Which version performed best for you?

[Majorjim!]

The issue here is there are lots of new players.

Also, it is FACT that versions prior to 1.0+ performed better.

- - - Updated - - -

To be brutally honest, I've only really started building big things in 1.0, but my computer isn't that great at running KSP so I never get very good results. That's just me though - I'm often envious of the people who can build huge interplanetary vessels with very little performance drop.

Did you vote?

[xtoro]

Thanks to everyone for contributing. Seems by the comments that the majority has seen better performance pre-1.0. But the votes show otherwise, and as others suspect, it's due to new players who haven't played older versions...

[SubzeroSpartan7]

I think the problems with 1.x and its performance drops are all the new calculations the game has to do with heat and aerodynamics compared to the previous versions.

[TheMoonRover]

I remember it lagging with fairly small ships in older versions. It improved significantly in 0.23 or thereabouts, although that was around the time I got this laptop with Windows 8, as opposed to a 4 or 5 year-old Vista. Since then, the performance has gradually worsened in each version, making 0.90 irritatingly slow at times. There was a noticeable improvement in 1.0 but 0.23.5 still seems to perform slightly better for me. The 0.18.3 demo feels extremely laggy now, although perhaps that's because I'm used to it running better so it seems worse than it is. So 0.23.5 is the best for me, but if Unity 5 makes even the slightest difference (and everyone seems to think it'll make a huge difference) then 1.1 will be the best.

[Streetwind]

I don't generally build ships large enough to cause significant performance degradation. I think my largest one went up to 174 if I recall correctly. Made the timekeeping widget go yellow, and the launch took a bit longer than usual, but it got to orbit just fine (and had lost a quarter of its parts by then). I think that was in .90. Not sure.

But yeah, I'm probably not qualified to vote on this poll, so I'll pass

[Gojira]

Why not include every version since release of the paid version of KSP? 0.14, 0.15, etc.. It's obvious they didn't run better, but why not include them in the polls as well?

To add to the thread, I'd have to say that, contrary to what everyone else is saying, 1.0.4 is running best for me. I'm having an easier time looking at the terrain than I was in 0.90.

[xtoro]

I think the problems with 1.x and its performance drops are all the new calculations the game has to do with heat and aerodynamics compared to the previous versions.

But previous versions running DRE and FAR ran smoother than stock 1.0.4... And they did heat and aerodynamics. But I wonder now if they ran on a different thread than physics...

I don't generally build ships large enough to cause significant performance degradation. I think my largest one went up to 174 if I recall correctly. Made the timekeeping widget go yellow, and the launch took a bit longer than usual, but it got to orbit just fine (and had lost a quarter of its parts by then). I think that was in .90. Not sure.

But yeah, I'm probably not qualified to vote on this poll, so I'll pass

No worries about not voting and thanks for commenting anyways

Why not include every version since release of the paid version of KSP? 0.14, 0.15, etc.. It's obvious they didn't run better, but why not include them in the polls as well?

To add to the thread, I'd have to say that, contrary to what everyone else is saying, 1.0.4 is running best for me. I'm having an easier time looking at the terrain than I was in 0.90.

I've been playing since the early teen versions, but i found it better to not include them since so few people on the forums played them.

[sam.johnson841]

0.90, 1.0.x has slowed down. I suspect it's due to the new aero & heat calcs. Though i don't remember FAR making it as slow (in 0.90) as it can be now(1.0.x).

Hopefully Unity 5 will do something and then the devs could perhaps find some more time for general optimisation, though it sounds as if they are doing quite a bit for 1.1

- Sam

[Robotengineer]

.24, maybe .25 (those are kind of blurred in my mind) I could have numerous mods without having to compromise much, or maybe I was just easier to please back then.

[Alshain]

You didn't get a performance drop with 1.x? I found that building the same ships I had in 0.25 yielded much more lag. Don't really recall 0.90 because it went by quick for me in my mind for some reason.

Nope. None at all.

[mikegarrison]

You guys all seem to be ignoring that so much else is going on now. More aero and heating, in particular. When the code has to do more physics, it can't do it as quickly.

There is more to "performance" than just lag time on large part counts.

[Alshain]

You guys all seem to be ignoring that so much else is going on now. More aero and heating, in particular. When the code has to do more physics, it can't do it as quickly.

There is more to "performance" than just lag time on large part counts.

I think the biggest issue is perception. You can sit here on the forums and say "It's FACT" as an earlier person did but it's opinion, not fact. There is no adequate measurement, so it can't be proved. I've notice people always seem to think the newest is the worst. A lot of people are saying 0.25 was the best but a year ago I remember everyone complaining how bad 0.25 was. Same for 0.90, it was terrible, until 1.0 came out and then it was great!

