I just heard of KSP 2. Are they officially using Unity?

[magnemoe]

On 9/21/2019 at 11:38 AM, Incarnation of Chaos said:

My prediction for what it's worth is that CNT chips will end up holding an advantage in clockspeed initally; but Laser Transistors will win out because not only do they have a density advantage (Apperently) but even more important since they use light instead of an electrical potential to switch they don't have the same issues with electromigration, ablation and others that Si and CNT transistors would (CNT wouldn't likely have migration issues, ablation and the low voltages needed to run them would be the primary ones)

do you know how to make an nano scale optical transistor?
Optical memory sounds hard. 
On the other hand an optical router sounds pretty plausible and would be very useful as the signal is optical. 
Same with DPS and other cases there you have lots of data but not much complex handling of them. 
With silicon you option is to go parallel but going faster works even better 

Edited September 22, 2019 by magnemoe

[Incarnation of Chaos]

1 hour ago, magnemoe said:

do you know how to make an nano scale optical transistor?
Optical memory sounds hard. 
On the other hand an optical router sounds pretty plausible and would be very useful as the signal is optical. 
Same with DPS and other cases there you have lots of data but not much complex handling of them. 
With silicon you option is to go parallel but going faster works even better 

They've already made CNT transistors; optical memory i didn't even think about. Also an "optical router" is already a thing; it's required for any fiber installations to convert optical signal to ethernet. 

[GoldForest]

4 hours ago, magnemoe said:

Optical memory sounds hard. 

 

2 hours ago, Incarnation of Chaos said:

They've already made CNT transistors; optical memory i didn't even think about.

 

Optical Ram would be extremely hard. They would have to find a way to continuously keep the light signal alive. 

No, I don't think there will ever be optical Ram. At least none small enough to fit in a regular computer.

Optical Storage on the other hand, maybe. A laser would etch or burn parts of a crystal disk with a pattern. Then a reader laser would read the pattern. Only problem is the crystal disk once etched can't be etched again, as etching over the current pattern would leave 2 patterns in one place, and I don't think the laser could erase the previous pattern. So basically the harddrive would become read-only once its capacity was full. On the other hand, if the laser could erase the pattern, it might harm the crystal leaving minute damage that will accumulate over time leading to hard drive failure.

Anyway I feel regular transistor RAM will be used with optical hardware, which would slow it down a lot. Although, optical hardware might be fast enough that RAM might not be needed. The optical CPU might be able to send everything out fast enough without RAM, but that's just theory. If we really need Ram, CBT ram would probably be best as CBT looks to be faster than regular transistors. 

[Incarnation of Chaos]

2 hours ago, GoldForest said:

 

 

Optical Ram would be extremely hard. They would have to find a way to continuously keep the light signal alive. 

No, I don't think there will ever be optical Ram. At least none small enough to fit in a regular computer.

Optical Storage on the other hand, maybe. A laser would etch or burn parts of a crystal disk with a pattern. Then a reader laser would read the pattern. Only problem is the crystal disk once etched can't be etched again, as etching over the current pattern would leave 2 patterns in one place, and I don't think the laser could erase the previous pattern. So basically the harddrive would become read-only once its capacity was full. On the other hand, if the laser could erase the pattern, it might harm the crystal leaving minute damage that will accumulate over time leading to hard drive failure.

Anyway I feel regular transistor RAM will be used with optical hardware, which would slow it down a lot. Although, optical hardware might be fast enough that RAM might not be needed. The optical CPU might be able to send everything out fast enough without RAM, but that's just theory. If we really need Ram, CBT ram would probably be best as CBT looks to be faster than regular transistors. 

The issue with that is it limits the amount of storage to wavelength; which is already seeing diminishing returns on bluerays. The rewritable issue could be solved with phase-change materials like used in CD-RW's; but optical formats are already a step back from current magnetic and solid state storage. Also speed isn't the primary reason we use RAM; it's capacity. CPU cashe is measured in Mbs; RAM in the GB's. So unless the increase in density is allowing gigabytes of L2 cashe then we're still going to have RAM in some form.

