View Issue Details
| ID | Project | Category | Date Submitted | Last Update | |
|---|---|---|---|---|---|
| 0003903 | Valley 1 | Note To Test | Sep 22, 2011 11:16 pm | Sep 23, 2011 10:08 am | |
| Reporter | Admiral | Assigned To | Chris_McElligottPark | ||
| Status | closed | Resolution | not fixable | ||
| Product Version | 0.420 | ||||
| Summary | 0003903: Windowed performance is much lower than fullscreen | ||||
| Description | To facilitate double-windowing with Mantis and AVWW I switched from fullscreen (2560x1600 on one of the two monitors) to windowed (2500x1500) so it wouldn't minimize every time I switched. The game felt very choppy once I switched. So, I booted up FRAPS and did a few minor tests. I had a consistent 58-60 FPS (vsync enabled) in fullscreen. Even just sitting in a static scene in windowed, pushing fewer pixels, I was getting between 44-50 FPS. Sometimes for no discernable reason FRAPS would drop down to the teens. Now, I realize windowed FPS is generally lower, but I thought such differences were minimal with modern OS (Win 7 x64) and GPU (AMD 5870) combos. | ||||
| Tags | No tags attached. | ||||
| Internal Weight | New | ||||
|
|
Going to the travel window is an almost complete bust. I'm getting 10 FPS and the background is super choppy. |
|
|
Out of curiosity, what do you see in similar tests on AIW? Same basic graphics engine, though Chris has added a ton of new tricks in AVWW. |
|
|
10 on the world map? Just tested it on my midrange laptop with an integrated graphics card and got 20 FPS on the world map (with very little space not taken up by regions). Can you attach your world directory? |
|
|
Yep, there should be a minimal difference, and I routinely get 350fps in the Unity editor, which is windowed (the actual compiled game is capped at 60fps). This isn't anything we can do anything about, though, as it's most likely at the driver level or something to do with whatever else is going on with the OS at the time, or VERY possibly something with the unity engine but I suspect they are actually off the hook on this one, too, as no-one else is reporting it on this or our other games. |
|
|
I can attach the world directory if you like, but sounds like Chris isn't convinced it's worth checking into. |
|
|
I'd like to run it by the profiler and see if anything obvious pops out. Not the cause of the windowed/fullscreen thing, certainly, but perhaps some inefficiency in my world map related stuff. |
|
|
Well, I can tell you this: our game has no power over anything to do with windowing. So the best I could do is log a bug report to unity, and let them ignore it for the next year. Given this hasn't shown up on any other machine, I'm thinking it's something local to your setup. I mean, our literal control over the windowing consists of: 1) check if it windowed; 2) tell it to be windowed or not. That's one thing that annoys me greatly with unity, because I used to be able to do many things much more directly with the old AI War engine. |
|
|
Keith, the world map in general is hugely inefficient, I can tell you. I know exactly what it is: 1. All the text, specifically the region numbers. 2. All the borders between the super-regions. 3. To some extent, on very large screens, all the various regions and their overlays doesn't help, but this should mostly be no worse than an actual side-view section. 4. All the various icons, although this also is like 3 and isn't as big a deal as 1 and 2. I started on the code to address 1 today, and that will ultimately lead to a solution for 2 as well, but it's slow going because it's a really radically different and bare-metal sort of thing compared to the basic unity stuff. |
|
|
I meant cpu related stuff, not graphics. I know the graphics are horribly inefficient right now because of the incessant texture swaps and whatnot with all that. |
|
|
And draw call count, yeah. |
|
|
|
|
|
