Frame Rate: A Short Guide for H2A

Discussion in 'Halo and Forge Discussion' started by SecretSchnitzel, Nov 19, 2014.

By SecretSchnitzel on Nov 19, 2014 at 9:13 AM
  1. SecretSchnitzel

    SecretSchnitzel Donald Trump
    Forge Critic Senior Member Temporarily Banned

    Messages:
    2,433
    Likes Received:
    1,882
    Frame Rate Issues

    Halo TMCC offers an amazing retooling of forge this time around and brings many great new things to the table, such as scripting, and has brought back functions that were dearly missed in Halo 4, such as precision editing. It also features, as expected, a brand new palette and three unique canvases. Unfortunately, all is not perfect. While the functions work excellently (save a few small discrepancies such as magnets), there are major issues with the palette itself. Many experienced forgers that have been around since Reach and Halo 4 are finding problems arising in what is referred to as “frame rate lag”. In this, the objects on the screen are causing the game’s performance to drop, resulting in a stuttered appearance in movement on the screen from the perspective of the player. This is an issue because it results in a loss of fidelity in precision aiming on the part of the player, and thus is to be avoided at all costs. Yet, how can one avoid “frame rate lag”? This article will be brief in explaining what causes it and how best to avoid it.

    Frame rate lag is caused when the engine struggles to render the various objects on the screen. This can be due to too many objects being on the screen at one (object density in field of view), or because of particularly taxing objects being on the screen. What are particularly taxing objects? Objects whose models are made up of significantly high complex polygon counts. Know that “pipe” on the bottom of many bridge pieces? That’s a complex polygon model. Now polygons are really nothing more than basic shapes that are used to make a more complex shape, and the more complex the shape is, the more polygons are required. A simple block piece in this game features a very low polygon count relative to more complex objects such as bridges, as it is essentially a 6-facet 3-dimensional shape. In that regard, blocks are not very taxing on the game’s engine to render. In fact, the blocks are probably one of the friendliest pieces for the engine to render.

    Alright, so what does all this mean? Well, the combination of object density and object complexity lend themselves to frame rate lag, so the individual forger should try to avoid them. This can be accomplished through using less complex objects and avoiding excessive merging of objects together. To tell if an object is complex or not, all the forger needs to do is zoom in and examine the object looking for 3-dimensional designs on the object. The flatter it is (such as a block), the less complex it is. The more 3-dimensional the “texture” looks (like the pipes and various cuts in the bridge pieces), the more complex it is.

    In short, try to keep your object usage simple. If you start experiencing frame rate lag, try swapping out the more complex pieces for similarly sized blocks. It’s a matter of trial and error really and you shouldn’t let frame rate lag hinder your design ambitions. In closing, I’ll leave you with a short list of objects that have been particularly bad for causing frame rate lag: bridge xl, brace large, and ramps (when used in excess for walling).

    Happy forging!
     
    #1 SecretSchnitzel, Nov 19, 2014
    Last edited: Nov 21, 2014
Tags: this article has not been tagged

Comments

Discussion in 'Halo and Forge Discussion' started by SecretSchnitzel, Nov 19, 2014.

    1. A Haunted Army
      A Haunted Army
      what? no mention of glass having to render an additional local effect and reflectiveness that can also hit performance hard? the glass pieces are fine..... when you can't see the map through them.

      now, where is that sticky option because this is important. i've noticed people on other forums struggling to pin point this even though it's pretty obvious.
    2. SecretSchnitzel
      SecretSchnitzel
      Glass and lights really aren't a big issue this time around... They can be if they're near each other, but really, they work fine.
    3. purely fat
      purely fat
      One thing I noticed with lights is they seem to effect the frame rate more when they are placed in the middle of the map as opposed to edges. Most likely has to do with object density when looking at the middle. I am also starting to think that larger scale objects might cause some issues as well. I have cut down object density on maps by using larger blocks, grass plains, etc. and have found little improvement in frame rate in some cases and none at all in others. The thing I have noticed is the most effective is keeping object count below 300. I think I am going to make a couple of 1v1's that only use one object to test specific pieces direct effect on frame drops.

      Good piece btw.
      Chronmeister and SecretSchnitzel like this.
    4. Goat
      Goat
      this post was done by a TurbTastic on another site:

      "I put together this rough guide so that we can try to gauge which pieces are taxing our forge maps the most. I tested some common pieces by placing the maximum amount of them and timing the lighting generation process. Feel free to reply with other interesting results. See the results below:

      Testing was all done on the Awash skybox.

