Frame rate discussion moved from Devnotes thread.

[Azimech]

Bugtracker gives a 502. I'm going to write a lengthy feedback about the performance increase that was promised for 1.2, was there during the earliest versions of 1.2 pre, now totally lost with 1.2 final.

I'm surprised with the apathy most members of the community display with the greatest disappointment of 1.2.

[Green Baron]

Yep, performance had a huge setback with 1.2.

[Padishar]

43 minutes ago, Azimech said:

I'm surprised with the apathy most members of the community display with the greatest disappointment of 1.2.

I suspect this could be because most people haven't really noticed such a slowdown.  The performance with some of my vessels does appear to have dropped slightly since the early builds but I don't have any firm figures.  Do you have comparable figures from different pre-release builds?

2 minutes ago, Green Baron said:

Yep, performance had a huge setback with 1.2.

I've certainly not seen this, the performance for me is definitely better than 1.1.3 was, just not quite as fast as the early pre-release builds of 1.2.

[Azimech]

9 minutes ago, Padishar said:

I suspect this could be because most people haven't really noticed such a slowdown.  The performance with some of my vessels does appear to have dropped slightly since the early builds but I don't have any firm figures.  Do you have comparable figures from different pre-release builds?

I've certainly not seen this, the performance for me is definitely better than 1.1.3 was, just not quite as fast as the early pre-release builds of 1.2.

I'm going to request the community to provide me with the installation files of the first 1.2 pre, unfortunately I use steam and didn't make a backup of every build (which I should have).

Anyway, it's noticeable with high part count. https://www.reddit.com/r/KerbalSpaceProgram/comments/57h5i0/high_part_count_12_is_just_as_slow_as_113_and/

Edited October 22, 2016 by Azimech

[Green Baron]

39 minutes ago, Padishar said:

*snip*

I've certainly not seen this, the performance for me is definitely better than 1.1.3 was, just not quite as fast as the early pre-release builds of 1.2.

Hmm, now that you mentioned it i measured (though strictly it doesn't belong here).

I haven't played much 1.1.3 but 1.0.5. To be precise, fast atmospheric flight with supersonic effects below 40km causes the framerate to drop, from ~60fps (i limited it because monitor frequency) to ~30. That is certainly acceptable for a mildly complex 4 stage rocket.

When i last launched that same craft it felt like the framerate was much lower (in the 10s). Surely the ascent profile was slightly different (and i updated mods from 1.2pre to 1.2 release); but seemingly my feelings fooled me.

So, to say it loud&clear(tm), when measured the peformance is nicer than when felt :-)

The first huge station will reveal more.

 

[Azimech]

23 minutes ago, Green Baron said:

Hmm, now that you mentioned it i measured (though strictly it doesn't belong here).

I haven't played much 1.1.3 but 1.0.5. To be precise, fast atmospheric flight with supersonic effects below 40km causes the framerate to drop, from ~60fps (i limited it because monitor frequency) to ~30. That is certainly acceptable for a mildly complex 4 stage rocket.

When i last launched that same craft it felt like the framerate was much lower (in the 10s). Surely the ascent profile was slightly different (and i updated mods from 1.2pre to 1.2 release); but seemingly my feelings fooled me.

So, to say it loud&clear(tm), when measured the peformance is nicer than when felt :-)

The first huge station will reveal more.

 

Please download this craft for a proper test: https://kerbalx.com/Azimech/77I--Performance-Tester

[regex]

1 hour ago, Azimech said:

I'm going to request the community to provide me with the installation files of the first 1.2 pre, unfortunately I use steam and didn't make a backup of every build (which I should have).

I'll bet you it has something to do with code obfuscation. I didn't check this time around but during 1.1 the pre-release was not obfuscated. Code obfuscation comes with performance setbacks.

[Azimech]

Just now, regex said:

I'll bet you it has something to do with code obfuscation. I didn't check this time around but during 1.1 the pre-release was not obfuscated. Code obfuscation comes with performance setbacks.

I'm no coder, I don't know what code obfuscation means. I do remember clearly the very first 1.2 pre had one of my turboshaft helicopters running in the green while with 1.2 final it's running in yellow. Just one example.

[Guest]

2 minutes ago, Azimech said:

I'm no coder, I don't know what code obfuscation means. I do remember clearly the very first 1.2 pre had one of my turboshaft helicopters running in the green while with 1.2 final it's running in yellow. Just one example.

Have you checked your hardware?  Are you sure you have the latest drivers?

[Green Baron]

17 minutes ago, Azimech said:

Please download this craft for a proper test: https://kerbalx.com/Azimech/77I--Performance-Tester

25fps lolling on the runway. If we could wrap it in tinfoil and fill it with He it would fly :-)

 

[sarbian]

4 minutes ago, Green Baron said:

25fps lolling on the runway. If we could wrap it in tinfoil and fill it with He it would fly :-)

 

Nice, but some number with that "early 1.2 build" that was faster would help.

