Jump to content

1.1's "massive" performace improvement


Recommended Posts

I first want to start out by saying that i love Kerbal Space Program so much, and with each new update i fell the game gets better and better in ways that amaze me every time I play, which is why i feel screwed over by this issue.

I tried out KSP 1.1 after it came out, getting all excited because of the amazing results in performance that i saw in the twitch streams. Then i started getting into it, and when i checked the performance, it capped at 60-65 FPS in a stock game, just like 1.0.5 (i did a comparison to make sure).

I assumed it was because i wasn't switched to 64-bit because it wasn't showing the (x64) in the version number on startup. I actually made another thread about how to switch to 64 bit on a Mac, and i was told it was a unified binary and to check my activity monitor. So i did, and it said i was on a 64 bit build so i figured my troubles were over.

I tried moving along with the performance issue by installing two mods: scatterer and EVE with configs. when I tested, i was getting 20-30 FPS on a 15 part ship in orbit, and same for a plane in atmosphere as the link shows: Performance Pics

This was not the great performance fixes we were promised, because i see little to no improvement to 1.0.5. Does anyone know how to fix this issue, or is anyone having the same issue?

SPECS:

iMac (Retina 5K, 27-inch, Mid 2015)

Processor: 3.3 GHz Intel Core i5

Memory: 16 GB 1600 MHz DDR3

Storage: 1.11 TB

Graphics: AMD Radeon R9 M290 2048 MB

OS: latest version of OSX El Capitan

Link to comment
Share on other sites

9 minutes ago, justspace103 said:

yeah it says every v-sync blank should i turn it off?

Yes, it will limit your FPS to a portion of your monitors refresh rate in order to reduce "screen tearing." You will get a higher FPS without it but it will fluctuate more.

That said, I use a Radeon card as well, and I haven't seen much if any noticeable improvement in performance in 1.1 either.

Also, I seem to get the same FPS regardless of playing on the lowest settings compared to the highest so it's prolly a CPU issue.

Link to comment
Share on other sites

34 minutes ago, Rocket In My Pocket said:

Yes, it will limit your FPS to a portion of your monitors refresh rate in order to reduce "screen tearing." You will get a higher FPS without it but it will fluctuate more.

That said, I use a Radeon card as well, and I haven't seen much if any noticeable improvement in performance in 1.1 either.

Also, I seem to get the same FPS regardless of playing on the lowest settings compared to the highest so it's prolly a CPU issue.

do i need a higher graphics card or are there settings on my mac that i can change to make performance better?

Link to comment
Share on other sites

8 minutes ago, justspace103 said:

do i need a higher graphics card or are there settings on my mac that i can change to make performance better?

I think your card should be fine, your specs seem plenty good enough for KSP.

Considering it's a laptop, does it have an integrated card? You should check in your graphics cards options and make sure KSP is using the better card.

I'm not very familiar with Macs though so I'm not terribly helpful in that regard.

Link to comment
Share on other sites

22 minutes ago, Rocket In My Pocket said:

I think your card should be fine, your specs seem plenty good enough for KSP.

Considering it's a laptop, does it have an integrated card? You should check in your graphics cards options and make sure KSP is using the better card.

I'm not very familiar with Macs though so I'm not terribly helpful in that regard.

its not a laptop its a desktop but thanks for your help!

Link to comment
Share on other sites

7 hours ago, Rocket In My Pocket said:

Oh, my bad. I saw the IMac 27" bit and jumped to conclusions.

In my experience the 3 most major settings you can turn down to improve frame rate in KSP are:

Pixel light count, Shadow cascades, and Physics delta time.

do you have any suggestions on what my settings should be for those?

Link to comment
Share on other sites

So you saturated your little Radeon mobile GPU not only with the 5K resolution but also by installing mods that add (only) additional GPU (and GPU memory) load and then wonder why you don't see the benefits of better CPU usage capability when specifically using a ultralow-partcount (and thus using as little CPU as possible) ship. Is this a joke thread?

Unity5 can handle a (sort of) CPU multiprocessing, especially when calculating more than one (high part count) ship in the same scene. As KSP is - due to the nature of the game (physics > graphical fidelity) - extreme CPU bound in most situations, especially when docking large ships to large stations, this can result in a HUGE performance boost (as shown by streamers like das valdez). But you created (deliberately?!) one of the few scenarios where KSP can act being 100% GPU bound by combining a mid range mobile GPU (even one with extremely little memory) with an ultra-high range resolution and mods that add additional gpu load.

 

Link to comment
Share on other sites

29 minutes ago, justspace103 said:

do you have any suggestions on what my settings should be for those?

You can turn Shadow cascades down to 1, you prolly won't even notice the difference visually.

Pixel light count I'd set to at least 2, and Physics delta time should be at it's lowest "0.3".

8 minutes ago, InsaneDruid said:

So you saturated your little Radeon mobile GPU not only with the 5K resolution but also by installing mods that add (only) additional GPU (and GPU memory) load and then wonder why you don't see the benefits of better CPU usage capability when specifically using a ultralow-partcount (and thus using as little CPU as possible) ship. Is this a joke thread?

