How Does KSP Work on PS4?

[RamptantFlamingo247]

Okay so some of you more "tech-savvy" KSPateers out there know that the physics calculations for KSP are sent to two cores in the CPU due to the cell architecture PC's run (32-bit and 64-bit).  My laptop at home has a dual-core 2.26GHz processing unit and the PS4 has a 1.6 GHz, 8-core processing unit, however, even though the calculations sent to the CPU are processed slower in the PS4, it can do them much quicker than my laptop due to there being more cores for it to process, or "think" with.

Recently someone jail breaked into his PS4 and installed linux, this kind of shows just how closely related new consoles and their architecture are to PC's

SO, my question is, does the restriction of only using two cores for physics calculations remain the same for consoles?  I'm very interested to know..

[Motokid600]

Me too. So far from what I've seen crafts of descent part counts do just fine. What I want to know is how it performs with much larger 300+ part crafts. Suppose well find out soon enough. 

[Deddly]

But your laptop is 2.2 ghz of what processor? You can't compare ghz to work out which one is more powerful. There was a Pentium 4 with over 4 ghz, but a moden cpu with 1.6 ghz will run rings around it, even if it's using just a single core.

Are you sure the PS4 just doesn't have a more powerful processor than your laptop?

[RamptantFlamingo247]

It's an intel i3, 3rd gen i believe.  It's old.  But a PS4 is made for "budget gaming" so it's processor wouldn't be anything super powerful.  Besides, I was just giving an example to help people understand, I know there's a lot more that goes into it but you get what I mean, I'm sure.  

From what I've read, the CPU in the PS4 can be boosted up to 2.75 GHz when the graphics card is at a low intensity, and KSP is not necessarily a graphics intensive game, just memory and CPU intensive and it does help that the PS4 has very fast memory thanks to it possessing GDDR5 source: http://www.pcper.com/reviews/Systems/PC-vs-PS4-vs-Xbox-One-Hardware-Comparison-Building-Competing-Gaming-PC

[Deddly]

It's true that KSP doesn't require much from the GPU. I also found something interesting according to Wikipedia:

"The graphics processing unit (GPU) is AMD's GPGPU-capable Radeon GCNarchitecture, consisting of 18 compute units (CUs) for a total of 1,152 cores (64 cores per CU), that produces a theoretical peak performance of 1.84 TFLOPS.[30] This processing power can be used for graphics, physics simulation, or a combination of the two, or any other tasks suited to general purpose compute."

Maybe that helps? But as I understand it, the PS4 is about three years newer than your laptop and laptops are not usually specced for gaming, wheras the PS4 is designed for gaming from the ground up. I would expect the newer technology in the PS4 is the main reason for the performance increase you see (unless your laptop has integrated graphics, in which case that could be a bottleneck for you)

[Temeter]

4 hours ago, Deddly said:

It's true that KSP doesn't require much from the GPU. I also found something interesting according to Wikipedia:

"The graphics processing unit (GPU) is AMD's GPGPU-capable Radeon GCNarchitecture, consisting of 18 compute units (CUs) for a total of 1,152 cores (64 cores per CU), that produces a theoretical peak performance of 1.84 TFLOPS.[30] This processing power can be used for graphics, physics simulation, or a combination of the two, or any other tasks suited to general purpose compute."

Maybe that helps? But as I understand it, the PS4 is about three years newer than your laptop and laptops are not usually specced for gaming, wheras the PS4 is designed for gaming from the ground up. I would expect the newer technology in the PS4 is the main reason for the performance increase you see (unless your laptop has integrated graphics, in which case that could be a bottleneck for you)

Won't help at all. KSP does not have multicore support for single, large ships (which are 90% of the CPU demand of KSP), these can only run on a single CPU core. So that outrules running anything on those 18 small cores (which aren't even good at this kind of physics).

PS4 will most likely face the same issues. They've got 6 cores for games (2 are reserved for some stupid reason), which are individually super weak, and won't be able to share much of the workload.

 

Btw, just saying: The PS4's CPU is something made for tablets and notebooks. It's that bad.