That is why polls like this are completely meaningless. I'm sure some of the people voting here are using mods and different combinations of mods in 1.0 than they did back then in 0.25 and 0.90 as well. Yet they fail to think that the mods might be adding to their performance issues, nope it has to be KSP, right?

[SciMan]

Well, the new aerodynamics and thermodynamics systems seem to have had a noticable negative impact on performance.

In v0.25, I would start having trouble with ships of 275-300 parts (heavily modded game).

Now, I'm having trouble with ships of ~100 parts. (running most of the same mods as my v0.25 install).

Of course, the KSP resource system and thermodynamics calculations aren't allowed to use a different core than the physics, which I think is where most of the problems come from.

Same processor core having to do more work for the same number of parts = slower game.

If at least the resource system could be split off and allowed to run on a different core, that would help quite a bit I think.

Also the resource system is doing calculations more than once per physics frame, per resource, per part that generates, consumes, stores, or converts resources. It REALLY shouldn't need to run more than once per frame. On the other hand, it's always done that ever since rocket fuel was different than jet fuel (way back, the game used to use generic "fuel" instead of LFO, and jet engines didn't need intakes, but couldn't go above ~700m/s).

Edited September 23, 2015 by SciMan

[mikegarrison]

On the other hand, it's always done that ever since rocket fuel was different than jet fuel (way back, the game used to use generic "fuel" instead of LFO, and jet engines didn't need intakes, but couldn't go above ~700m/s).

Rocket fuel is still the same as jet fuel. Do you mean that back in the day, rockets didn't need a separate resource for oxidizer?

[Red Iron Crown]

Rocket fuel is still the same as jet fuel. Do you mean that back in the day, rockets didn't need a separate resource for oxidizer?

That's how it used to be, yes.

[katateochi]

0.90 by far.

I could launch an 800 part craft and record, with a smooth framerate, no time slowing.

Version 1+ has taken a massive nosedive in performance. Fingers crossed for unity 5.

This.

I voted for both 0.24 and 0.25 (can't quite remember which was the best for me), but the recent versions performance has really upset me (to the point I've not touched KSP in quite a while). Like Majorjim I used to build things that easily exceeded 600 parts, 800 was pretty common too and performance (in terms of FPS, not crashes) was much better. Now 300 parts seems to be the limit for me (on the same rig) and that's just frustratingly small.

The later version have been much more stable, but tbh I'd rather have the game crash more often and have better FPS while playing.

[Justicier]

For the most part I build space planes, and from that perspective, either .25 or .90 was the best performing. Every version post 1.0 has been a significant performance decrease... I can now only build the simplest (< 100 part) space planes without the framerate taking a nose dive. I am uncertain if the new aero model or heat model is the culprit, or if they both work well together as a team for framerate destruction. Once out of atmosphere there is a definite performance increase, but building aircraft and SSTO's has now become less of a joy and more of a frustration.

[HafCoJoe]

I'm having to agree with you MajorJim and Katatiochi... I'm actually pulling worse FPS in post 1.0 on my (far) superior rig in comparison to pre 1.0 on my potato PC. You can tell the difference in my Project Inquiry video. That rocket was around 600 parts and it recorded pretty well on my old computer. With the newest updates I can't do nearly that well. My guess is that the addition of heating and the new aero are eating up processing space. - The addition of multi-core threading will hopefully alleviate this... *knocks on wood*

Edited September 23, 2015 by Avera9eJoe

[cantab]

I'm not 100% sure, but I voted 0.90. I certainly don't remember it having objectionable performance, and it added nice features over 0.25.

1.0.x as many have said is a performance disaster, which is one of the main reasons I rate it as the worst release since I bought the game on 0.23. It's doubly frustrating because I blame the heating and aero. I use FAR so the aero is irrelevant, and the heating system is so fundamentally broken I just go around with the heatproof cheat on, meaning my processor is doing totally, utterly, useless work when it could be doing the stuff that matters.

1.0.x holds the dubious distinction of hosting my laggiest ever launch. 1100 parts on the pad, 700 in orbit, two hours to get there in a near-stock install. That's not my launchpad part count record either, though it may be the record on an orbital ship.

EDIT: The pessimistic side of me says that unless Squad do adequate optimization, 1.1 might only perform as well as 0.90. The improvements brought by Unity 5.2 and PhysX 3.3 could easily be eaten up by the CPU-hogging code from 1.0.x.

Edited September 23, 2015 by cantab

[xtoro]

I'm going to run some tests.

0.25 vs 0.90 vs 1.0.4 all stock.

Then the same tests again but with DRE and FAR installed on the pre-1.0 versions.

My PC is a beast, so we'll see what the differences are.

[Majorjim!]

I'm not 100% sure, but I voted 0.90. I certainly don't remember it having objectionable performance, and it added nice features over 0.25.