Unity5 can handle a (sort of) CPU multiprocessing, especially when calculating more than one (high part count) ship in the same scene. As KSP is - due to the nature of the game (physics > graphical fidelity) - extreme CPU bound in most situations, especially when docking large ships to large stations, this can result in a HUGE performance boost (as shown by streamers like das valdez). But you created (deliberately?!) one of the few scenarios where KSP can act being 100% GPU bound by combining a mid range mobile GPU (even one with extremely little memory) with an ultra-high range resolution and mods that add additional gpu load.

 

Really helpful, you should go into customer support. Seriously. You have a gift with people.

Link to comment
Share on other sites

Yea sorry for being salty and I can only say that it is absolutely not meant to be personal, but, you know the constellation of hardware (gpu vs monitor) plus the strange setting of adding only mods that only add gpu load (and then on top especially these that add shader AND texture load - again, 2GB mobile with half the memory amount and half the bandwith of a regular 290 card, only 80 TMUs instead of 176, 1280 shader units instead of 2816 @5K) with the unusual 15 part ship. This is a strange constellation especially as every KSP player should have seen that KSP 1.0.5 and below suffered from partcount. There is even a ultra old and widely used mod to cure this issue in the form of the welding tool.

 

EDIT:

  • disable Vsync (or turn on triple buffering - if supportet on a mac) (this will avoid capping fps to integer divisions of your refresh rate (60-30-15))
  • reduce resolution (to free GPU resources)
  • reduce FSAA (to free GPU resources)
  • disable scatterer (to free GPU resources)
  • remember: the most benefits of 1.1 will bee seen with high part count scenes.
Edited by InsaneDruid
Link to comment
Share on other sites

1 hour ago, InsaneDruid said:

Yea sorry for being salty and I can only say that it is absolutely not meant to be personal, but, you know the constellation of hardware (gpu vs monitor) plus the strange setting of adding only mods that only add gpu load (and then on top especially these that add shader AND texture load - again, 2GB mobile with half the memory amount and half the bandwith of a regular 290 card, only 80 TMUs instead of 176, 1280 shader units instead of 2816 @5K) with the unusual 15 part ship. This is a strange constellation especially as every KSP player should have seen that KSP 1.0.5 and below suffered from partcount. There is even a ultra old and widely used mod to cure this issue in the form of the welding tool.

 

EDIT:

  • disable Vsync (or turn on triple buffering - if supportet on a mac) (this will avoid capping fps to integer divisions of your refresh rate (60-30-15))
  • reduce resolution (to free GPU resources)
  • reduce FSAA (to free GPU resources)
  • disable scatterer (to free GPU resources)
  • remember: the most benefits of 1.1 will bee seen with high part count scenes.

thanks so much for the help!!

Link to comment
Share on other sites

I think there is some confusion on measuring performance.

In 1.0.x I was getting an FPS count of very stable 25 fps; yet the game was slow. An ascent to LKO took me about 15 to 20 minutes real-time ( during which 6-8 minutes of ingame time passed). This was due to the physics load, not graphics, and I saw an indication of this in the color of the time-display, which would be red rather often. The game would respond sluggish to input, too. That was expected, though, since I ran some mods. The game was slow, yet video was rendered at a good 25 fps.

 

Now in 1.1.0 I can run SVE and scatterer, and despite physics being strained a bit, the game goes smooth. I can steer in realtime, ascent takes me 6-8 minutes.

The fps counter shows 16 fps when facing Kerbin, physics is showing some stress (I increased delta to 0.06), the time is displayed in yellow. Yet the game runs smooth and I totally can steer in realtime.

 

I do count this as a performance boost!

 

tl;dr: fps is not the only performance indicator.

Link to comment
Share on other sites

15 minutes ago, c_space said:

In 1.0.x I was getting an FPS count of very stable 25 fps; yet the game was slow

Then you almost certainly weren't getting 25 fps.  The in-game fps counter in the debug window does not display rendered frames per real-time second, it displays rendered frames per in-game second.  This means that it bottoms out at 1/(Max Physics Delta Time) due to how the game slows down real time when it can't keep up with the physics simulation, anything lower than that will require a different tool to measure the fps.  I generally use the GCMonitor mod for this but external tools such as Fraps will also give a real-time number.

The biggest confusion about the performance improvements seems to be among people whose game was already GPU limited.  In these cases, you will not see much of an improvement at all and may even see a slowdown due to differences in the graphics shaders being used.  The large performance increases happen where your game is heavily CPU bound which tends to mean having a significantly large part count vessel.  E.g. I have an i7-4770K and GTX560ti and my 1410 part station that got between 2 and 3 fps in 1.0.5 now gets 8.4 fps in 1.1.  A smaller station (just over 300 parts) that managed a little over 20fps in 1.0.5 is now in the high 50s in 1.1.  Those are seriously significant performance increases...

Link to comment
Share on other sites

This thread is quite old. Please consider starting a new thread rather than reviving this one.

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...