[Deddly]

2 hours ago, Temeter said:

Won't help at all.

...

They've got 6 cores for games (2 are reserved for some stupid reason), which are individually super weak, and won't be able to share much of the workload.

Btw, just saying: The PS4's CPU is something made for tablets and notebooks. It's that bad.

And yet @RamptantFlamingo247 and others report that performance is pretty good. So where's the performance coming from?

[RamptantFlamingo247]

My only idea is that they changed the engine to be able to run on more than just two cores or maybe they're using the RAM and the GPU to help it.  I'm really not sure though.  I've never had any bad problems with it yet and I'm making ships with around 70 parts rn and I've never seen it dip below 30 fps, or a noticeable dip at least unless it's engaging the physics engine of course

[Majorjim!]

Just now, Deddly said:

And yet @RamptantFlamingo247 and others report that performance is pretty good.

What does that mean? If you are talking about KSP performance you need to quote part counts and FPS or timer status otherwise it's meaningless.

Just now, RamptantFlamingo247 said:

My only idea is that they changed the engine to be able to run on more than just two cores or maybe they're using the RAM and the GPU to help it.  I'm really not sure though.  I've never had any bad problems with it yet and I'm making ships with around 70 parts rn and I've never seen it dip below 30 fps, or a noticeable dip at least unless it's engaging the physics engine of course

Have you tried 100 parts? Is the game locked to 30FPS?

Edited July 16, 2016 by Majorjim

[RamptantFlamingo247]

17 minutes ago, Majorjim said:

What does that mean? If you are talking about KSP performance you need to quote part counts and FPS or timer status otherwise it's meaningless.

Have you tried 100 parts? Is the game locked to 30FPS?

I think with smaller ships I've noticed it run at 60 FPS but I haven't tried 100 ships yet (that I know of at least, we'll see how my kerbin system orbiter pans out in career mode) 

and let's try to stay nice please   we all love KSP, no hate  

[Deddly]

5 minutes ago, RamptantFlamingo247 said:

we all love KSP, no hate  

Don't worry, there's no hate coming from any of us replying so far, and I don't think you'll see any hate coming from @Majorjim at all. He makes a fair point - performance is subjective and depends a lot on part count. There are some stock planes with over 100 parts. How about trying one out in a sanbox game and telling us how it runs.

One other thing we could talk about that greatly affects performance is resolution. What screen resolution does the PS4 run at?

[Majorjim!]

Just now, RamptantFlamingo247 said:

I think with smaller ships I've noticed it run at 60 FPS but I haven't tried 100 ships yet (that I know of at least, we'll see how my kerbin system orbiter pans out in career mode) 

and let's try to stay nice please   we all love KSP, no hate  

I'm not hating,  just stating a helpful fact that if you want to compare performance you need numbers. How do you see FPS on the console?

Edited July 16, 2016 by Majorjim

[RamptantFlamingo247]

Ok good!  Just making sure  

well a lot of intensive PS4 games are running at 720-900 so I'd say around there.  I wouldn't be surprised if it was 1080 but I wouldn't mind 900 if it gave me better frames  

[Temeter]

Unity 5 performas a lot better, so I wouldn't expect problems with very small starter ships. What if you go 100, 200, 300?

[soulsource]

9 hours ago, Deddly said:

It's true that KSP doesn't require much from the GPU. I also found something interesting according to Wikipedia:

"The graphics processing unit (GPU) is AMD's GPGPU-capable Radeon GCNarchitecture, consisting of 18 compute units (CUs) for a total of 1,152 cores (64 cores per CU), that produces a theoretical peak performance of 1.84 TFLOPS.[30] This processing power can be used for graphics, physics simulation, or a combination of the two, or any other tasks suited to general purpose compute."