I did upload the world file in case you really want to investigate it. |
|
|
Couple other notes: 1. 0.421 is now out, and won't be drawing the sky behind the various maps anymore. It was both visually distracting as well as a real performance drain on those screens. That ought to help, but to what degree I don't know. 2. You might also try changing to static skies while in windowed mode just to see what happens. The dynamic skies do millions of calculations on your GPU, to procedurally create and animate those clouds (thanks to the awesome Unisky). I've put in some changes of my own to help improve the performance of Unisky for our purposes here, and there's a Sky Detail (or something like that) in the settings window that you can use to tune this down even further if you want. All that said, the only reason it should differ in windowed versus fullscreen is still if the OS or driver is either doing something wrong, or really busy with something else. I don't use the Aero interface in Windows 7, for instance, so that might be part of it. There's really nothing we can do either way, but static skies drastically reduces the GPU load in general of the game (something like 90% of the GPU load comes from the dynamic clouds), so if you're having troubles there that would be one way to solve it. The static skies aren't as pretty, sadly, but they aren't too bad. :( And with that I'm off for the night -- more tomorrow! |
|
|
Night :) (Note: if you read this tonight, you're in deep trouble.) |
|
|
Deep trouble it is. But now I'm going, for real. The baby wakes up at 6am either way, so I know how much sleep I'm (not) going to get if I push things, heh. |
|
|
0.421 without the clouds (and no modifications to my system) now has the travel map moving at 20 FPS. I would strongly suggest putting in a subtle non-black static background. |
|
|
Yea, looking at the initial map is now tons less visually interesting, though I use static skies and perhaps that's why I didn't feel the sky was distracting. And in theory a mature game-state is probably not going to have uncovered spaces a lot of the time... but I still want the sky back :) I can put in the toggle myself :) |
|
|
Iggg... ok. Make sure it's defaulting to off, then. |
| Date Modified | Username | Field | Change |
|---|---|---|---|
| Sep 22, 2011 11:16 pm | Admiral | New Issue | |
| Sep 22, 2011 11:19 pm | Admiral | Note Added: 0013386 | |
| Sep 22, 2011 11:19 pm | keith.lamothe | Note Added: 0013387 | |
| Sep 22, 2011 11:25 pm | keith.lamothe | Note Added: 0013389 | |
| Sep 22, 2011 11:30 pm | Chris_McElligottPark | Internal Weight | => New |
| Sep 22, 2011 11:30 pm | Chris_McElligottPark | Note Added: 0013393 | |
| Sep 22, 2011 11:30 pm | Chris_McElligottPark | Status | new => closed |
| Sep 22, 2011 11:30 pm | Chris_McElligottPark | Assigned To | => Chris_McElligottPark |
| Sep 22, 2011 11:30 pm | Chris_McElligottPark | Resolution | open => not fixable |
| Sep 22, 2011 11:36 pm | Admiral | Note Added: 0013396 | |
| Sep 22, 2011 11:37 pm | keith.lamothe | Note Added: 0013399 | |
| Sep 22, 2011 11:39 pm | Chris_McElligottPark | Note Added: 0013400 | |
| Sep 22, 2011 11:41 pm | Chris_McElligottPark | Note Added: 0013401 | |
| Sep 22, 2011 11:44 pm | keith.lamothe | Note Added: 0013407 | |
| Sep 22, 2011 11:46 pm | Chris_McElligottPark | Note Added: 0013408 | |
| Sep 22, 2011 11:46 pm | Chris_McElligottPark | Note Edited: 0013401 | |
| Sep 22, 2011 11:52 pm | Admiral | File Added: Doug 2 - Mantis bug 3903.zip | |
| Sep 22, 2011 11:52 pm | Admiral | Note Added: 0013409 | |
| Sep 23, 2011 12:18 am | Chris_McElligottPark | Note Added: 0013417 | |
| Sep 23, 2011 12:26 am | Admiral | Note Added: 0013420 | |
| Sep 23, 2011 12:27 am | Chris_McElligottPark | Note Added: 0013421 | |
| Sep 23, 2011 12:37 am | Admiral | Note Added: 0013423 | |
| Sep 23, 2011 10:05 am | keith.lamothe | Note Added: 0013435 | |
| Sep 23, 2011 10:08 am | Chris_McElligottPark | Note Added: 0013437 |
