Oh well.  I doubt we'll see any results from this for months and maybe they'll change their minds in the meanwhile

At least a year, would be my guess. There are a LOT of preexisting assets to recreate.

TBH, as interesting as Mortoc’s post was to read, a big voice in the back of my head was screaming, “Why are you JUST NOW considering these changes? None of this is a surprise discovered two weeks after launch.”  And reading the optimism and gee-whizzery of this thread reminds me of the early/mid-KSP development era when great big performance improvements were going to pop up any time now as soon as (A) the devs updated to a newer Unity; (B) 64-bit!; (C) Et cetera, et cetera …

 

[JoeSchmuckatelli]

9 hours ago, Rosten said:

What's the target date for completing the switch to HDRP?  Also, how will CBT be extended to support planetary-scale terrain?

 

8 hours ago, RayneCloud said:

I literally wrote in my post that I do not work on KSP 2. I do not work for IG. Did you read it?  

We don't know.  We are working off two pieces of information - a Shadowzone (maybe others) pre-launch interview with Nate where he refers to a massive overhaul of the PQS and the post by Modoc (Dev 18).  No timelines were suggested. 

Rayne...  

[LeroyJenkins]

Thank you @RayneCloud for sharing your insight on that technology. As I am not in that industry it was very helpful to understand what those are, and what it possibly could bring to KSP2. I really liked the Unity Terrain video, I can imagine the potential with such a detailed environment on different worlds to explore. It could take KSP2 beyond just "Build a rocket and fly to a destination" but expand the science exploration and resource gathering to new levels.
Thank you.

[Periple]

(Edit: Deleted, I was speculating on details of other people's work and I really shouldn't be doing that!)

Edited March 12, 2023 by Periple

[LameLefty]

2 minutes ago, Periple said:

I don't think all the assets need to be recreated, just the materials and any custom shaders they have. And it might even be possible to auto-convert some of the materials. Also some of the assets need a fair bit of optimization work anyway. Converting to HDRP/CBT isn't trivial but it might not be quite as big a thing as it seems, it certainly shouldn't take a year!

The game has been in development for years already and two weeks after launch we still don’t even have fixes for obvious game-breaking bugs. I guess you’re just a lot more optimistic than I am about the rate of development based on what we all already know from past history here if you think a complete rewrite of the same broken terrain rendering system is going to occur on a scale of months rather than at least a year.

[JoeSchmuckatelli]

13 minutes ago, Periple said:

(Edit: Deleted, I was speculating on details of other people's work and I really shouldn't be doing that!)

Welcome to the internet - we do that all the time here! 

Actually - I agree with what you wrote.  This development process will be very interesting to watch. 

@RocketRockington 's link to the Reddit discussion is kind of telling. 

There is fantastic potential to do something great / beautiful with the game... But also making it performant on potato machines may be a bridge too far. 

Although... I'm remembering from the SZ interview that the current focus of the work is to have Mortoc make the game performant at low settings. 

Edited March 12, 2023 by JoeSchmuckatelli

[Periple]

Just now, LameLefty said:

The game has been in development for years already and two weeks after launch we still don’t even have fixes for obvious game-breaking bugs. I guess you’re just a lot more optimistic than I am about the rate of development based on what we all already know from past history here if you think a complete rewrite of the same broken terrain rendering system is going to occur on a scale of months rather than at least a year.

I am optimistic, yes! And it could very well turn out that my optimism is unwarranted.

The reason I'm not too worried about how long it's taking them to get the first patch out is that I know how hard it is to get to a proper, regular release cycle, and what you need to do to and what needs to change when you switch from internal-only builds to published builds that have to go through QA. Even if you know exactly what you're doing it takes a while for everything to fall into place. It's hard and it takes practice!

If things are like I think they are, they will be settling into a regular 2-, 3-, or 4-week release cycle over the next few months. It could be that I'm wrong though and they do have bigger problems they need to solve, but we'll find out soon enough!

[Sturmalex]

I suppose it'll take a good chunk of the year to accomplish, but in the end may very well be worth it.

I trust them being able to deliver the new tech. Companies have swapped gfx engines already in the past.

[JoeSchmuckatelli]

1 hour ago, LameLefty said:

if you think a complete rewrite of the same broken terrain rendering system is going to occur on a scale of months rather than at least a year

I fully expect the port to the other system to be a very long term project.  Nate mentioned it without giving any clue as to an expected timeline - but from what he said pre - release, the implications are that it was underway back then.  The other thing that he said in pre release interviews was that they were aware of performance hits in other areas (fuel, etc) and again those were being worked on (parallel process). 

