KSP2 System Requirements

[Bej Kerman]

2 hours ago, Grenartia said:

Same. 1k parts is just a meme 99.99% of the time.

It's nowhere near as ridiculous as you make it out to be. Especially when large NERV-powered vehicles depend on MkI fuselage spam.

[Haustvindr]

Do not worry guys, through telemetry, I will single-handledly bring down those requirements with my trusty GTX 560ti... as long as they don't put a softlock on hardware checks that prevents me to even try (I hate those). 

I would have a better thing already but... [points everywhere], life, yeah.

I can even try to update at a later point about the experience! XD

[magnemoe]

5 hours ago, Grenartia said:

Same. 1k parts is just a meme 99.99% of the time. Like, maybe if you're attempting a Jool 5 or grand tour or putting bootprints on Eve and returning, in a single launch, I could see it, but otherwise, there's no serious need for anywhere close to 1k parts on a single craft in KSP 1, and I would be shocked if that is any different for KSP 2. 

I'm pretty sure my Minmus bases has had more than thousands part at times. as it was refueling multiple bases and tankers at once. Standard bases was a bit over 200 parts as they had ISRU life support and part production. The Minmus main base was close to 500 build in lots of parts.  The rocket to get the base to the planet was also around 200 parts as it included an reusable tug as upper stage. 

[kerbiloid]

When people make their bets on future optimization, do they presume that obviously experienced software developers were writing things from scratch in a school student manner, making stupid mistakes which make the game several times slower than it should be?

Ten nested loops with memory leaks in each?

9 hours ago, Bej Kerman said:

I just know how developers tend to prioritise things.

Yes. Noone will optimize the already sold project unless it's a complete failure.

Because he is paid for the new one.

Nobody will allow him spend his worktime on optimization of the previous one.

[K^2]

14 hours ago, mcwaffles2003 said:

Any reason in your opinion that running physics on a GPU wouldn't be worked toward? As I understand GPU's are better for working out those kinds of calculations. I get that doing so may require better hardware, but I would assume it would at least be worth looking into

Almost nobody writes custom physics solvers these days. If you're running Unity, you're either using PhysX or Havok, both of which (in Unity, at least) are running on the CPU.

And you also have to ask, why doesn't PhysX or Havok run on the GPU? And for Havok I actually have an answer from one of their engineers, back when I interviewed with them nearly a decade ago. The answer is still relevant. Games sell on visuals. Nobody wants to cut their graphics budget to increase physics performance. KSP might be one of the very few games where it'd actually make sense, but overall, there isn't a market for this sort of thing. And nVidia did actually try to make PhysX on GPU a feature, and that didn't take off for likely much the same reason. Unity never even added support for GPU acceleration in their implementation of PhysX.

Finally, if you are a studio that decides that this is absolutely worth pursuing, because your game is all about the physics, you do have GPU budget to spare, and you want to make your own solver that runs on the GPU, it's a lot of work. I know exactly what needs to be done, I've worked on every relevant component in several studios, and I don't know if I'd have the time to do it all and come up with something that results in better overall performance than just having Havok run on the CPU. And this is my niche - there's a handful of engineers in this industry more qualified, and we're all very expensive.

If Intercept seriously considered going with GPU physics, we'd see several physics engineers working on this project at the same time, and I'd recognize at least one of these names. The physics engineers who have been working, instead, have very good academic backgrounds, but very little experience in the games industry. These are the kind of people who can take Unity's physics and make it work with a space game, and I'm sure that's exactly what they've been doing.

[Leopard]

guess a chunk of this is highballing the "minimum" to avoid a lot of day one videos on youturnip showing clunky graphics - i.e. the minimum is set so its good not just "borderline playable, maybe"

pity as it utterly rules my GT730 2GB card but then thats not exactly new, nor is the rest of the PC so wasn't overly expecting to play this.. but the jump to recommended is.. quite high especially considering this was said to be playable in mid spec machines

I get this is the early version, I get its unoptimised and I get there will be a slew of diagnostic and debug code in there still, yup get all that however what this has just done is label KSP 2 as "eck that needs an expensive computer the play it" and I wonder how many who see that won't bother looking back if it drops?

[Vl3d]

28 minutes ago, K^2 said:

Almost nobody writes custom physics solvers these days. If you're running Unity, you're either using PhysX or Havok, both of which (in Unity, at least) are running on the CPU.