Maybe that helps? But as I understand it, the PS4 is about three years newer than your laptop and laptops are not usually specced for gaming, wheras the PS4 is designed for gaming from the ground up. I would expect the newer technology in the PS4 is the main reason for the performance increase you see (unless your laptop has integrated graphics, in which case that could be a bottleneck for you)

While meanwhile one can in principle calculate everything on the graphics card, there are two issues here:
First, KSP is built using Unity, which uses nVidia PhysX as physics engine, which relies on nVidia CUDA, which is nVidia only (as PhysX is closed source, not even the Boltzmann Initiative can help here...).
Second, and that's much worse, the way KSP is simulating a single vessel is not well suited for parallel computation (that's why there's also only one core per vessel on the PC).

I wouldn't expect KSP to run better on the consoles than it does on a PC with comparable specifications, but I hardly can wait to test it myself as soon as it gets released in Europe.

[RamptantFlamingo247]

Hmm ok give me a list of what you guys want me to test in KSP and I'll see what I can get done today/tomorrow!

[Deddly]

1 hour ago, RamptantFlamingo247 said:

Hmm ok give me a list of what you guys want me to test in KSP and I'll see what I can get done today/tomorrow!

Great!

Test 1: A single 200 part craft. Any slowdown? Improvment if looking up so that the planet and ocean is not visible?

Test : 2 x 100 part crafts close to each other. Results compared to test 1? 

Does the framerate stay steady at a cost to slow time simulation (yellow mission timer) or is it balanced?

[RamptantFlamingo247]

1 hour ago, Deddly said:

Great!

Test 1: A single 200 part craft. Any slowdown? Improvment if looking up so that the planet and ocean is not visible?

Test : 2 x 100 part crafts close to each other. Results compared to test 1? 

Does the framerate stay steady at a cost to slow time simulation (yellow mission timer) or is it balanced?

Test 1: I'll get this done later tonight (hopefully) 

Test 2: Who is a skilled kerbonaught capable of bringing two ships close together in orbit?  I am not one of these...  We need your help!

[RamptantFlamingo247]

Test 1: I got 198 problems but frame rate ain't one son 

[klesh]

2 hours ago, RamptantFlamingo247 said:

Test 2: Who is a skilled kerbonaught capable of bringing two ships close together in orbit?  I am not one of these...  We need your help!

 

Make a 100 part plane, launch it and just taxi off the runway.  Go back to Space center, Spaceplane Hangar, and launch a second one and taxi alongside the first. 

 

Rendezvous comes later.

[Deddly]

8 hours ago, klesh said:

 

Make a 100 part plane, launch it and just taxi off the runway.  Go back to Space center, Spaceplane Hangar, and launch a second one and taxi alongside the first. 

 

Rendezvous comes later.

Yes this is what I had in mind. A plane or a rover and just drive it out of the way.

But it's interesting to hear the results of test 1, good job!

[Majorjim!]

11 hours ago, RamptantFlamingo247 said:

Test 1: I got 198 problems but frame rate ain't one son 

Urge to use sarcasm is great..

 Dude, numbers please other than part count. So 198 parts, where though and what was the framerate or timer status, pics would be nice if possible.

[RamptantFlamingo247]

5 hours ago, Majorjim said:

Urge to use sarcasm is great..

 Dude, numbers please other than part count. So 198 parts, where though and what was the framerate or timer status, pics would be nice if possible.

I used the large cubic girders connected to an SRB with a probe as the core. Frame rate was around 15-20. No difference in frame in flight vs on the ground

Test 2:

I brought two 88 part planes within wing tip touching distance.  There was a little frame loss, went down between 10-15 frames

[klesh]

What color was the timer in the upper left in each instance?

[Majorjim!]

Just now, RamptantFlamingo247 said:

I brought two 88 part planes within wing tip touching distance.  There was a little frame loss, went down between 10-15 frames

Odd.. Multiple craft should result is a large performance boost. So you got 15-20 FPS with a single 198 part count craft but a much lower frame rate of 10-15 FPS with two 88 part craft, a total of 176 parts? That seems to indicate that the PS4 version is not mulit-threaded for multiple craft.

 And even if so that is a super odd result and 10-15 FPS with even 200 parts is super poor performance considering the mega boost we got moving to Unity 5..