KSP2 at GDC23

[Dantheollie]

Hey everyone, the KSP2 team had a dev-booth at GDC23. Does anyone have any info on what was being presented in their booth?

A recent tweet alludes to them showcasing something they've "been working on"

[Superfluous J]

Well, they've been working on KSP2, so my guess is they'll show that.

 

 

 

 

[The Aziz]

https://schedule.gdconf.com/session/houston-we-have-a-planet-the-spherical-terrain-of-kerbal-space-program-2-presented-by-unity-technologies/894320

 

[Gotmachine]

17 hours ago, regex said:

https://schedule.gdconf.com/session/houston-we-have-a-planet-the-spherical-terrain-of-kerbal-space-program-2-presented-by-unity-technologies/894320

The video should be made available on the Unity YouTube channel the week of April 3.

It will be interesting to see what they have to say, especially since they admitted last week that the terrain system they have put together doesn't work and that they are planning to scrap it.

Edited March 23, 2023 by Gotmachine

[Strawberry]

While the terrain is definitely a big step down in terms of performance, I think its undeniable that the terrain system is a big step up look wise from KSP1 (If you want to see it from orbit look at duna, if you want to see what the terrain system can do look at the geography of Pol). Im guessing the talk will go basically "We started with PQS because ksp1 did it and it was easy, heres all the little knick knacks we've added to PQS, here's when we ultimately realized that we couldnt get PQS as performant as we'd like with the graphics we like, here's what we're doing with CBT". 

[Fluke]

2 minutes ago, Strawberry said:

If you want to see it from orbit look at duna, if you want to see what the terrain system can do look at the geography of Pol

Duna is amazing compared to KSP1, Ike is pretty impressive too. I avoid looking at Pol screenshots because I want to see it for myself, it seems the most overhauled/complex to me.

[RocketRockington]

Anyway, staying on topic:

Hearing 2nd hand from people who went to the talk, I hear that the talk consisted of KSP1 PQS  and more expensive textures and shaders.  Nothing really that you'd call 'PQS+'.  (Yet again, Nate was overselling/overhyping) And apparently at least the group that went to GDC doesn't seem to be talking about giving up PQS, just a brief reference to something that might be CBT.  And of course they didn't fess up to how god-awful the performance is.

So overall maybe the mostly-artists that gave the panel don't know what Mortoc is working on, or CBT is considered to be a long shot they're not banking on.

Edited March 24, 2023 by RocketRockington

[Vanamonde]

Keep it on-topic, please. Quite a few comments removed. 

[J.Random]

10 hours ago, RocketRockington said:

Hearing 2nd hand from people who went to the talk, I hear that the talk consisted of KSP1 PQS  and more expensive textures and shaders.

So I was right in my [snip] prediction. This. Is. Pathetic. And judging by "Format: Sponsored Session", it was probably a management's brilliant idea, agreed upon months before the release. Probably thought up as a "success story" or something.

When the video inevitably emerges, here's something to compare it to:

2013.

Edited March 24, 2023 by Gargamel
Portions redacted by moderator

[Kerbart]

46 minutes ago, J.Random said:

So I was right in my [snip] prediction. This. Is. Pathetic. And judging by "Format: Sponsored Session", it was probably a management's brilliant idea, agreed upon months before the release. Probably thought up as a "success story" or something.:

Well, yes, this is how these conferences work. Months in advance tickets go on sale, with a detailed agenda, as no one wants to spend $1500 on a conference without knowing what the presented topics are. So months before that speakers are requested to submit their topics and the most interesting ones are picked. At least six months ago, if not longer.

It's not inconceivable that IG was working on a better system and initial tests looked promising. Then some showstoppers were discovered during testing. Do you go ahead and hope you can fix the issues on time? The agonizing decision was made to revert to PQS because, well, Feb 24 was on the horizon. This is all speculation and conjecture of course. But it's also a good illustration of what have they been doing in the past three years. Working on improvements, that not always work out. And reverting back to the original system isn't just a matter of swapping out two libraries, as there probably is a ton of code depending on it as well. If improving code was only as simple as adding the line make_code_go_fast = true;

Edited March 24, 2023 by Gargamel
Edited Quote

[J.Random]

25 minutes ago, Kerbart said:

Well, yes, this is how these conferences work. Months in advance tickets go on sale, with a detailed agenda, as no one wants to spend $1500 on a conference without knowing what the presented topics are. So months before that speakers are requested to submit their topics and the most interesting ones are picked. At least six months ago, if not longer.