And you also have to ask, why doesn't PhysX or Havok run on the GPU? And for Havok I actually have an answer from one of their engineers, back when I interviewed with them nearly a decade ago. The answer is still relevant. Games sell on visuals. Nobody wants to cut their graphics budget to increase physics performance. KSP might be one of the very few games where it'd actually make sense, but overall, there isn't a market for this sort of thing. And nVidia did actually try to make PhysX on GPU a feature, and that didn't take off for likely much the same reason. Unity never even added support for GPU acceleration in their implementation of PhysX.

Finally, if you are a studio that decides that this is absolutely worth pursuing, because your game is all about the physics, you do have GPU budget to spare, and you want to make your own solver that runs on the GPU, it's a lot of work. I know exactly what needs to be done, I've worked on every relevant component in several studios, and I don't know if I'd have the time to do it all and come up with something that results in better overall performance than just having Havok run on the CPU. And this is my niche - there's a handful of engineers in this industry more qualified, and we're all very expensive.

If Intercept seriously considered going with GPU physics, we'd see several physics engineers working on this project at the same time, and I'd recognize at least one of these names. The physics engineers who have been working, instead, have very good academic backgrounds, but very little experience in the games industry. These are the kind of people who can take Unity's physics and make it work with a space game, and I'm sure that's exactly what they've been doing.

Then why are the CPU requirements so low? Even with multithreading we should expect higher loads for physics calculations. Is it possible that CPU requirements will increase as EA advanced?

[Leopard]

at a guess the way physics is handled has been rethought a bit to lighten the load on larger craft, especially if its a single thread for physics. Just thought out from the ground up could lead to a fair bit of optimisation there from the off

[magnemoe]

1 minute ago, Vl3d said:

Then why are the CPU requirements so low? Even with multithreading we should expect higher loads for physics calculations. Is it possible that CPU requirements will increase as EA advanced?

Good question. 
On the other hand I don't see physic in KSP who is mostly stresses on ship parts would run well on an GPU as its lots of dependencies. Upper stage depend on mass and air resistance from the payload and this is then transmitted down the parts. 
It would do well for lots of independent objects on a trajectory like after the rocket breaking up  But this is simpler calculations. 

[purpleivan]

17 hours ago, JoeSchmuckatelli said:

I was watching the development with interest - my posts were in support of the effort. 

But you can look at the previous 10 posts in this thread and see what I'm concerned about.  People who think 'midrange' or 'high end' is a price point - because of the baked in presumption that everyone uses a 1080p screen. 

Mid range isn't a price point - it's a performance metric.  Or at least it used to be, and of course price correlated with performance - which caused the misperception. 

3080 is super high end / way overkill for most games at 1080p... But it's putting up mid to high fps numbers at 4k.

So if you have a 1080p monitor with a computer housing a 3080 - you have a high end machine. 

If you have a 4k monitor with a computer housing a 3080... You have a mid range machine. 

Both are bloody expensive. 

I'm sorry, but the relative performance of specific hardware and that hardware's perception of being "high", "low" or otherwise, is irrelevant to a discussion of the content of performance tests (the subject that was being discussed on the other thread). We simply wanted to put together a set of standardised perf test, that anyone who could get the game to run (regardless of perceptions of their hardware's performance), could carry out and so collect a set of useful data.

[Leopard]

what would be useful is a way to see if a machine that can currently play KSP1 can play KSP2, into a few categories such as

• no, just no
• technically yes, but really no
• yes, as long as you set all the settings to minimum and don't plan on anything more than 20-30 parts and accept a low frame rate
• yes, but will be laggy with low settings

etc

[JoeSchmuckatelli]

58 minutes ago, purpleivan said:

I'm sorry, but the relative performance of specific hardware and that hardware's perception of being "high", "low" or otherwise, is irrelevant to a discussion of the content of performance tests (the subject that was being discussed on the other thread). We simply wanted to put together a set of standardised perf test, that anyone who could get the game to run (regardless of perceptions of their hardware's performance), could carry out and so collect a set of useful data.

Two players run tests they find on the interwebz. 

Player 1 is pleased as punch with her 63 FPS 

Player 2 complains the game is horribly optimized and laggy because she's only seeing 27 FPS 

Player 1 is using her brother's AlienThing Gaming xTREME running a sweet 6300m while Player 2 is stuck using her mother's old crappy work laptop by Lenotho that has a janky 6300m inside. 

 This information is useless if you don't also know that Player 1 is running native 1080p and Player 2 got a 1440p gaming monitor from Santa and has the laptop hooked up to that. 

