Arcania: Gothic 4 Demo
http://www.youtube.com/watch?v=gR2IA6H-LFc
Avg:21.219fps Min-Max:17.5-29.0fps Tear:32.747%(32.259%)
http://www.youtube.com/watch?v=gR2IA6H-LFc
Avg:21.219fps Min-Max:17.5-29.0fps Tear:32.747%(32.259%)
The Framerate Analysis Thread part 2
- Thread starter grandmaster
- Start date
brain_stew
Regular
Hmm, is this the worst performance in any measured PS3 or 360 game so far? Can't recall any other game with such a low average framerate this generation, certainly not one which isn't v-synced. Can anyone think of something worse?
corduroygt
Banned
Gundam Crossfire
brain_stew
Regular
Ah forgot about that, although it was a launch game so I guess it can be given a little slack. Did it use vsync though (so not exactly an apples-to-apples comparison).?
Might be but a comparision should be made with similar scenarios. I wonder though if one reason for the bad perfomance is the shadow quality. If they kept the shadow quality from PC version which is fantastic in quality then that would explain how it even runs at ~25fps at 1:50 into video. Judging from small video but it also looks like a step up graphically vs Risen.
Two Worlds.
Though I haven't seen any kind of performance analysis on that game to know for sure.
IIRC Gundam Senki (not Crossfire) would be the game released this gen with lowest framerate, it literally ran at 15fps....& that's not the average I am talking about. But it did have a rather fantastic IQ.
Gundam Senki? no way, I have that game, and it definitely doesnt run at 15 fps. Crossfire does.
update: this http://zoome.jp/ps360/diary/364/ is senki. Crossfire is the one with good IQ running at 15 fps.
update: this http://zoome.jp/ps360/diary/364/ is senki. Crossfire is the one with good IQ running at 15 fps.
Last edited by a moderator:
do hard v-synced games while under heavy stress run at 20fps? noticed that games like halo reach, mgs4, and dead rising 2 drop from 30fps abruptly to 20fps.
but proper triple buffered games typically can run at any variable frame rate for example god of war, infamous, modern warfare 2
is this typically the case or just a coincidence?
but proper triple buffered games typically can run at any variable frame rate for example god of war, infamous, modern warfare 2
is this typically the case or just a coincidence?
Its not coincidence double buffer+ hard vsync will always give you drop to 20FPS like that.
Also Halo Reach shouldn't be included there since that game is double buffered + soft V sync.
Digital Foundry did an extensive analysis.
corduroygt
Banned
Yes it let the engine run at its opitimal speed.That's the whole point of triple buffering!
Still people are clearly overdoing it, it doesn't prevent the game to drop to 20fps as in the end it's the display refresh rate that rules. If it refreshes every 16.6ms that's it. Display can't refresh @ 28fps/29/whatever because the refresh happens every 16ms no matter what you have is a succession of frame rates 20, 30, 45, 60.
Triple buffering doesn't improve this it just help by not letting the render runs at "full" speed.
corduroygt
Banned
That would only be true if we had games that run at more than 60fps (16.6ms refresh). But we don't in the console world.
You can have any framerate you want between 1 and 60 with triple buffering on a 60hz display without tearing. Some frames will just be displayed more than once if the game runs at less than 60.
TheWretched
Regular
Well... I don't know about the 360 version or if it was done in-house, but Risen and Gothic 4 were done by completely different teams (Risen was Piranha Bytes (Gothic 1-3 devs), whereas Gothic 4 was done by Spellbound).
Either way... waiting for a bug free PC version...
Oops I missread your post.
Clearly no the display doesn't adapt what you see on screen is is succession of 60fps, 45fs, 30fps, etc.
It can't be otherwise, the render on the other hand... But basically say for some frame of a 30fps render that exceed its render time by say 10% so it took actually 36ms to render.
The triple buffered render will continue to render as soon as it is done with the previous frame vs a double buffered one, so there is a serie a rendering heavy frame it takes 10 frames to render 1 frame vs a double buffered one (but two "fresher" frames will be displayed, in my example stretching it a bit after five frames ~3x5ms a "fresher frame will be displayed as it may match display refresh rate which 60fps). The bigger the hit the bigger the advantage for the triple buffered render both in displayed frames and input lag.
Still it doen't change the fact that if rendering time is exceeded frame will be displayed at 20fps (could be less) for some frame. Point is triple buffered render kind of "recovers" quicker as the triple buffered render doesn't lose time.
This has been discussed a lot and I still haven't seen something that convinced me otherwise.
Basically things as fraps are missleading as the measure the speed of the render not what you "see". The same for fps which is an average.
The triple buffering is clearly superior and especially critical when you render is faster than refresh rate, as while avoiding tearing it consistently lowers the input lag.
Last edited by a moderator:
That depends on your sampling frequency to dervie an average. If you count frametime, ms on screen, which is what you seem to be doing, you either have a frame shown for 1 refresh, or 2, or 3, leading to 60, 30 or 15 fps for a frame. But if in a one second period you show 59 individual frames, with one repeated for two refreshes, that's 59 fps. And if you have a mess of 1, 2 and 3 refreshes-perpframe images, you can end up with a very choppy...47 fps, or 31, or 20. So yes, it's possible to have any range of integer fps up to 60 fps on a 60fps display, whether you v-sync with triple buffering or allow tearing. If you allow tearing, you can exceed 60fps, although you'll only see parts of a frame. I dunno what you'd call partial frames. But then if you are exceeding the refresh of your monitor, it makes sense to lock to the refresh rate, and disable v-sync only when you drop below that.
Last edited by a moderator:
Similar threads