[GoldForest]

20 minutes ago, Incarnation of Chaos said:

The issue with that is it limits the amount of storage to wavelength; which is already seeing diminishing returns on bluerays. The rewritable issue could be solved with phase-change materials like used in CD-RW's; but optical formats are already a step back from current magnetic and solid state storage. Also speed isn't the primary reason we use RAM; it's capacity. CPU cashe is measured in Mbs; RAM in the GB's. So unless the increase in density is allowing gigabytes of L2 cashe then we're still going to have RAM in some form.

Right, but speed is also a factor. The faster the cpu can access the ram, the faster the operation can be completed.

As far as L2 and L3 cache, I forgot about that, whoops. Still, light based cache would have the same thing as light based ram, keeping the signal alive. So optical cpus might rely on transistor based RAM and cache more heavily than transistor cpus.

The only way I see ram being able to be optical is if it had a "light tunnel" that looped. Basically a fiber optic cable linked to a signal repeater. But that would make the RAM big and bulky.

[Incarnation of Chaos]

1 minute ago, GoldForest said:

Right, but speed is also a factor. The faster the cpu can access the ram, the faster the operation can be completed.

As far as L2 and L3 cache, I forgot about that, whoops. Still, light based cache would have the same thing as light based ram, keeping the signal alive. So optical cpus might rely on transistor based RAM and cache more heavily than transistor cpus.

The only way I see ram being able to be optical is if it had a "light tunnel" that looped. Basically a fiber optic cable linked to a signal repeater. But that would make the RAM big and bulky.

True; which may mean you could use optical interconnects and just have some form of active conversion between the electrical signals coming from RAM and the optical from CPU for decent speed gains.

[Gargamel]

We'd be more than happy to split this off into it's own thread if you guys would like.

[K33N]

No offense but the people who complain about Unity as an engine are computer / programming illiterate. Its so bizarre to see people have such a strong, bonkers opinion on a game engine that so wrong. 

[Vanamonde]

5 hours ago, K33N said:

No offense but the people who complain about Unity as an engine are computer / programming illiterate. Its so bizarre to see people have such a strong, bonkers opinion on a game engine that so wrong. 

That's a rather harsh way to put it. Care to state a case for your position? 

[Nuke]

i dont really have any problem with unity. especially as it made ksp1 work just fine. i don't think any other engine could have pulled it off. 

though i would like it to go back to the old ways where game devs wrote their own engines which can better fit their requirements so games aren't made out of bodges and workarounds. and usually results in a lot of very smart people on the dev team that can solve unforeseen problems. of course nobody does this anymore (unless they intend to sell the engine to other devs with the base game essentially being a tech demo) so im not going to freak out ksp2 doesn't do this. 

[magnemoe]

4 hours ago, Nuke said:

i dont really have any problem with unity. especially as it made ksp1 work just fine. i don't think any other engine could have pulled it off. 

though i would like it to go back to the old ways where game devs wrote their own engines which can better fit their requirements so games aren't made out of bodges and workarounds. and usually results in a lot of very smart people on the dev team that can solve unforeseen problems. of course nobody does this anymore (unless they intend to sell the engine to other devs with the base game essentially being a tech demo) so im not going to freak out ksp2 doesn't do this. 

The problem is that this takes lots of time. 
it was not much of an problem back in the old days but today you have to get everything from lighting to physic pretty right. And the dev tool also let you set up stuff as fast as making Skyrim mods. 
 