1.0.x as many have said is a performance disaster, which is one of the main reasons I rate it as the worst release since I bought the game on 0.23. It's doubly frustrating because I blame the heating and aero. I use FAR so the aero is irrelevant, and the heating system is so fundamentally broken I just go around with the heatproof cheat on, meaning my processor is doing totally, utterly, useless work when it could be doing the stuff that matters.

1.0.x holds the dubious distinction of hosting my laggiest ever launch. 1100 parts on the pad, 700 in orbit, two hours to get there in a near-stock install. That's not my launchpad part count record either, though it may be the record on an orbital ship.

EDIT: The pessimistic side of me says that unless Squad do adequate optimization, 1.1 might only perform as well as 0.90. The improvements brought by Unity 5.2 and PhysX 3.3 could easily be eaten up by the CPU-hogging code from 1.0.x.

Given the current state of performance having something like it was in 0.90 would be fantastic.

[xtoro]

So I don't have time this evening but I started doing a bit.

I'm using DMagic's 600-part benchmark rocket for this, and all future tests. I'll also be downloading an FPS counter. For now, I use the "physics realtime ratio" displayed by going into the F12 menu and clicking on "show flight debug stats".

What I found so far (haven't downloaded DRE or FAR yet):

KSP 0.25 completely stock, new install, all graphics settings on max.

From VAB to launch, smooth as silk.

Physics ratio between 0.7-0.8x and timer shows steady yellow at launch.

Launch is smooth and ratio is between 0.6-0.8x.

After dropping a few stages, ratio is 0.9-1x and timer is green.

The rest of the flight is same high FPS, 1x ratio, and smooth.

KSP 1.0.4 completely stock, new install, all graphics settings on max.

VAB is smooth and launch is smooth until physics loads (the point when the camera drifts down to the CoM). At this point it is noticeably choppy, when moving the camera around.

Physics ratio is between 0.4-0.5x, timer is steady yellow.

Launch is still choppy, but as it approaches Mach 1, the ratio drops to between 0.3-0.4, even after dropping a few stages. This is when aero and heat are taking effect.

Ratio only gets to 0.6-0.7x even when approaching the edge of the atmosphere and FPS is still choppy. There are less than 20 parts left at this point.

During the entire ascent, the F12 menu shows a steady 25 FPS all the way up, but it clearly is not 25 FPS. It also shows Max and Min FPS to be 25. I don't think it works properly in atmosphere.

The second I hit 70KM and escape the atmosphere (with the the last 2 stages) the FPS counter shoots up to 165 FPS and becomes very smooth for the rest of the flight (didn't reenter atmosphere).

/END OF COMPARISON

I still have yet to install and run the test on 0.25 again with DRE and FAR, but I will, along with 0.90. I also want something to display actual FPS outside of the game itself. I also want to hyperedit the ship into orbit and see how the performance is with a 600-part ship outside of atmosphere. And I want to run the 1.0.4 test again with 0% heating and the old aero which I think can be switched to in the settings config.

Finally, I will retry with all the lowest graphics settings but will keep the resolution the same.

More to come...

[cantab]

Yeah, that's the test rocket I use. There are two big confounding factors with comparing 0.90 with 1.0.4 though. Firstly the rocket will fly differently because of the aerodynamic and engine changes. FAR has changed a lot too since 0.90. Secondly the extended physics range in 1.0.4 is a big part of the extra lag.

[quasarrgames]

This poll and its shape are very interesting. Wonder why so many people said 1.x?

Also, .90 was best for me. 300+ part ships with playable framerates.

Now, it seems like once a ship gets to 100 parts or above, there's a massive wall of lag. Am i the only one who thinks this?

150+ parts now is subtle psychological torture...

[xtoro]

Yeah, that's the test rocket I use. There are two big confounding factors with comparing 0.90 with 1.0.4 though. Firstly the rocket will fly differently because of the aerodynamic and engine changes. FAR has changed a lot too since 0.90. Secondly the extended physics range in 1.0.4 is a big part of the extra lag.

Yeah I'm aware of the added stock aero and heat calculations which is why I mentioned doing a test with the latest version with heat turned off and drag being applied as acceleration vice force to be more like pre-1.x. It may not completely eliminate null calculations, but when comparing it to pre-1.x versions running both DRE and FAR it would be more of a fair comparison.

This poll and its shape are very interesting. Wonder why so many people said 1.x?

Also, .90 was best for me. 300+ part ships with playable framerates.

Now, it seems like once a ship gets to 100 parts or above, there's a massive wall of lag. Am i the only one who thinks this?

150+ parts now is subtle psychological torture...

I think the shape has a lot to do with new players compared to veteran players. In hindsight I should have maybe put 2 options of "1.0.4 but never played pre-1.0 versions" and "1.0.4 and I have played pre-1.0 versions" to make a better assessment... But then I could have also added playing with and without mods, AMD and Intel processors... Really, the possibilities are endless but it's hard to get specific data using a simple poll