      200 Wall Coliseums = 28 SECONDS
      200 Cylinder large = 21 SECONDS
      200 Walkway large = 16 SECONDS
      200 Brace large = 11 seconds
      100 Flat large terrain = 10 seconds
      100 Cliff large terrain = 8 seconds
      200 Ramp circular large = 7 seconds
      100 Room triple = 6 seconds
      200 Dish = 6 seconds
      200 Window Coliseums = 6 seconds
      200 Platform Xlarge = 6 seconds
      200 10x10 Blocks = 4 seconds
      200 Ramp stunt = 3 seconds
      200 Seastack rocks = 3 seconds

      200 Bridge Xlarge = Little to no delay
      200 Landing pad = Little to no delay
      200 Platform Large = Little to no delay
      200 Tunnel long = Little to no delay
      200 Cover glass = Little to no delay
      100 Container open large = No delay
      100 Grass planes = No delay
      200 Glass panel 3x3 = No delay
      64 Heavy man cannon = No delay
      64 Tin cups = No delay
      32 Light red = No delay
      200 Tree 01 large = No delay, but obvious effect on performance meter"

      Am sure it isn't terribly accurate, but have found similar results with most of the items in the first list. I've found that terrain and "tan" rocks tend to be most taxing on the engine (and causes huge drops for anyone else in your forge session).
      Last edited: Nov 21, 2014
      SecretSchnitzel likes this.
    5. SecretSchnitzel
      SecretSchnitzel
      I don't think the time to render lighting directly correlates to causing frame rate lag. It seems more to be based more on object size than complexity of the object. Of course large objects that are also polygon heavy are going to be taxing on both regards, but small objects like Bridge XL are horrible for frame rate but have next to no lighting delay.

      Now why would a large object cause lighting longer to render? Easy, the game has to process where the lighting is blocked by the object. The more lighting that's blocked, the longer it'll take to render.
    6. oxitran
      oxitran
      Still we need a testable method, anyone got any ideas?
    7. SecretSchnitzel
      SecretSchnitzel
      Yeah, trial and error. If map has frame rate, delete pieces you think might be the problem and try something else. Keep note of what pieces turn out to be problem pieces.

      There is no empirical testing method that can measure which pieces are the worst.
    8. Zombievillan
      Zombievillan
      Yeah hopefully someone breaks this down into great detail. I hate to put it off on others but it's gonna be a few weeks until I have an Xbox one so I'm not gonna lie, it'd be pretty sweet to have a guide already when u start forging again.
      SecretSchnitzel likes this.
    9. SecretSchnitzel
      SecretSchnitzel
      It's hard to test empirically (free of other pieces influencing it) to get an accurate assessment of which pieces are the worst. I'm building my assumptions based off pieces I've noticed as problematic and the traits they share off of gameplay, as well as the distinct similarity the problem has to Reach frame rate issues.

      Welcome back Zombie.
    10. Zombievillan
      Zombievillan
      Good to be back, thanks. Guess it will take time. Same thing happened with Halo 4 forge I just hope that once mcc gets going strong it holds up longer than Halo 4 did.
    11. SecretSchnitzel
      SecretSchnitzel
      All said and done, H2A has the best frame rate performance out of it, H4 and Reach. It's frustrating that there are still glaring issues with it, but I'd take H2A's forge over the other two a thousand times over. Clean textures, beautiful stone blocks, and better object counts. I'm happy. Coulda been heaps better, but this will do for now. With h5 in a year, this is just a pass time to enjoy before the next forge (hopefully) blows our socks off.
    12. the_suicide_fox
      the_suicide_fox
      I've discovered that lights + glass = framerate issues. At one point, my Disco Slip map had lights instead of shields on the walls and I was having serious performance issues (bar was dropping below the P in Performanace). At first I thought it was the dancefloor, but even after deleting every shield it was still having the same problem. I found that putting lights near glass pieces (or in them as I had) was the root cause and now I have very little performance issues despite all the shields.
      a Chunk likes this.
    13. leegeorgeton
      leegeorgeton
      I've loved all of the versions of forge except for 3 which I didn't use, but I was really expecting that there wouldn't be any frame rate issues with this new one, particularly if I was just remaking a map from one of the previous forges. I'd expect if you're running a re-tooled version of Halo 4's forge on a machine that's 8 or so years advanced in technology compared the machine that was already running Halo 4's forge, that the new machine would be able to easily handle the task.

      The new Halo 2 Anniversary Forge does look pretty, but not significantly better than Halo 4's forge. I'm just curious what it is about this very similar-seeming forge that is bogging down this brand new Xbox One.