[Radar]

5 hours ago, Nuke said:

i dont really have any problem with unity. especially as it made ksp1 work just fine. i don't think any other engine could have pulled it off. 

though i would like it to go back to the old ways where game devs wrote their own engines which can better fit their requirements so games aren't made out of bodges and workarounds. and usually results in a lot of very smart people on the dev team that can solve unforeseen problems. of course nobody does this anymore (unless they intend to sell the engine to other devs with the base game essentially being a tech demo) so im not going to freak out ksp2 doesn't do this. 

Time as @magnemoe said but also money.. Nowadays it's all about using off-the-shelf products to save both time and money. Why reinvent the wheel when other companies (Unity, Unreal, Cryengine the list goes on) can essentially give you what you want. They all have their advantages and disadvantages like one could provide better visuals with less cpu/gpu  usage while others will give you better physics formulation. 

Like I said, why reinvent the wheel when you can do what you want with the tools already available off the shelf and do it in less time, less money not to mention employing staff more easily.

Can you imagine a job ad from one firm saying 'must have experience in Unity' over another that said 'must have experience in KSP engine (an in-house developed gaming engine)'. Unless the staff that developed the engine stay around, the costs of retraining new staff far out ways using what's already around and can essentially start from day 1.

[DStaal]

Also worth noting is that, sure, a hand-optimized full stack solution designed just for a particular game is going to be better...

*If* you're an expert in coding realistic physics, and lighting, and AI, and progressive loading, and memory management, and every one of the dozen other things the game engine has tools for.

Yes, you could in theory write a solution that's better for your specific case - but most likely you'll mess up, vs. the publicly tested solution that has been optimized over time and exposure from dozens of games.  The game engine company can afford to hire experts in each of those as they know it'll help their game engine sell, and they'll be able to recoup the costs over many games - where an individual game likely can't hire experts in everything.

Realistically, the game is likely to perform better if you're using an established game engine that suits your game.  In theory could *could* do better doing it all yourself - but you'd *probably* end up spending more money and doing a worse job.

[Guest]

9 hours ago, Nuke said:

though i would like it to go back to the old ways where game devs wrote their own engines which can better fit their requirements so games aren't made out of bodges and workarounds. 

That would raise the cost of development by an order of magnitude or more. Comparable software exists in other domains. Licenses typically cost $1000 to $10,000 or so. Would you be willing to pay that much for a game?

Games in the old days were much simpler beasts than they are now. You could go that way too, to keep the cost reasonable. Do you think KSP would work as a 320x240 resolution 2D sprite game with only Kerbin and the Mun, saving only at KSC, and only one vessel in flight at a time?

Edited September 27, 2019 by Guest

[Incarnation of Chaos]

23 minutes ago, Brikoleur said:

That would raise the cost of development by an order of magnitude or more. Comparable software exists in other domains. Licenses typically cost $1000 to $10,000 or so. Would you be willing to pay that much for a game?

Games in the old days were much simpler beasts than they are now. You could go that way too, to keep the cost reasonable. Do you think KSP would work as a 320x240 resolution 2D sprite game with only Kerbin and the Mun, saving only at KSC, and only one vessel in flight at a time?

I'm honestly not so sure of this myself; building an engine is an expensive project without a doubt requiring experienced staff. But it's a one time cost; spread out over the entire time you're selling games. Even if upgrades are needed later; they're normally done along a normal development cycle which blunts the cost signifigantly.

4 hours ago, Radar said:

Can you imagine a job ad from one firm saying 'must have experience in Unity' over another that said 'must have experience in KSP engine (an in-house developed gaming engine)'. Unless the staff that developed the engine stay around, the costs of retraining new staff far out ways using what's already around and can essentially start from day 1.

Personally i think this is the biggest reason; training is always an expense and having your own custom stack makes it even more expensive since now you have to spend months training crew before they can actually start producing. Someone that knows unity needs far less before they can start being productive; unlikely day 1 but far closer.

[Curveball Anders]

It would be interesting to know if the people who have so strong opinions about the pros and cons of Unity vs engine-X vs home brew actually have participated in creating a game using any of those.