That said - I fully agree that they were ill served by the Alpha / Beta process - if they existed at all 

[Streetwind]

3 hours ago, LameLefty said:

Why are you JUST NOW considering these changes?

Personal guess: because "now" isn't as late as you might think it is.

The game clearly wasn't as far ahead as we thought it would be when it entered early access, and while I personally don't mind - I enjoy watching things develop - I also understand how other people may have come in with different expectations and feel disappointed. I don't know who decided to go public now, but it was too early. It's not a beta by any definition; it's deep in alpha territory.

Meaning, the game is closer to the prototyping stage of development than we think, and Mortoc's blog post did imply that they went with PQS in the beginning to have something familiar to build systems off of. Maybe they did carry their prototyping setup a bit further than perhaps wise, sure - but it's not like they've built a finished product on top of it. It's far from finished. The fact that the team knows exactly what is wrong, and apparently knew it all along in the lead-up to early access, just confirms in my mind that someone in management pushed for EA too early, and we are now simply witnessing the "switch from prototyping implementation to production implementation" step in the development roadmap that would normally get handled away from the eyes of the public.

[Periple]

Okay I'm going to speculate again!

HDRP was considered production ready in Unity 2019.3. That means that it was pretty new tech when IG started working on KSP2 in 2020. At that time, KSP2 was slated for release in Q3/2021. That was extremely ambitious! If they were targeting anything like that kind of timeline, it would be prudent to stick with the devil you know rather than to take a risk on new, unproven tech that you're not familiar with. 

Fast-forward a couple of years. By now, HDRP has matured and stabilized and there are plenty of developers familiar with it. At the same time it's clear the original solution for terrain rendering just doesn't scale to the ambitions of the game. So they decide that the terrain solution needs a rewrite, regardless of the render pipeline. And since they have to rewrite it, why not at the same time switch to the newer, more full-featured render pipeline that lets them do more cool stuff visually?

— Also, just because Mortoc published the devblog today doesn't mean that these decisions were also done today! It's quite possible the decision was made already some time ago.

Finally: ideally hard performance limits of various systems will be caught earlier rather than later so you don't paint yourself into a performance corner you can't optimize your way out of without major changes to your systems. But hindsight is 20/20 and sometimes unfortunate things happen, especially if your ambition level goes up bit by bit. Perhaps the terrain system scaled perfectly well to what they had in mind in 2021 but the visual targets outgrew it between then and now. These things happen!

[MarcAbaddon]

If we are going to speculate my guess would be that they didn't have a graphics expert for some time before Mortoc, either because they never had one or (probably more likely) due to turnover. If they had someone that person would be more senior and would probably take the lead on it since both render pipelines and terrain rendering are very central to the game. 

[Periple]

6 minutes ago, MarcAbaddon said:

If we are going to speculate my guess would be that they didn't have a graphics expert for some time before Mortoc, either because they never had one or (probably more likely) due to turnover. If they had someone that person would be more senior and would probably take the lead on it since both render pipelines and terrain rendering are very central to the game. 

That would be very weird! Do you really think it wouldn’t occur to them that they might need a graphics programmer?

[Strawberry]

19 minutes ago, MarcAbaddon said:

If we are going to speculate my guess would be that they didn't have a graphics expert for some time before Mortoc, either because they never had one or (probably more likely) due to turnover. If they had someone that person would be more senior and would probably take the lead on it since both render pipelines and terrain rendering are very central to the game. 

Theres credits in the extra tab of the main menu, KSP2 has 4 graphics engineers total, but appears to only have 1 from the star theory day (though its possible some of them got carried over, the people who got carried over arent listed on both sides of the credits). I think its geniunely just sunk cost fallacy, they sunk so much time into refurbishing, expanding  and optimizing this system that for a long time they believed that if you just spent more work on it you could squeeze out enough juice to where it would look good and be performant, but as they went along they slowly realized they would need something else and now we're in our current state.

[RocketRockington]

7 hours ago, Periple said:

That would be very weird! Do you really think it wouldn’t occur to them that they might need a graphics programmer?

It would occur to them, but engineering turnover on that project has been pretty extreme, even if you don't count the star theory mess.  The fact they didn't manage to keep their physics programmer or hire a new one - when this is the sort of project physics programmers typically would be extremely excited to work on - seems like a bad litmus test all by itself.

It's not uncommon for games to have turnover, but so many senior/lead/tech director programmers coming during just the 2.5 intercept years is pretty telling - on linkedIn 8 out of the 16 people that worked there but stopped are engineering staff, and I know there are a couple of others not listed on linkedIn.

Edited March 13, 2023 by RocketRockington

[Periple]

Let’s hope they’ve got things on a more solid footing by now, then!