[Azimech]

4 minutes ago, sarbian said:

Nice, but some number with that "early 1.2 build" that was faster would help.

If you can provide me with the upgrade installation files, I shall give you the exact numbers (when I get home).

[Green Baron]

 

Build 1473: same, but the framerate shown in the mod-f12 window is fixed (no green line, no slight changes). Seemingly rates below 25fps are not shown .... ? Or am i missing something ?

Lights on and whole part it's framerate is around 30, slowly dropps to 25 and then frameratecount seemingly stops ...

Edited October 22, 2016 by Green Baron

[Azimech]

20 minutes ago, CliftonM said:

Have you checked your hardware?  Are you sure you have the latest drivers?

My hardware is as such that I had this running on the screen during 1.2 pre at almost playable levels while in 1.2 final it's absolutely unplayable. During the transition of 1.1.3 to 1.2 I didn't change hardware or drivers, I doubt there would be any need to, IMHO. Intel, AMD nor nVidia would update their drivers for such a small player as KSP. For Unity ... maybe ... for the big updates. So in my opinion, we can leave the whole hardware subject for what it is ... because this depends on code. A lot of younger KSP players have hardware at the very lower edge of the system requirements, the ones that start to experience lag at 60 parts, those are the ones who would benefit the most of these optimizations.

The first optimization run did wonders ... until Squad added a lot of features not on the initial features list. These features well ... they seem unoptimized and dragging everything down like a Siberian hunting spider with three limp legs.

5 minutes ago, Green Baron said:

 

Build 1473: same, but the framerate shown in the mod-f12 window is fixed (no green line, no slight changes). Seemingly rates below 25fps are not shown .... ? Or am i missing something ?

Lights on and whole part it's framerate is around 30, slowly dropps to 25 and then frameratecount seemingly stops ...

The internal FPS counter is not to be trusted. Most of the time I don't get a reading at all.

[Speadge]

27 minutes ago, Green Baron said:

 

Build 1473: same, but the framerate shown in the mod-f12 window is fixed (no green line, no slight changes). Seemingly rates below 25fps are not shown .... ? Or am i missing something ?

Lights on and whole part it's framerate is around 30, slowly dropps to 25 and then frameratecount seemingly stops ...

what is ur hardware alike when u are running THIS @25fps?

Edited October 22, 2016 by Speadge

[Green Baron]

I'm not sure about the 25fps. The mod-f12 seemingly doesn't show framerates below 25. It's probably lower, but not much.

i7 3.58 Ghz, gtx660, 16gb ram

Edited October 22, 2016 by Green Baron

[Padishar]

6 minutes ago, Green Baron said:

25fps lolling on the runway. If we could wrap it in tinfoil and fill it with He it would fly :-)

 

There's no way you're getting 25 fps (unless you somehow have a PC clocked at over 10GHz ).  You are presumably looking at the built in fps graph which never gives any reading below 25 fps (with the default max physics delta setting).  On my i7-4770K I get the following framerates (approximated using my MemGraph mod) when put on the runway and viewed from the side with the whole vessel visible:

1.1.3   Between 7 and 8 fps
1451   Closer to 7 than 8 fps
1484   Between 7 and 8 fps
1533   Pretty solid 8 fps
1586   Pretty solid 8 fps

So, I don't see any really significant performance change at all with your vessel though the official 1.2 1586 does look like the fastest.

With my own test vessel I see slightly more difference between the builds and build 1533 is actually slightly quicker than 1586.

[Azimech]

Just now, Green Baron said:

I'm not sure about the 25fps. The mod-f12 seemingly doesn't show framerates below 25. It's probably lower, but not much.

i7 3.5Ghz, gtx660, 16gb ram

More interesting would it be with a test between 1.1.3, 1.2 early and 1.2 final, using an external FPS counter, Delta-t ratio at 0.04 and counting both FPS and physics delay (how many real world seconds go in a KSP second). Then all the other tests: object on the runway without forces acting on it, object with SAS on roll full.

4 minutes ago, Padishar said:

There's no way you're getting 25 fps (unless you somehow have a PC clocked at over 10GHz ).  You are presumably looking at the built in fps graph which never gives any reading below 25 fps (with the default max physics delta setting).  On my i7-4770K I get the following framerates (approximated using my MemGraph mod) when put on the runway and viewed from the side with the whole vessel visible:

1.1.3   Between 7 and 8 fps
1451   Closer to 7 than 8 fps
1484   Between 7 and 8 fps
1533   Pretty solid 8 fps
1586   Pretty solid 8 fps

So, I don't see any really significant performance change at all with your vessel though the official 1.2 1586 does look like the fastest.

With my own test vessel I see slightly more difference between the builds and build 1533 is actually slightly quicker than 1586.