Because there's a lot of opinions floating around the Interwebs but much of it appears to be based on the opinions of others who are based on the opinions of others and very little that is based on first hand experiences.

 

 

[runner78]

Regarding quantity of new games, Unity is the market leader (50%, according to Unity, worldwide an all platforms) So its natural that a large amount of this game are bad and poorly developed, and that pushes the image of Unity down. So Unity work on the promise "performance by default" with DOTS (Data-Oriented Technology Stack)

But don't underestimate bad developers, they can also create bad games with the best engine. 

Edited September 27, 2019 by runner78

[K33N]

On 9/25/2019 at 7:05 PM, Vanamonde said:

That's a rather harsh way to put it. Care to state a case for your position? 

A game engine is a living piece of code architecture that is constantly changing; modern Unity is a game engine of 2019.  Most of the people that are upset think its an old engine, while having no knowledge of programming or engine architecture. 

[birdog357]

On 9/14/2019 at 11:07 PM, ronson49 said:

On 9/14/2019 at 10:15 PM, GoldForest said:

Unity works really well when OPTIMIZED! 

Hit me with a 60FPS 1080 AAA Unity game. 

Battlestar Galactica Deadlock.

[ronson49]

5 hours ago, birdog357 said:

Battlestar Galactica Deadlock.

 

It is just more of the same tired 2d/2.5d/3dplane (boardgame style).    The bulk of this discussion is pointing out all these features but nothing exists beyond the pen and paper stage.  

 

And for all the people saying that Unity is somehow Assembly,  that, the better you are, the better it is.   Well not really.   You hit the ceiling of the framework and the framework has to make assumptions so that my grandmother can make a game as well.  

 

[Guest]

30 minutes ago, ronson49 said:

You hit the ceiling of the framework and the framework has to make assumptions so that my grandmother can make a game as well.

What ceiling, specifically? 

What assumptions, specifically?

[ronson49]

I am thinking (and please correct me! I am interested).   You can see Unity like Java.    In the sense that, Unity has a runtime executable written in C++ and then,  developers author their work in C#.     Effectively what happens now is that the unity runtime is put into the final executable first,  and then,  the authored code is compiled on a separate layer.   Meaning that your final code is still C# managed output running on the thread, CALLING INTO the C++ runtimes?

 

 

 

[runner78]

2 hours ago, ronson49 said:

I am thinking (and please correct me! I am interested).   You can see Unity like Java.    In the sense that, Unity has a runtime executable written in C++ and then,  developers author their work in C#.     Effectively what happens now is that the unity runtime is put into the final executable first,  and then,  the authored code is compiled on a separate layer.   Meaning that your final code is still C# managed output running on the thread, CALLING INTO the C++ runtimes?

Yes and no, 

• Unity has IL2CPP, that converts IL-code to C++. Was designed for devices where no JIT is allowed. It exists also as option for Desktop, but Mod-support will be difficult.
• Unity's Burst compiler create machine code AOT (JIT in development) for C# Jobs.

[Incarnation of Chaos]

22 hours ago, ronson49 said:

And for all the people saying that Unity is somehow Assembly,  that, the better you are, the better it is.   Well not really.   You hit the ceiling of the framework and the framework has to make assumptions so that my grandmother can make a game as well.  

 

Even if this is true; unity exposes API's to allow you to extend on it's base functionallity (The application built on unity can also expose API's also). You could even code the entire game as a mod it's self; then you can pretty much do whatever you want. It's more about structure than anything; you have to build the game from the ground up to support mods or to run well. 

[Cloakedwand72]

On 9/14/2019 at 9:29 AM, GluttonyReaper said:

Yes, they are, it's been officially confirmed.

EDIT: It's been confirmed in a few places, for example the FAQ in this thread: https://forum.kerbalspaceprogram.com/index.php?/topic/187315-kerbal-space-program-2-master-post/

Does that mean I could run it Just fine on my current PC?