No, no, that part is obvious. I was commenting on the "sponsored" part. I'm assuming that it means that IG paid for their devs to appear, so topic being "interesting" doesn't really matter much.

As for what they were working on: well, it was PQS+, the stuff they released. I don't recall them mentioning [snip] CBT at all before the release. So I still have no idea what they were doing for the past 3-5 years.

Edited March 24, 2023 by Gargamel
Portions redacted by moderator.

[Kerbart]

1 hour ago, J.Random said:

No, no, that part is obvious. I was commenting on the "sponsored" part. I'm assuming that it means that IG paid for their devs to appear, so topic being "interesting" doesn't really matter much.

Ah, I see. I would still be optimistic about the interestingness of the presentation though.

My experience with non-(game-) dev conventions is that if you're a corporate sponsor at a certain level you do get slots. And it would be a waste not to fill them. Not sure if it was IG, PD or T2 who was the sponsor but given that KSP2 runs on Unity, one way or another they'd seem like a good candidate to fill that slot, one assumes (I doubt the sponsorship money came from the IG budget).

And yes while corporate and marketing will quickly label anything "interesting" (I once signed up for an "interesting" podcast that suggested data science and the next three episodes were interviews from HR with VP's celebrating their 25th anniversary with the company, but I digress), the developer who's actually doing the presentation doesn't want to look like a total tool, so it probably will be interesting. Even if it's just regular PQS he might give insights and what was tried and why it didn't work; no doubt others can learn from that as well (no experiment is ever a complete failure; at the very least it can serve as an example to everyone else).

Edited March 24, 2023 by Kerbart

[whatsEJstandfor]

Looks like the recording of their presentation went live a little bit ago!

 

[JoeSchmuckatelli]

5 hours ago, whatsEJstandfor said:

Looks like the recording of their presentation went live a little bit ago!

 

I posted this in another thread, but it's relevant here.

...

That was interesting.  3/4 of the presentation was all about how they've done what they've done up til now - using the PQS+ system they got and tweaked from KSP. 

Moving into the CBT gets mentioned, if not talked around, in the section of the video labeled "Early Access - Tool Plans".  It's the shortest segment of the video.  Given that we got Mortoc's Dev Insights # 18 on the 10th of March and the GDC presentation happened 10-15 days later... it sound like the artists and engineers on the panel are new to the whole idea.  They like it ("super pumped") - but it doesn't read like something that's been in the pipeline for a year - rather something that they only started maybe in February, if not March itself.  The actual phrase used was "Investigating a new terrain system".

A few other tidbits:

• "Science mode - which will introduce a campaign"
• "Gurdama... Introduced during the Interstellar update"
• "[they were using a built-in render pipleline right up to EA Launch and did not have access to HDRP]"  but they do now - or will have moving forward.  Given some of what the Engineers and Artists said before... there may be some additional time spent by the engineers building out tools for the artists with the new capabilities, or to match what they did using the old tools - before they can get started.  The note here, is that there are four biomes per planet and four textures per biome (currently) - and the artists are excited that HDRP and CBT might give them more freedom, both in increasing textures per biomes, but also biomes per planet/ CB.
• They're shifting to a coordinate system called S2 (developed by Google) that is supposed to be considerably faster and lighter on storage.  By comparison, the PQS system uses Cartesian with double precision.  (Not pretending I understand all of this... but someone will).
• Tides are not going to be a thing.
• Blocky shorelines are a problem - mostly an artifact of rendering and how that scales with distance; it sounds like it will continue to be a problem, but one that can... time allowing... be fixed - it's just not a high priority at the moment.  (lot of interpretation on my part)

[K^2]

No real surprises in this talk. A lot of what they ran into are the sort of things many teams working on large, open world games have been finding, with a caveat that because the world is spherical, Intercept had to re-implement a lot of the tools from scratch. (Having worked on an in-house engine that went from canned maps to open world, I'm pretty familiar with all the rakes hidden in that tall grass, as we had to make similar kinds of tools for our teams.)