4 minutes ago, Leopard said:

what would be useful is a way to see if a machine that can currently play KSP1 can play KSP2, into a few categories such as

• no, just no
• technically yes, but really no
• yes, as long as you set all the settings to minimum and don't plan on anything more than 20-30 parts and accept a low frame rate
• yes, but will be laggy with low settings

etc

Meh.  Some people have kept KSP going on zombie comps that should have been retired long, long ago 

Edited February 20, 2023 by JoeSchmuckatelli

[Leopard]

Just now, JoeSchmuckatelli said:

Two players run tests they find on the interwebz. 

Player 1 is pleased as punch with her 63 FPS 

Player 2 complains the game is horribly optimized and laggy because she's only seeing 27 FPS 

Player 1 is using her brother's AlienThing Gaming xTREME running a sweet 6300m while Player 2 is stuck using her mother's old crappy work laptop by Lenotho that has a janky 6300m inside. 

 This information is useless if you don't also know that Player 1 is running native 1080p and Player 2 got a 1440p gaming monitor from Santa and has the laptop hooked up to that. 

Meh.  KSP is so old that this is an unfair test

not thinking of it as a "fair" test in that way, more just managing expectations, suspect a lot who are looking to play KSP2 will be KSP1 players - sticking a KSP1 update thats basically a "will this run KSP2?" info box that can score what settings you are likely to want or just a note that "this machine may run KSP2 but we would not recommend the user experience yet"

and then update that whenever the early access requirements change

so it may well be someone with a decent machine (better than mine, mines not going to work I accept that) should probably hold off now but maybe in six months could get decent performance

its about perception management, when it was noted the game was aimed at "mid spec PCs" the comment should really have been expanded with something like "but during early access requirements will be higher" just to manage expectations.

I think the only issue here is people have been led to expect that the requirements would be lower than they are, I suspect what has been put out as the "minimum" is the sort of "minimum" that other games etc should use, i.e. "minimum to get a good experience of this" and not a "microsoft minimum" of "yes you can, but for the love of Kraken, don't"

[Piatzin]

Holy smokes, I don't even have that much free disk space in total, never mind any of the other things.

Does anyone know why it's so bulky? Is it just graphics or what? KSP was always one of those games you could play on any old laptop, assuming part count was kept low, but these requirements are pretty high-end stuff and I'm not really seeing where it could all be from.

[JoeSchmuckatelli]

Just now, Leopard said:

not thinking of it as a "fair" test

We are cross posting/editing.  'fair' wasn't what I wanted to say but I haven't had coffee, yet. 

 

Just now, Leopard said:

perception management, when it was noted the game was aimed at "mid spec PCs

That is the problem.  People assuming they had a mid spec machine or that they even knew the definition of mid-spec. 

Fact is that it is a definitionless phrase.  The meaning typically shifts over time. 

Very commonly it has been used to mean 2 Gen old mainstream card... Which I should point out does not include the 970, 1050m, etc 

[Vl3d]

18 minutes ago, Leopard said:

what would be useful is a way to see if a machine that can currently play KSP1 can play KSP2, into a few categories such as

• no, just no
• technically yes, but really no
• yes, as long as you set all the settings to minimum and don't plan on anything more than 20-30 parts and accept a low frame rate
• yes, but will be laggy with low settings

etc

Forget about KSP1 performance. The games should not be compared anymore.

I'm sure people will be able to play KSP2 with KSP1-looking graphics in the future. But what's the point? You have KSP1 for that.

Why go back to playing GTA 2 or Doom 2 after the release of GTA 3 or Quake 3 Arena? Look to the future. I'm not going back to Pentiums.

Edited February 20, 2023 by Vl3d

[darthvader15001]

I DONT CARE IF YOU RELEASE IT ON TIME JUST LOWER THE REQUIREMENTS!!

[Vl3d]

Just now, darthvader15001 said:

I DONT CARE IF YOU RELEASE IT ON TIME JUST LOWER THE REQUIREMENTS!!

UPGRADE!

[Leopard]

my point on comparing to KSP 1 is that a lot who will want KSP 2 likely are KSP 1 players so "how will this run on the same hardware?" and "will this run on the same hardware?" are reasonable questions

for me "high spec" is a gaming PC, "mid spec" is a more general purpose machine, probably 5 years old, or a gaming machine maybe 7 years old, and less is "low spec", its a moving feast as you say

[WelshSteW]

1 minute ago, darthvader15001 said:

I DONT CARE IF YOU RELEASE IT ON TIME JUST LOWER THE REQUIREMENTS!!