What about the game timer? I've read multiple reports of reasonable FPS but more lag.

[Green Baron]

Sorry guys, sunset. Must go to a pub now.

Padishar, pls read my doubts about the framerate. Unfortunately the native nvidia driver doesn't show me the actual framerate, i would think ksp is the only other process that would be able to tell if it wanted to tell. I would say it's lower than 25, but not much.

This is debian linux 8, kernel 3.16, all 8 cores @3.8Ghz.

 

Edited October 22, 2016 by Green Baron

[Padishar]

9 minutes ago, Azimech said:

What about the game timer? I've read multiple reports of reasonable FPS but more lag.

They were all virtually indistinguishable, which is exactly what I'd expect.  At 0.04, when under 25 fps, the game does exactly 2 physics updates for every rendered frame which means the clock slowdown is directly related to the framerate you are getting.  If you are only getting 5 fps then the game is doing 10 physics updates each second.  This is a total of 200 ms of in-game time resulting in a 5x slowdown (time running at 20% speed because fps is 20% of the 25 fps effective limit).

I think we need to know which build it was that you think was significantly quicker and some idea of how much quicker it was...

11 minutes ago, Green Baron said:

Padishar, pls read my doubts about the framerate. Unfortunately the native nvidia driver doesn't show me the actual framerate, i would think ksp is the only other process that would be able to tell if it wanted to tell. I would say it's lower than 25, but not much.

Try installing something like Fraps which can add a framerate overlay to any Direct3D program.  I will be very surprised if your machine is managing 10 fps, let alone "not much" slower than 25...

We should probably take this somewhere else as it isn't really on topic in this thread...

[Azimech]

31 minutes ago, Padishar said:

They were all virtually indistinguishable, which is exactly what I'd expect.  At 0.04, when under 25 fps, the game does exactly 2 physics updates for every rendered frame which means the clock slowdown is directly related to the framerate you are getting.  If you are only getting 5 fps then the game is doing 10 physics updates each second.  This is a total of 200 ms of in-game time resulting in a 5x slowdown (time running at 20% speed because fps is 20% of the 25 fps effective limit).

I think we need to know which build it was that you think was significantly quicker and some idea of how much quicker it was...

Try installing something like Fraps which can add a framerate overlay to any Direct3D program.  I will be very surprised if your machine is managing 10 fps, let alone "not much" slower than 25...

We should probably take this somewhere else as it isn't really on topic in this thread...

Maybe not on topic but ... the topic I started earlier was fused with a main feedback thread. There must be a good reason for it but I don't agree.

I do understand what you are saying. I have a lower spec machine with a pretty strong CPU core but weak GFX card. We need more test results from different people to get a complete picture.

Anyway ... if I can get a hold of those early versions I'll test them again. I've got the feeling @Majorjim! might be interested as well.

[Majorjim!]

Just now, Azimech said:

Maybe not on topic but ... the topic I started earlier was fused with a main feedback thread. There must be a good reason for it but I don't agree.

I do understand what you are saying. I have a lower spec machine with a pretty strong CPU core but weak GFX card. We need more test results from different people to get a complete picture.

Anyway ... if I can get a hold of those early versions I'll test them again. I've got the feeling @Majorjim! might be interested as well.

Yup, very interested indeed. I have noticed a decrease in performance too with 1.2, I felt it immediately on a craft I was testing a lot in 1.1.3.

 Especially during ascent and anything with a lot of fairings. Fairings( if there are a lot of them) also cause 2 second pauses in the editor every time I right click or add a part or delete one. They did not before 1.2. I know NOTHING about coding or such things so I have no idea what could be causing it but it's there so I will let more knowledgeable others look further into this.

Edited October 22, 2016 by Majorjim!

[Kerbal Kommander]

Odd, I noticed a nice jump in performance on 1.2  Overall game play is smoother, and I get good FPS with mods like scatterer (with all extras turned on), SVE, texture replacer with 8k resolution textures.

Admittedly, fairings do seem to have an impact on performance.

I don't know what my exact FPS is, but it feels closer to 60 than it did before.

Specs for reference...

Windows 7 64 bit

3.9 GHz Ivy Bridge

8 GB Ram

GTX 1060 6 GB

Samsung SSD

[Green Baron]

1 hour ago, Padishar said:

Try installing something like Fraps which can add a framerate overlay to any Direct3D program.  I will be very surprised if your machine is managing 10 fps, let alone "not much" slower than 25...

We should probably take this somewhere else as it isn't really on topic in this thread...

I'm back.

Fraps ? Direct3D ? What are you talking about ? .-)))

To cite myself:

"This is debian linux 8, kernel 3.16, all 8 cores @3.8Ghz"

To get a correct framerate, a mod that reads out the variable process timer would solve the problem.

[0111narwhalz]

2 hours ago, Azimech said:

...like a Siberian hunting spider with three limp legs.

Now that one's new to me!