Things that was a bit unexpected for me is how often the Intercept went to more artist-tunning over more physics-based parameters for lighting and scattering. This isn't strictly good or bad, but I would have guessed Intercept going more towards physics on that. Might be good news for modders. Want some really wacky color combination for sky and sunset colors? Just change some of the atmospheric parameters to the desired colors. Nothing in the game, apparently, is going to stop you because that combination is non-physical.

[JoeSchmuckatelli]

6 hours ago, K^2 said:

artist-tunning over more physics-based parameters for lighting and scattering. This isn't strictly good or bad, but I would have guessed Intercept going more towards physics on that.

I found the 'Lighting guy's part of the talk interesting;  his initial math-based solution annoyed the artists because Moho was too bright/washed out and things around Jool too dim.  Only, later in the video he talks about the 'not having access to the HDRP tools'.

Would that have any impact on leaning towards more physics in lighting design? 

(I write this aware of their caveat that little in Nature would make Jool green, but they wanted to keep that).  Or are they likely to simply tweak things to keep the artists happy, rather than trying to play around with reverting to a more realistic model?

[Scarecrow71]

On 3/23/2023 at 7:19 PM, Strawberry said:

While the terrain is definitely a big step down in terms of performance, I think its undeniable that the terrain system is a big step up look wise from KSP1 (If you want to see it from orbit look at duna, if you want to see what the terrain system can do look at the geography of Pol). Im guessing the talk will go basically "We started with PQS because ksp1 did it and it was easy, heres all the little knick knacks we've added to PQS, here's when we ultimately realized that we couldnt get PQS as performant as we'd like with the graphics we like, here's what we're doing with CBT". 

See, I don't get what everyone else is seeing with the terrain.  Of the bodies I've landed on so far, none of them has been anything to write home about.  From space the terrain looks great.  Once landed, it's all flat and smooth and stuff.  Even the rings on Dres were just meh.

[Observe]

3 hours ago, Scarecrow71 said:

See, I don't get what everyone else is seeing with the terrain.  Of the bodies I've landed on so far, none of them has been anything to write home about.  From space the terrain looks great.  Once landed, it's all flat and smooth and stuff.  Even the rings on Dres were just meh.

I concur. I've been struggling with this myself, as I've been unable to see how KSP2 terrain is any better than the original. Sure, there are a few more low res trees here and there on Kerbin, but otherwise, personally I think KSP1 terrain is as good if not better.

[AtomicTech]

On 3/22/2023 at 8:19 PM, Dantheollie said:

something they've "been working on"

A working version of KSP 2.

[Bej Kerman]

16 minutes ago, AtomicTech said:

A working version of KSP 2.

Which we already have

[Alexoff]

It's amazing that the developers of KSP2 took the time to participate in the GDS23 just a couple of weeks after the successful release of the working version of KSP2! They probably didn't have more important things to do in the office, like fixing bugs or creating features from the roadmap.

[asmi]

40 minutes ago, Alexoff said:

It's amazing that the developers of KSP2 took the time to participate in the GDS23 just a couple of weeks after the successful release of the working version of KSP2! They probably didn't have more important things to do in the office, like fixing bugs or creating features from the roadmap.

How do you know what was/is important for them?

[Alexoff]

4 minutes ago, asmi said:

How do you know what was/is important for them?

This is easy to understand - if they spend time on this, then it is important to them. Four developers have prepared a long presentation, in my experience I can say that it takes more than one day.

[Bej Kerman]

4 minutes ago, asmi said:

How do you know what was/is important for them?

Spoiler

Queeg!

[K^2]

6 hours ago, JoeSchmuckatelli said:

Would that have any impact on leaning towards more physics in lighting design? 

Could have been a factor, but there are other ways to fake HDR.