 

If you don't care about the release date, wait for the actual release? They've said they'll be optimising, so just don't get the EA version?

 

[Leopard]

Just now, Vl3d said:

UPGRADE!

but upgrade to what? what specs will cover the game and cover it well? currently no one really knows, what will a modded install need?

 

also "upgrade" is easy to type, but when the graphics card alone is £700 - £1,000 its slightly harder to actually do

meah, anyway for me there is no chance of playing this, PC won't be upgraded until about five years from now, so wish the good ship KSP 2 a safe journey, but I won't be aboard

maybe come five years hence I'll have actually rescued everyone in my current save

[LoSBoL]

I have KSP2 on my wishlist on steam since 2019, and knowing what a resource hog KSP was and is, I knew I needed to upgrade since 2019.

[Rpatto92]

On 2/17/2023 at 11:23 PM, Kerbart said:

Let's see, we've been promised:

• Large vessels with amazing detailing
• Better physics with thrust under timewarp
• Micrometer precision at interstellar travel
• Transparent cockpits with animated Kerbals inside
• Clouds! Water! Trees!

And everyone seriously thought this all came for free?!

Seriously, y'all need a reality check. I'm just happy my local PC Seller has hardware meeting requirements for less than $1000, not as bad as I thought it would be. But I'll let others (that means, yes, YOU who are reading this) do the dirty work of finding out what works and what doesn't before I spend a dime on that.

KSP players are no fools, they do not expect something for nothing but preposition is clearly wrong. 

Your bullet points are pretty generous. Most of those things are planned features (not yet in) and the functionality of those other features is an unknown. 

What we do know however is KSP 1 has plenty of mods that replicate many of the up-and-coming features (in some cases betters) and it doesn't have those same system requirements and runs reasonably well even when heavily modded. 

Any pc gamer can look at those system requirements, the released footage, game details, KSP1, other games specifications and reasonably conclude that those requirements are out of whack. That the game is more than likely going to have severe optimizations issues that those requirements are an obvious symptom of. 

So who are you kidding. 

[mcwaffles2003]

1 hour ago, Vl3d said:

Then why are the CPU requirements so low? Even with multithreading we should expect higher loads for physics calculations. Is it possible that CPU requirements will increase as EA advanced?

Perhaps all the optimization has been in the physics and almost none in the graphics?

[mcwaffles2003]

2 hours ago, K^2 said:

Almost nobody writes custom physics solvers these days. If you're running Unity, you're either using PhysX or Havok, both of which (in Unity, at least) are running on the CPU.

And you also have to ask, why doesn't PhysX or Havok run on the GPU? And for Havok I actually have an answer from one of their engineers, back when I interviewed with them nearly a decade ago. The answer is still relevant. Games sell on visuals. Nobody wants to cut their graphics budget to increase physics performance. KSP might be one of the very few games where it'd actually make sense, but overall, there isn't a market for this sort of thing. And nVidia did actually try to make PhysX on GPU a feature, and that didn't take off for likely much the same reason. Unity never even added support for GPU acceleration in their implementation of PhysX.

Finally, if you are a studio that decides that this is absolutely worth pursuing, because your game is all about the physics, you do have GPU budget to spare, and you want to make your own solver that runs on the GPU, it's a lot of work. I know exactly what needs to be done, I've worked on every relevant component in several studios, and I don't know if I'd have the time to do it all and come up with something that results in better overall performance than just having Havok run on the CPU. And this is my niche - there's a handful of engineers in this industry more qualified, and we're all very expensive.

If Intercept seriously considered going with GPU physics, we'd see several physics engineers working on this project at the same time, and I'd recognize at least one of these names. The physics engineers who have been working, instead, have very good academic backgrounds, but very little experience in the games industry. These are the kind of people who can take Unity's physics and make it work with a space game, and I'm sure that's exactly what they've been doing.

That genuinely makes me a bit sad. Not for this game specifically, but just as a fan of gaming. I remember being excited when PhysX cards were coming out and imagining a similar climb in the development for game physics would happen as graphics had in the early 10's and thought since nVidia bought them out that they would be implemented on GPUs similar to how ray tracing is implemented along side rasterization now. I'm guessing there's just been a great lack of need overall for such a thing though in gaming so far and that's why there would be such a need for such a specialized group to implement anything of the sort. Were it to be done, what extra potential do you think is there in such a kind of implementation over running on a CPU?