      Is there something added to this forge that I'm not seeing? Is there that much more under the hood?
      Chronmeister likes this.
    14. Hulter
      Hulter
      Bring back the Reach brace large, Bungie.
    15. SecretSchnitzel
      SecretSchnitzel
      Bring back that style of aesthetic. I miss the clean forerunner aesthetics... Make the pieces "flatter" (not stupid polygon intensive models) and give us a color shared, and of course a better budget, and I'll be fine as rain.
    16. leegeorgeton
      leegeorgeton
      Is any one of the three forge base maps preferable to the others as far as frame rate issues are concerned? I'm wondering if the cloud/fog in skyward causes any issues. I'm just a bit annoyed that I've rebuilt a map from Halo 4 and there are frame rate issues that weren't in Halo 4 even though I've significantly simplified the structural pieces used.
    17. Maximus IL
      Maximus IL
      Complex-appearing pieces don't always cause framerate issues. For example, you can load up a map with a bunch of pyramids without causing issues, but covering the same area with landing pads will cause issues. Here's a list I came up with by trial-and-error on a rather complex map I'm still building (mostly still attempting to reduce frame drops). Lots of these are familiar from H4, but there are a few unusual suspects in there like Coliseum Walls and Landing Pads.

      Bad for framerate:

      Cliffs (all of them)
      Grass planes
      Flats (all of them)
      Walkway
      Coliseum Walls
      Coliseum Windows
      Brace Large
      Landing Pads
      Platform XL (platform large seems to be okay)
      Bridge XL (other sizes seem okay)
      Glass (any kind - but especially in the presence of lights, as mentioned above)
      Buildings (when using many for floors/walls)
      Teleporter frames
      Lights


      Possibly bad for framerate:

      UNSC Man Cannon
      Forerunner Gravity Lift
      Brown Rocks
      Large Y-Crosses
      10x10 Blocks (though when covering large areas, seem better than bunches of small pieces; however, when just having a small portion of them showing, using 2 - 3 smaller pieces shows improvement)


      Easy on framerate:

      Blocks (seemingly all of them except maybe the 10x10s)
      Circular Wall
      Wall
      Double Wall
      Y-Cross
      Windows / Corner Walls (without glass)
      Pyramid
      Platform Large
      Trim
      Walkway Cover
      Strut
      Black Rocks

      I haven't found too many issues with shields.


      Part of the issue with the larger pieces may not be the polygon count but rather the resolution. If Forge attempts to always render a particular object in a uniform resolution, then even simple objects (geometrically speaking) could cause some issues. A larger number of smaller objects may be less taxing because it can downgrade the resolution on the furthest ones, while it may have to attempt to render a grass plane in full resolution for the entire distance. I did a test where I laid out a line of 3 grass planes next to my map (so there were already objects present to tax the engine) and saved it. Then I recreated the same surface area in the same spot using 27 10x10 blocks. Despite there being 9 times the number of blocks used, performance was better with the blocks than with the grass planes.

      For small objects, the general rule that frame drops follow complexity seems to hold true. Landing pads are a notable exception.

      I've also had some luck closing gaps by merging pieces. On the pyramids there's a hole in the geometry where the polygons from the two sloping sides of the tip piece don't actually connect (you can see it if you find the right angle to look at it). It seems as if Forge sees that gap and attempts to render everything beyond it as if the more distant objects were in the field of view. But when I closed the gap by burying the edges of both planes into a trim piece, the performance improved noticeably. I don't know how many of the objects have gaps that prevent them from being used for cutting down the number of objects in the FoV (which only works to an extent anyway), but I imagine there are more. I have been using significant overlaps (0.2 units total) wherever feasible to reduce the number of times gaps cause problems.
      Last edited: Jan 28, 2015
    18. Maximus IL
      Maximus IL
      The difference is partially due to resolution. H4 ran 720p. H2A runs 1328x1080 - or 1.43 million pixels - while H4 ran at 720p with 0.92 million pixels. H2A also attempts to run at 60fps as opposed to 30fps, which means H2A has to process the larger number of pixels at twice the rate to avoid a drop. In the MCC version of H4 it's an even larger change, as MCC H4 runs at 1080p (2.1 million pixels). Where you just to run H4 forge (720p, 30fps) on the Xbox One, it would probably do beautifully all the time.

      So where the starting conditions resolution and framerate wise equal, I'm quite sure Xbone would blow away the 360 in performance. However, given the resolution adjustment, the Xbone has to work 3x as hard on H2A and 4x as hard on H4 to avoid frame drops.
      leegeorgeton likes this.
    19. Goat
      Goat
      wait, so platform XLs are okay and it's the larger platform cylinders that are the problem?

      will 12ish teleporter frames beside one another be bad for framerate?

Share This Page