Well I voted for “Better FPS and same stutters”
And I will quantify that below,
System specs:
I7-920 @ stock speed
6gb of ram (3 x 2gb in 3 cannel mode)
GF GTX275 (stock speed) with NV 270.51 drivers.
Win 7pro 64bit (running like classic win2k desktop/folders etc I.e no XP/vista/win7 desktop/folders FX)
OS & Steam on separate dedicated HDD.
Running NO pagefile (I haven’t for >6 years, but it has been enabled for shutter testing on its own dedicated HDD but it makes no difference as there is >3gb free RAM that windows will use as VM before paging VM to HDD pagefile)
Game settings:
1280x720 @ 60hz full screen, AA = x2
Modal detail, Damage decals, Building detail, Land detail, Visual effects = Medium settings
Buildings amount, Land shading, Forest = Low
Grass & Roads = On
Shadows, Anti epilepsy filter = Off
Quantification:
Test1.trk
Pre & post patch frame rate info was gathered with FRAPS using a 160sec benchmark using a repeatable .trk file of a decent from 2200ft above RAF Ford to land at RAF Tangmear (200ft) with 8 other planes circling to land, this .trk has never exhibited any shuttering either when originally flying it or subsequent replays.
As above game settings:
Pre patch = 8609 frames, 28 Min, 79 Max, 53.806 Avg
Post patch = 9946 frames, 39 Min, 78 Max, 62.163 Avg
As above game settings except for 1600x900 @ 60hz full screen, AA = off:
Pre patch = 7614 frames, 26 Min, 62 Max, 47.588 Avg
Post patch = 7891 frames, 31 Min, 61 Max, 49.319 Avg
Test2.trk (perceived shutter testing with buildings)
Pre & post patch frame rate info was gathered with FRAPS using a 160sec benchmark using a repeatable .trk file of a decent from ~2500ft to low level over a larger town which is over flown twice in opposite directions
Buildings in numbers larger than a small village are about the only cause of shutter for me when flying low, given frame rate is independent of “shutter” for me, with buildings the increase in frame rate matches similar % increases to the minimums & average frame rates collected above for pre and post patch though many more permutations of building detail level & building amount were tested with the aim of perceiving shutter.
A: With Building detail = Low & Buildings amount = Low Vs. With Building detail = High & Buildings amount = Low, has minimal impact on frame rate and the same amount of shutter take place
B: With Building detail = Low & Buildings amount = unlimited Vs. With Building detail = High & Buildings amount = unlimited, has minimal impact on frame rate and the same amount of shutter take place, the frame rate is slightly lower than A: above (~5fps lower but we are still talking >30fps) and the shutter has a slightly different nuance to it and could be said to shutter fractionally less than A: above especially when over the centre of the town as there is a perceived smooth spot ware all the buildings are in view so have loaded.
To me the problem with buildings is threefold, 1st: the mechanism in how they are loaded into the game world and when. 2nd: the amount of LOD’s used for “flying” situations when there is a lot of irrelevant detail being shown for flying at even at just above roof height but at high speed. 3rd: how setting Building detail = Low only serves to reduce the distance each LOD will be seen but NOT also the highest detailed LOD that can be seen.
So e.g.
on Building detail = Low with Buildings amount = Low, you get a performance hit from the quantity of buildings loading and unloading rapidly from view in a short space of time, you get a performance hit from successive LOD’s loading rapidly and a performance hit from the detail of the highest detail LOD being shown.
|